Public Disclosure Authorized
Policy Research Working Paper
Public Disclosure Authorized
5253
Measuring Financial Access around the World
Jake Kendall
Nataliya Mylenko
Alejandro Ponce
Public Disclosure Authorized
Public Disclosure Authorized
WPS5253
he World Bank
Financial and Private Sector Development
Financial Access Team
March 2010
Policy Research Working Paper 5253
Abstract
his paper introduces a new set of inancial access
indicators for 139 countries across the globe and
describes the results of a preliminary analysis of this data
set. he new data set builds on previous work using a
similar methodology. he new data set features broader
country coverage and greater disaggregation by type of
inancial product and by type of institution supplying the
product—commercial banks, specialized state run savings
and development banks, banks with mutual ownership
structure (such as cooperatives), and microinance
institutions. he authors use the data set to conduct a
rough estimation of the number of bank accounts in
the world (6.2 billion) as well as the number of banked
and unbanked individuals. In developed countries,
they estimate 3.2 accounts per adult and 81 percent of
adults banked. By contrast, in developing countries,
they estimate only 0.9 accounts per adult and 28 percent
banked. In regression analysis, they ind that measures
of development and physical infrastructure are positively
associated with the indicators of deposit account, loan,
and branch penetration.
his paper—a product of the Financial Access Team in Consultative Group to Assist the Poor, Financial and Private
Sector Development—is part of a larger efort in the department to improve measurement of access to inancial services.
Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. he author may be contacted
at nmylenko@worldbank.org.
he Policy Research Working Paper Series disseminates the indings of work in progress to encourage the exchange of ideas about development
issues. An objective of the series is to get the indings out quickly, even if the presentations are less than fully polished. he papers carry the
names of the authors and should be cited accordingly. he indings, interpretations, and conclusions expressed in this paper are entirely those
of the authors. hey do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and
its ailiated organizations, or those of the Executive Directors of the World Bank or the governments they represent.
Produced by the Research Support Team
Measuring Financial Access around the World*
Jake Kendall
Nataliya Mylenko
*
Alejandro Ponce
The World Bank / CGAP; we are grateful to Amrote Abdella, Maximilien Heimann, Yehia Houry, Joana Pascual
and Valentina Saltane for research assistance. This paper’s findings, interpretations, and conclusions are entirely
those of the authors and do not necessarily represent the views of CGAP, CGAP’s Council of Governors or
Executive Committee, the World Bank, its Board of Executive Directors, or the countries they represent. Data
available at: http://www.cgap.org/financialindicators
1
1. Introduction
Financial inclusion—providing access to financial services for the poor—has gained increasing
prominence in the past few years as a policy objective for national level policy makers,
multilateral institutions, and others in the development field. The United Nations designated
2005 the International Year of Microcredit, adopting the goal of building inclusive financial
systems, and most other development institutions and multilateral donors have financial access
on their agenda. Consequently, both private and public funds are flowing to fund various
financial inclusion initiatives around the world.1
The rise of financial inclusion as an important policy goal is due in part to mounting evidence
that access to financial products can make a positive difference in the lives of the poor. From the
field, the evidence comes in the form of rapid take-up of financial services when they are made
available to poor households and the very high rates of repayment that the poor exhibit in order
to maintain access. The results of the Financial Diaries Project; summarized in the recent book
“Portfolios of the Poor”, by Collins, Morduch, Rutherford, and Ruthven (2009); show how
dependent the poor are on various financial instruments, both informal and formal, to manage
what little money they have on a day to day basis. And – though the results do not always
support preconceived notions prevalent in the microfinance community – an increasing number
of academic studies show that granting the poor access to financial services can make a
difference in their lives in various ways [see, e.g. Burgess and Pande (2005), Karlan and Zinman
(2005, 2009), Dupas and Robinson (2008), Banerjee, Duflos, Glennerster, and Kinnan (2009),
Bruhn and Love (2009)].2 In short, as a policy goal, developing more inclusive financial systems
will continue to hold a place on the policy agenda.
1
For example, in 2008, there was over $11Bn in outstanding investments in MFIs, nearly 50% of which were
private funds.
2
For example, three recent randomized control trial studies do not support the vision of microfinance’s main goal
being to lend to those poor wanting to start or expand a business. Instead, they find that access to consumer credit
can also improve welfare by (e.g.) keeping people in their jobs (Karlan and Zinman (2005)), that access to
microenterprise oriented credit is fungible with other forms of household debt and is often used to improve risk
management rather than to invest in a business (Karlan and Zinman (2009)), and that the presence of a new MFI in a
neighborhood has no impact on consumption or health and education spending of micro-entrepreneur households
15-18 months later, though it does seem to improve households ability to borrow, invest, and create and expand
businesses.
A necessary step towards achieving an inclusive financial system is to evaluate its status in each
country. To assist policymakers in designing effective policies and tracking progress in the area
of financial inclusion at global level, this paper introduces a new set of financial access
indicators for 139 countries across the globe, and describes the results of a preliminary analysis
of this data set.
This data collection effort has its roots in two previous projects carried out at the World Bank.
The first set of similar indicators of financial access was collected in 2003 for 99 countries (see
Beck, Demirgüç-Kunt, and Martinez-Peria (2007)). In this initial effort, data were collected on
number of loans, deposit accounts, bank branches, and ATMs associated with deposit money
banks (as defined by the IMF). These indicators were updated for a select set of 54 countries in
2007 with the “Banking the Poor” report (World Bank (2008)) and augmented with a set of
survey questions regarding various regulatory features and policy initiatives present in the
country as well as a survey of the 5 largest banks in the country by total assets which collected
data on banks’ fees and the procedures clients had to go through to access loans or deposit
accounts.
The current set of indicators builds on these two previous works using a similar methodology to
collect new data through a survey of financial regulators in 139 countries. In addition to being
more recent and having broader country coverage, this new database features finer
disaggregation by type of financial product and by categories of the financial institutions
supplying such products – commercial banks, specialized state run savings and development
banks, banks with mutual ownership structure (e.g. cooperatives), and microfinance institutions.
Next, we use this data set to conduct a scoping exercise to estimate the number and distribution
of bank accounts worldwide as well as the number of banked and un-banked individuals. Our
estimates suggest that there are approximately 6.2 billion deposit accounts in the world, more
than one for each adult on the planet. However, these accounts are not evenly distributed. We
estimate that around 160 million adults in developed countries have no bank account (or 19% of
all adults) whereas somewhere near 2.7 billion adults (or 72%) of the adults in developing
3
countries are un-banked. While these figures are back-of-the envelope calculations, they give a
sense of the scale of the problem of delivering financial services to the poor.3
Finally, we conduct a preliminary analysis of the data and focus on the national level factors
associated with greater deposit account and loan penetration (measured as number of accounts or
loans per 1000 adults) and of bank branch density (both relative to population and geographic
area). We conduct basic cross-country regressions, controlling for income and population
density, two of the best predictors of the penetration of deposit and loan products in the
population and of the density of bank branching.4 Consistent with previous literature, we find
that controlling for these two factors, the best predictor of deposit account and loan penetration
as well as branch penetration are measures of the development of physical infrastructure
including electricity consumption and phone line density. Additionally, lower inflation and the
presence of explicit deposit insurance are associated with greater deposit penetration. On the
lending side, higher concentration in the banking sector is significantly associated with lower
loan penetration, while measures of creditor rights and creditor information availability are
positively associated with penetration. Finally, none of the policies our survey respondents
reported employing to boost financial inclusion (such as mandates for basic “no-frills” bank
accounts, or allowing for agent-based banking) show any significant relationship with our
measures of loan and deposit penetration. This result however, does not imply that these policies
are ineffective, but rather an indicator of the large variations the countries pursuing them.5
Further analysis, including micro-level studies, is essential to evaluate the effectiveness of
measures aiming to improve access to financial services.
3
Additionally, these figures do not necessarily equate to percentage of households with bank accounts.
Our hypothesis is that population density is a proxy for the profitability of bank branching both because more
customers can be reached per branch in dense areas, and because infrastructure and other services are often more
available in dense areas. When both bank branch density and population are included, branch density is highly
significant, and the coefficient on population density becomes negative. We use population density because branch
data are not available everywhere in all countries.
5
Cross-country regressions are not appropriate for measuring the impact of specific policies, as they rarely capture
changes in the variables over time. For example, the coefficients of a cross-country regression of a certain policy on
the number of accounts are likely to be identified by the differences between countries with a large number of
accounts and no policies in place (perhaps because those countries do not need them) and countries with few
accounts but recent policies addressing this issue.
4
4
The paper is structured as follows: Section 2 describes the collection of the data and the survey
design, Section 3 gives some facts describing the patterns of bank and non-bank supervision
around the world, and the related patterns of data availability for different elements from the
survey, Section 4 describes the procedure for estimating the number of accounts world wide and
the number of banked and unbanked individuals, Section 5 contains the results of the regressions
with the correlates of deposit, loan, and bank branch penetration. Section 6 concludes.
2. Description of Data Collection Methodology
The primary purpose of the Financial Access Survey was to assemble a dataset of measures of
the breadth of usage of the most basic financial products including deposit accounts, loans, and
payments. A secondary goal was to measure the pervasiveness of physical access points such as
bank branches and ATMs which are necessary elements to facilitate broad inclusion. Finally, the
survey attempts to collect information on some of the policies and practices in each country
which might affect financial inclusion.
It is important to distinguish between usage of financial services, which is what we can collect
data on, and the broader concept of access to finance. A widely cited definition of financial
access is that outlined by Claessens (2006) who breaks the population into three parts, those who
have access and use it (group A, included), those who have access but don’t use it (group B,
voluntarily excluded), and those who don’t have access (group C, involuntarily excluded) where
“access” is defined equal A + B. He contends that while data on the usage of financial products
measures A, B is almost impossible to measure, which highlights the difficulty in finding true
measures of “access to finance”, so defined. While in concept, not being able to measure the
voluntarily excluded is a major issue, in reality, there are very few households anywhere which
live in true voluntary financial autarky (though they may choose only informal financial options
if formal ones are not readily available).6 The real issue is that usage data is only widely
6
In Portfolios of the Poor, by Collins, Morduch, Rutherford, and Ruthven (2009), the authors show that all 250 of
the very poor slum residents they study have some form of debt and savings and none use fewer than 4 types of
instruments (be they formal or informal) throughout the year. The true picture is one where poor households are
5
available in terms of the volume or number of loans or deposit accounts, but not in terms of the
number of individuals who own them and thus does not give a clear picture of what percentage
of households use formal financial services.7 The Financial Access Survey focuses on measuring
the usage of a set of basic financial products from various types of formal intermediaries. We
confirm in Section 3 that the measures we collect correlate strongly with the few available data
points from household surveys that measure the number of individuals using similar products.
To achieve the goal of measuring usage of financial services in the world, access to finance
questionnaires were sent to 144 countries around the world: 13 countries in East Asia and
Pacific, 27 countries in Europe and Central Asia, 23 countries in high income OECD region, 21
countries in Latin America and Caribbean, 14 countries in Middle East and North Africa, 6
countries in South Asia and 40 countries in Sub-Saharan Africa. For practical reasons, most of
the small islands and countries at war were excluded from the sample. The survey was directed
to the central bank governor’s office or head office of the financial regulator for approval and
assignment to a contact person, often in the statistics department, who would be responsible for
gathering the appropriate information. A response was received from 139 countries representing
90% of the world’s population and 97% of world GDP. For more detail on the coverage of
individual data elements, see Section 4.
2.1 Description of Survey
Data collection was implemented through a regulators’ questionnaire. The questionnaire consists
of two main parts: statistical questions and regulatory questions. The statistical part of the survey
collects data on the numbers and volumes of deposit accounts, loans, banking infrastructure (e.g.
branches and ATMs) and other measures the usage of financial services usage. Table 1 has a list
continuously substituting between a variety of formal, semi-formal, and informal financial products based on
availability, product features, pricing, and other non-price barriers.
7
This implies a certain rate of double counting since individuals can – and regularly do – own more than one deposit
account for instance.
6
of the variables we surveyed in each of three categories. Except for ATMs, debit cards, and POS
terminals, all elements were asked separately for each institutional category.8
Since many low-income individuals get deposit services through institutions which are not
commercial banks (henceforth, we refer to bank-like institutions which are not commercial banks
as non-bank financial institutions or non-banks for short), we ask for most of the data
disaggregated into commercial banks as well as a number of non-bank categories.9 This addition
represents a major advantage of this study over others that have come before it. One difficulty in
implementing this approach is that there are many types of financial institutions and each can be
different from country to country. India, for instance, has commercial banks, Area Banks, Rural
Banks, various types of Urban and Rural Cooperatives, and many other types of bank-like
financial institutions, each of which differs in subtle ways from similar institutions in other
countries.10 To make the cross-country comparison of the different types possible and
meaningful, regulated institutions are divided into four main categories and a catchall “other”
category:
1) Commercial banks - banks with a full banking license. In some countries, majority
government/state owned banks are included in this category to the extent that they perform a
broad set of commercial banking functions rather than a specialized development role.
2) Institutions with a mutual ownership structure (or mutually owned financial institutions:
MOFIs), such as cooperatives, credit unions or mutual banks.
3) Specialized state owned financial institutions (SSFIs) or extensions of the government whose
main purpose is to lend to support economic development and/or to provide savings,
payment, and deposit services to the public. This group includes postal banks, government
savings banks, SME lending facilities, agriculture banks, development banks, etc.
8
World Bank (2008b) presents more detailed data on the status of national payment and securities settlement
systems worldwide.
9
In CGAP Occasional Paper No. 8, (2004) the authors refer to these non-bank financial institutions as AFIs for
“Alternate Financial Institutions”.
10
The survey asked respondents to list the types of retail financial institutions active in their jurisdiction. Table 1
gives a fictitious example of the variety of institutions that might be reported (the table was used as an example for
respondents.)
7
4) Microfinance institutions – financial institutions whose primary business model is to lend to
(and possibly take deposits from) the poor.
5) Other – other institutions providing deposit or retail lending services which don’t fall into
the above categories.
These categories were chosen in part to facilitate the disaggregation of some of the institutions
which are typically more active in serving financial inclusion policy goals (such as SSFIs) or
which tend to target middle class or low income clients with basic banking services (MOFIs and
MFIs). The descriptions of the various categories as they were presented to the survey
respondents are provided in Table 2. The regulators were asked to use their best judgment to
choose only one category in which each type of institution fits. In a few cases, regulators could
not separate data for two types that likely would have fallen into different categories (e.g. credit
union data could not be separated from commercial bank data in Switzerland) and were asked to
put them into the category which was dominant, or to leave both categories as missing and
provide only the system wide total. An example of how the form was filled by a regulator is
shown in Table 3.
The MFI category is somewhat problematic in that many institutions which would likely be
categorized by regulators as falling into the other categories consider themselves to be MFIs. For
example, deposit taking MFIs are registered as banks in many countries small rural cooperatives
conduct micro-lending exclusively. Since our main focus is the use of financial services by the
poor, we decided to keep the MFI category and recognize that in many countries where no data
was provided, MFIs are grouped in with one of the other categories.
A second part of the survey contains questions on regulations and policies relating to access to
financial services, including:
Financial services provided though the post offices
The use of agents and correspondents
Bank account management
8
Bank branch regulations
Collateral and lending
Transparency and consumer protection
Policies promoting of access to finance.
In this part of the survey, we ask regulators to respond with answers that reflect the rules and
regulations that commercial banks must follow if there is any discrepancy in how the regulations
apply across institutional categories.
2.2 Implementation of the Data Collection
The questionnaires were sent directly to the head office of the main commercial bank regulator,
usually the Governors’ offices of Central Banks or Banking Supervisory Agencies. Depending
on the country and the structure of the Central Bank, the questionnaires were filled out by one or
several of the following departments: research, statistics, supervision, and/or foreign relations.
Out of 144 questionnaires sent, 129 countries responded; Eastern Caribbean Central Bank
responded with seven additional unsolicited questionnaires which are included in the dataset.11
Central Bank of West African States (BCEAO), which accounts for eight countries, did not
provide a response. Also, Egypt, Saudi Arabia, United Arab Emirates, Sudan, Liberia, Nigeria
and Sierra Leone did not return the surveys. We discuss response rate and data availability in
greater detail in Section 3 below.
To verify the data we conducted multiple phone and email follow-up sessions with any country
for which clarification was needed. We also conducted a full set of internal consistency checks
and verified the data against external data sources where available including the IMF-IFS
statistics, the MIX and data from World Bank (2008) and Beck, Demirgüç-Kunt, and Martinez-
11
Additional surveys were received from Anguilla, St. Kitts and Nevis, St. Vincent and the Grenadines, St. Lucia,
Dominica, Grenada and Montserrat.
9
Peria (2007). Where any discrepancies were found, respondents were contacted by telephone or
email for further clarification.
Scatter plots of the comparison between the Financial Access Survey commercial bank deposit
account penetration measure and measures from World Bank (2008) and Beck, Demirgüç-Kunt,
and Martinez-Peria (2007) can be seen in Figure 2. In the two comparisons, differences come
from changes over time and from the fact that the two previous efforts requested data on all
deposit money banks which often included cooperatives and credit unions in the total, as is the
case of Spain, Italy, and Austria.
2.3 Advantages and Limitations of the Data
Our approach has three important limitations which impact the comparability of our measures
across countries and the uses to which they can be put:
Differential reporting: Not all countries report data for the same types of institutions which has
implications for comparing data across countries. Specifically, data from non-banks seemed to
suffer from greater underreporting bias than data from commercial banks. This bias expresses
itself in two ways. First, and most common, many countries did not report anything for certain
categories of institution (cooperatives for instance) though that type of institution was clearly
operating within their jurisdiction. Second, in just a few cases, countries reported incomplete
data for a given category (for instance, if the data reported comprised urban cooperatives but not
rural cooperatives). Desk research showed this issue was not pervasive and was of small
magnitude in most cases. For example, data from the Microfinance Exchange (MIX) showed that
the number of loans and deposits from unregulated NGO-MFIs (which were unlikely to be
represented in data sent to us by the central bank, since these institutions were not regulated)
were a small fraction of those from regulated MFIs in most countries, which were again a small
fraction of total loans. Having data disaggregated into multiple categories and data on which
specific institutions are included in each category also helped mitigate these problems.
10
Unavailability of Certain Variables: By approaching Central Banks, the variables we can collect
are often limited to data they have on hand, e.g. that provided in balance sheets, income
statements, and standard regulatory reports. Many regulators were not able to provide our main
statistics of interest, which were number of deposit accounts, number of loans, and bank
branches. In addition, respondents often had trouble disaggregating data in the ways we had
asked for it. Section 3 below discusses data availability in greater depth.
Inability to Measure Individuals: Finally, the main outcome variable, number of accounts or
loans per 1000 adults, is not a perfect indicator of individuals with a loan or a deposit account.
Double counting is a problem as people may have more than one bank account and/or have
different accounts with different banks. Also, most countries do not make a distinction between
government and corporate vs. individual deposit accounts. Another inconsistency stems from the
different treatment of dormant accounts – some banks close dormant accounts after six months of
inactivity, while other banks keep dormant accounts open for many years. See section 4.4 for a
discussion of the correlation between accounts per 1000 adults, and the number of individuals
with an account, which is strong despite the potential for double counting.
The main advantages of the approach taken here are the levels of disaggregation achieved both
by type of product, and type of institution delivering the product. In many cases, respondents
were able to provide data on the individuals and businesses having loans, as well as finer
breakdowns for type of deposit account into time, savings, and checking. Disaggregation makes
the data more useful, but also facilitates greater accuracy by showing more clearly what is
missing and what is not.
3. Patterns of Bank Supervision and Data Availability
Modern financial systems are complex and feature a great variety of regulated and unregulated
financial service providers. In most countries there is no single central supervisor or coordinating
entity for all financial institutions. However, the main financial authority, usually a central bank
or bank supervisory agency, regulates some non-bank financial institutions along with banks in
11
about half of the countries in the world (Table 4). In many countries there are also non-bank
regulatory authorities that may supervise or regulate (to varying degrees) cooperatives and credit
unions, finance companies, MFIs, etc.
The patterns of supervision documented in Table 4 have consequences for the response rates for
different data elements in the survey. Financial regulators tend to collect and publish data on the
institutions they supervise and the data tend to reflect their policy priorities. This pattern shows
clearly in Table 5, where volumes of deposits and loans; and data on branches and other forms of
physical outreach are relatively available, the data on numbers of deposits and loans and
numbers of individuals are relatively less populated. Volume data comes from balance sheets and
is reported regularly for monitoring purposes and many countries require registering bank
branches with the authorities. The collection of data relevant to financial inclusion is rarer,
however, and is often lower priority and done on an ad hoc basis.12
3.1 Data Available on Mutually Owned Financial Institutions (MOFIs)
Even when financial cooperatives are regulated, few countries are able to provide financial
access relevant data on them. In countries where cooperatives are supervised by the main
financial regulator, less than 60 percent have data on the values of deposits and loans, and just a
third on the number of accounts and loans. There is a significant difference in data availability
by region. Latin America has the best data coverage with 80 percent of countries collecting data
on values and 60 percent on numbers of loans and deposits. Sub Saharan Africa faces the
greatest challenge in collecting data on cooperatives. In the region, only 20% of countries where
cooperatives were regulated by the financial authority, had data on the number of deposits and
loans. Many cooperatives are small and many of them do not have an MIS or even a computer,
which likely accounts for some of this trend.
12
Many respondents reported collection the data on numbers of loans and deposit accounts in response to our
request. In contrast, in many developed countries information on financial inclusion is collected through household
surveys.
12
3.2 Specialized State-owned Financial Institutions (SSOFIs)
In most countries (62 out of 80 who report having specialized government institutions) the main
bank regulator supervises government banks.
Only in 18 countries these institutions are
supervised by other government agencies, such as ministries of finance for development banks
and ministries of post and communications for postal banks. Even though specialized state
owned institutions are an important provider of services very few countries were able to provide
values on the outreach of these institutions.
3.3 Data Available on Microfinance Institutions (MFIs)
A specific challenge in measuring microfinance is that it is not necessarily defined by the type of
institution, but by the market segment these institutions serve. For example, many cooperatives
operate in rural and poor areas and provide microfinance services. In some countries, banks
entered the space traditionally served by microfinance, such as ICICI in India, Equity bank in
Kenya, or BRI in Indonesia. Accordingly, only countries with separate licensing and regulatory
requirements for MFIs were able to report data separately for these institutions - though many
others presumably had healthy MFI sectors where the MFIs are registered as cooperatives,
banks, or non-bank financial corporation of some other kind.
There are 57 countries in the world where MFIs are defined for regulation purposes as a separate
institution type and regulated by the main financial regulator (Table 4). In 10 of these countries,
multiple forms of MFIs exist, where at least one is regulated by another regulator. In these cases,
there are often NGO MFIs which are loosely regulated by a ministry of the government and
which provide credit services only, and deposit taking MFIs which are regulated and supervised
by the main financial regulator. MFIs are supervised by the main financial authority in about a
half of developing countries. A notable exception is Africa where in 80 percent of countries
central banks supervise MFIs. This is a result of the recent drive to formalize microfinance
operations and recognize the important role MFIs play in serving a large part of the market in
13
African countries.13 Less than 10 percent of high income countries regulate or even have MFI
institution as a category within regulatory framework.
3.4 Summary Statistics
Table 6 presents the summary statistics of the main indicators used in our analysis. As already
mentioned, data on commercial banks are the most comprehensive. The table also shows
substantial variance among countries in the different indicators. The mean (median) of deposit
account penetration in banks is 996 (702) accounts per thousand adults with 25th and 75th
percentile values of 270 and 1498, respectively, while the average balance lies around $7,725
(3,070) USD. On average, the average balance represents 2.6 times the GPD per capita (median
= 0.94). The mean (median) of loan penetration in commercial banks is 299 (215) loans per
thousand adults. The 25th and 75th percentile values for such variable are 42 and 431,
respectively. The number of loans and the number of deposits per thousand adults in commercial
banks are positively correlated, with a correlation of 0.72. In contrast, the average number of
accounts or loans per thousand adults is much smaller in cooperatives, government banks, and
microfinance institutions; for cooperatives, for example, the mean of such figure is 129. While
the number of observations is smaller, these figures are not driven by sample selection.
The average (median) number of branches of commercial banks per thousand adults equals 16
(13). The penetration of branches among the population is positively associated with deposit and
credit penetration, with correlations of 0.45 and 0.44, respectively. In contrast, the equivalent
association with the proxy for geographic penetration is a bit weaker, with correlations of 0.34
and 0.35, respectively.
13
BCEAO countries did not respond to the survey. Recent years have seen a large number of African countries
develop national microfinance strategies including the concept of special regulatory windows for deposit taking MFI
banks designed to encourage the sector. See, e.g. Duflos and Glisovic-Mézières (2008).
14
4. Financial Access around the World
In this section, we use the Financial Access Survey data to roughly estimate the distribution of
rates of deposit account ownership and access to credit around the world. We first use regression
models and other data sources to fill in missing values for the numbers of deposit accounts and
loans in different countries. From the filled in estimates we can generate worldwide totals for
deposit accounts, and the breakdown by country (expressed as maps). We then use household
surveys to estimate the average number of accounts per account holder which allow us to
estimate the number of banked and unbanked individuals worldwide and broken into developing
and developed countries.
There are two main difficulties which we face in these exercises. First, our estimate of the
number of accounts per active accountholder comes from a very limited set of household surveys
and so is quite rough (our error bands are very wide). This is a fundamental limit imposed on us
by the availability of household survey data and our estimate should be treated as only a rough
guess. Second, there is significant bias in the reporting of non-bank data relative to commercial
bank data which we were not able to overcome when filling in the missing data with estimates or
outside data sources. While we were fairly confident in our ability to generate a reasonable
econometric model to predict rates of commercial bank account ownership, there were a number
of factors which made us less confident in any similar attempt with the other categories,
including the fact that the development these other categories tend to be influenced by specific
factors in each country. We tried to fill in the missing values from other data sources, especially
for the two main non-bank categories of MOFIs and SSFIs, but were not confident that the data
we found were very comprehensive. Thus we expect that our estimates of non-bank accounts are
extreme lower bounds and should be interpreted as such. Our estimates of commercial bank
accounts are generated by statistical models and are as likely to be too high as too low for any
given country, so our estimates of commercial bank account numbers are not lower bounds but
best guesses. This implies that where we estimate a ratio of non-bank to bank accounts this ratio
is also an extreme lower bound on the true ratio.
15
We also made a limited attempt to predict the distribution of commercial bank loans around the
world. Here, we focused only on commercial banks because of data limitations. We did not
attempt to estimate an aggregated figure for the number of loans because, unlike deposit service
providers, there are a large number of non-regulated credit suppliers worldwide, rendering the
estimation of a meaningful figure through regulatory data practically impossible.
4.1 Counting the Number of Accounts
The broad coverage of our data allows us to develop a model to extrapolate out of sample to
estimate the total number of bank accounts and their distribution around the world. To reach
these estimates, we start with our survey data and develop a simple regression framework to
impute values for some countries for which we don’t have commercial bank or cooperative data.
In addition, we attempt to fill in gaps in the non-bank data with data from other sources,
including the MIX, CGAPs Occasional paper no. 8, the World Council of Credit Unions annual
survey, and country sources. In order to express the uncertainty in the estimate we calculate a
conservative “low estimate”, a “preferred estimate” and a “high estimate” to give a sense of the
range of values which may apply when we vary our assumptions. To generate the map and the
count of accounts worldwide, we use the sum of the preferred estimates for commercial banks,
cooperatives, government banks, and MFIs. The details of how numbers for each bank category
were estimated are provided below, including the preferred estimate, and the high and low
estimate reflecting different assumptions:
4.1.1 Commercial banks
[Preferred estimate] We first construct a set of regression models to fit the relationships between
log of account penetration on the LHS and per capita income, size of financial sector, population
density, value of deposits per capita, number of commercial bank branches per person, and
whether the country is in the high-income OECD category on the RHS. This specification choice
reflects a balance between wanting a parsimonious model with reasonable explanations as to
why each variable should be included (so that we can think through whether the model will hold
up well in conducting out of sample predictions) vs. the desire to get the best within sample fit
16
(which would imply more variables and/or a highly non-linear terms). We use log of deposit
accounts/1000 adults as the LHS variable in the regressions because the relationship with
observable factors is more straightforward, and because doing so avoids predicting negative
values for the total number of accounts. Using the log, however, presents a challenge because the
predicted values must be retransformed and multiplied by adult population to calculate the total
number of accounts in the country, and this process can induce a bias due to the effects of
Jensen’s Inequality. There are various methods to correct for this “retransform bias”, but we
choose the non-parametric smearing method (see Duan (1983)) retransformation correction
which is robust to non-normal residuals (the residuals from some of our specifications do not
appear to be normal based on a joint test of skewness and kurtosis). After generating a model, we
then predict out of sample the number of deposit accounts per 1000 adults for the remaining
countries and retransform the series to get the total number of accounts in each country. Where a
given piece of data is missing for a given country (e.g. if we do not know the number of
commercial bank branches in a country), a model with only the known elements is estimated and
used to predict the number of deposits for the countries with that data element missing. In all the
models were able to explain more than 70% of the variation in deposit accounts (within sample)
and greater than 80% when all predictors were available. Table 7 contains the results of the
different regression specifications; Table 8 describes the variables used in the different models.
4.1.2 Cooperatives, credit unions, and other mutually owned banks
[Preferred estimate] For countries which did not report data on the number of deposit account in
these institutions, we use the reported numbers of clients from the World Council of Credit
Unions (WOCCU). In general, the WOCCU numbers should understate the true number of
accounts in all mutually owned institutions, since they enumerate clients rather than accounts,
and are voluntarily reported data for credit unions only, not all mutually owned institutions.
Thus, we believe the WOCCU numbers are a very conservative estimate of the number of
accounts.14 For 13 countries in Europe for which there were no WOCCU data, we have data –
14
Nevertheless, there is a chance that some of the WOCCU members are unregulated institutions, and thus that the
numbers reported by WOCCU are higher than what would have been reported by the financial regulator, implying
17
again enumerating the number of clients, not accounts – from the European Association of
Cooperative Banks (EACB). For a few countries which do not report numbers of accounts or
clients in the Financial Access Survey, WOCCU, or EACB but which report value of accounts,
we conduct a simple regression based prediction of numbers of accounts from per capita income
and the value of accounts in mutually owned institutions relative to total adult population. Table
9 contains the regression table which has an R2 of 0.8 with 22 observations. The parameter for
log of income is -0.93 and the parameter for log of deposits per adult population equals 0.97,
both significant at 1%. While it might seem somewhat surprising that the coefficient on income
should be negative, this likely reflects the fact that, conditional on the amount of deposits relative
to the total population, the higher income the country, the fewer individuals (and thus fewer
accounts) these deposits represent. As we have a large number of small countries for which no
data is available, we can also make assumptions about the distribution of the cooperative
penetration variable for the missing values to improve our estimate. Specifically, in the preferred
estimate, we assume that the countries with missing values (no reported values, no WOCCU or
EACB data, and no predicted value from the regression) have an average cooperative deposit
penetration which is close to the 25th percentile of the distribution of reported values. This would
be a reasonable approximation, for instance, if many of the countries which did not report
numbers had smaller cooperative sectors than the ones who did. As table 10 shows, the bulk of
the accounts we count (92%) come from actual reported data including our survey data, data
reported to WOCCU or EACB or from direct communication with regulators (CGAP numbers).
[High estimate] The “high estimate” is the same as the preferred estimate with the exception that
we assume the average of the missing values are closer the 50th percentile of the observed
distribution.
[Low estimate] Here we assume that the missing values are zero.
4.1.3 Government savings banks
that compared to countries for which we have survey responses, WOCCU will overstate the number of accounts in
regulated MOFIs.
18
[Preferred estimate] The great heterogeneity in the form and function of government banks and
the likelihood that there is only one or two (or zero) in any given country renders suspect any
econometric regression estimates as were attempted above for commercial banks. Similar to the
MOFI estimate, we start with the reported data we have and fill in missing values with data from
a previous counting exercise conducted by CGAP presented in CGAP Occasional Paper no. 8.
Since this older data is likely to underestimate the true numbers and because there are many
countries which likely have government run deposit taking institutions such as national savings
banks or postal banks but for which we have no data, we believe our estimates understate the true
value.
[High estimate] We don’t make a high estimate (i.e. we assume it’s the same as the preferred
estimate) given the likelihood that there is only one or two (or zero) specialized government
banks in any given country.
[Low estimate] We use only our reported numbers and leave out the numbers from CGAP
Occasional Paper no. 8 on the grounds that in a few cases, they may be double counted as
commercial banks, despite their state ownership (we believe this is may have occurred in just a
few cases).
4.1.4 Microfinance Institutions
For MFIs, we take data from the MIX and use it to supplement our own data for countries which
reported nothing. Many MFIs are licensed as a commercial bank, cooperative, or NGO and thus
might be counted in the numbers for other categories supplied to us by the regulator so we have
removed cooperatives and banks from the MIX data. Since the MIX data are voluntarily
reported, and because the data reported to us in our survey covers only 17 countries, we believe
these estimates are significantly on the low side. Also, though most of the reported account
numbers from the survey data will be from regulated institutions that have a supervisory
relationship with the main bank regulator, the MIX includes many institutions which are not
supervised or regulated.
4.2 Mapping Bank Accounts around the World
19
The map in Figure 5 shows a total number of deposit accounts per thousand adults using a
combination of account numbers collected through Financial Access survey where available and
the estimates generated through the above procedures. The deposit account penetration ratios
represent all deposit accounts summed across the various categories (in most countries,
commercial banks are the vast majority).
As the map shows, rates of deposit account ownership in formal institutions vary greatly around
the world. Of the seven countries which have fewer than 100 bank accounts per 1000 adults,
five are in Africa —Congo D.R., Burundi, Madagascar, Mauritania, and Ethiopia. High income
countries exhibit the greatest levels of deposits penetration with an average of over 2000
accounts per 1000 adults.
Underlying the wide variation in rates of account ownership are large differences in the ability of
poor households to access formal savings. In recent household surveys, Rwanda, Pakistan,
Malawi, and Uganda15 all reported less than 20 percent of households save through formal
institutions, and Financial Access Survey data shows them to have fewer than 225 bank accounts
per 1000 adults. In contrast, a recent study by the European Commission calculates that in
Belgium and the Netherlands, greater than 98 percent of households have bank accounts and
survey data show them to have over 1500 accounts per 1000 adults in commercial banks.16 The
trend is that higher rates of bank account ownership equate to more banked individuals in the
population (see also section 3 above for the results of regressions of deposit penetration on
reported numbers of banked adults in a country).
4.3 The Number of Accounts Worldwide
Adding all the predicted and reported values puts global number of bank and non-bank accounts
in the world at approximately 6.2 billion or more than 1 for each adult on the planet (and nearly 1
for each of the 6.7 billion people on the planet).17 While there are more than enough accounts to
15
FinScope Rwanda (2008); FinScope Pakistan (2009) ; FinScope Malawi (2008) ; FinScope Uganda (2006).
In Belgium 3,724 and the Netherlands 1,772 accounts per 1000 adults. European Commission (2008)
17
These numbers are necessarily rough and likely to have a wide margin of error. In the estimation for commercial
bank accounts, the estimated margin of error is approximately 6%. The margins of error for the non-banks are likely
16
20
go around, they are not distributed equally. If we divide the number of accounts in each country
by the adult population, we obtain an estimated figure of 3.2 accounts per adult in developed
countries, but fewer than 0.9 accounts per adult in developing countries. It should be noted that
these accounts are not all held by individuals, but also include accounts held by businesses and
government agencies.18 Neither are they evenly distributed within countries; in poorer countries
household surveys show that the majority of accounts are often held by the well off.
According to the estimates presented in Table 13, banks are the main providers of deposit
services holding over 80 percent of all deposit accounts in the world (Table 13). At least 20
percent of accounts are held outside the commercial banking sector in cooperatives, credit
unions, government banks, and MFIs. Due to data limitations and the conservative nature of our
assumptions during the counting exercise, the estimate of the number of accounts in non-bank
institutions is likely more conservative than the one for banks, understating the true size of nonbank sector.
Though the numbers are rough, Table 13 shows large differences in the structure of the non-bank
markets in developing and developed countries. In developed countries, nearly 16 percent of
accounts are held with cooperatives, credit unions, and other institutions with a mutual
ownership structure compared to an estimated 3 percent in developing countries (though, again,
the true share of cooperatives in developing countries is likely to be underestimated). Public
institutions such as postal and specialized state savings banks are also important providers of
savings services in developing countries, holding 14 percent of total deposits. Microfinance
institutions, as a separately regulated type of institution, hold only about 2 percent of deposits
and are concentrated in developing countries.
larger as it is difficult to tell how many accounts are not being counted in the countries for which no data was
reported, and how many do not get reported in the countries which did provide numbers.
18
The Financial Access Survey did not ask for the breakdown of deposit account data by individuals and businesses,
but it is likely that these business and government accounts to be an important fraction of the total in some cases.
21
4.4 Conversion Factor: Converting the Number of Accounts to the Number of Account
Holding Individuals
It is possible to convert our data and estimates of the number of accounts in each country into
rough estimates of the number of banked individuals in each country. To do so, we have to
estimate a conversion factor, which is the number of accounts in a country, per banked individual
in each country.19 We use data from household surveys to estimate the number of individuals
who have savings accounts in formal financial institutions (which could be any of our
institutional categories) and divide the number of accounts in the country by this number.
We begin by evaluating whether our main measure of deposit account penetration, namely the
number of accounts per thousand adults, are significantly correlated with some more accurate
indicators of financial access, which are available for some countries through household surveys.
A regression of the reported rates of formal account ownership from household surveys on our
proxy for deposit penetration, namely number of accounts per thousand adults (in logarithms),
shows that the deposit penetration indicator is a good predictor of true financial inclusion, at least
in the limited sample of countries for which we have comparable household level data. Figure 1
shows the added variable plot for the log of deposit penetration; the fit is very close. Next, Table
14 shows the results of three bivariate regressions with rates of account ownership from
household surveys on the LHS and log of deposit penetration on the RHS. In the first model,
only surveys completed since 2003 are included, in the second model, the sample is further
restricted to surveys which have all adults as the survey frame (as opposed to all households), the
final column are all available surveys (which go back to 2000). The coefficients are largely
similar across the three samples, and the R2 is 0.69 and 0.71 when only more recent (and thus
more accurate) surveys are used. These results confirm that the deposit penetration ratio can be
used to predict the percent of the population with a bank account to a fair degree of accuracy.
Table 17 provides a list of household survey data and the sources.
19
This is not necessarily the same as the average number of accounts held by each account holder since business and
government accounts will be counted in the numerator.
22
To create our preferred estimate of the conversion factor, we select the available 15 household
surveys conducted in 6 developing and 9 developed countries since 2003 (see Table 15 for list of
survey sources) and which have reported data for two categories or more (commercial banks plus
one of MOFIs or SSFIs).20 To assure numbers are comparable across countries, we restrict our
sample to only those surveys which use all adults as the survey universe. The mean value of the
conversion factor is 2.96 and the 90% confidence interval is from 2.0 to 3.9 (Table 16). As a
round number, there appear to be approximately 3 accounts per person on average.
Table 17 also shows two variations to calculate a conversion factor. In the one, we use all
countries with household survey data and use both predicted and reported values of account
numbers. This expands the sample to 28 countries and produces an estimate of 3.2 total accounts
per banked individual for the conversion factor. The final method uses only reported data for
commercial banks under the assumption that the number of commercial bank accounts provides
a reasonable signal for estimating the number of individuals with bank accounts. This method
uses 17 countries and calculates a conversion factor of 2.5 commercial bank accounts per banked
individual.
This approach to calculating the conversion factor is admittedly a very rough estimate (as is
confirmed by the wide confidence intervals). Though we believe the simplicity of the calculation
is a virtue, there are a few simple corrections which we contemplated, but which the data
indicated were not warranted.
First, because the Financial Access survey data on number of accounts comes from 2008 data but
the household survey data come from various prior years, any worldwide upward trend in
account ownership would systematically inflate the conversion factor for older surveys. If older
household surveys systematically underestimate the true number of banked adults who currently
own accounts, one might expect a negative relationship between survey year and number of
20
We calculate the conversion factor as the number of accounts from all sources relative to the number of banked
adults. In the conversion factor calculation we drop countries for which we have only commercial bank data and no
reported number for government or cooperative accounts (3 surveys) so that the conversion factor better measures
this ratio (rather than measuring the number of commercial bank accounts per banked adults). When we do the final
calculation, we use predicted values for cooperatives and government accounts in countries where no data were
reported.
23
banked adults. However, a regression of accounts per adult on the year the household survey
does not show a significant relationship.
Second, one might also imagine that the conversion factor would vary significantly with the level
of economic development of the country, however there was not a significant difference in the
means between the developing and developed sample nor was there a significant relationship
when regressing the conversion factor on log of GDP per capita. This result seems surprising,
given the anecdotal evidence that individuals in richer countries often hold more accounts per
person. One possible explanation could be that though the typical adult in a developed country
may have more accounts on than her developing country counterpart, the ratio of business and
government accounts to individual accounts is likely higher in developing countries (due to
fewer individual accounts) which drives back up the ratio of accounts to banked adults.21
4.5 Estimating the Number of Unbanked Individuals in the World
To make a final estimate of the number of banked adults around the world, we divide the number
of accounts in each country by the conversion factor. When the resulting number is greater than
the number of adults in the country (as was the case in a few countries with higher than 3000
accounts per 1000 adults) the number of banked adults was set to the total number of adults.
Assuming 3 accounts per banked adult on average puts the number of un-banked adults in
developed countries at approximately 160 million or 19 % of all adults, and at 2.7 billion adults
or 72 % of the adults in the developing countries.2223 Table 17 shows how this breakdown varies
using the high and low confidence interval values for the conversion factor.
21
Desk research into a few country examples showed a high ratio of government and business accounts in
developing countries.
22
CGAP Occasional Paper no. 8 estimates the number of savings and loan account holders in non-banks to be 500
million, leaving 2.5 billion poor around the world who do not get services form non-banks. CGAP Occasional Paper
no. 8 numbers are not directly comparable to the ones presented in this report as they don’t include commercial
banks and because they add together loan and deposit accounts. Never the less, their estimate of 2.5bn unbanked
poor is similar in magnitude to the estimate presented here of 2.7bn adults in developing countries (some of which
may not be poor).
23
It should be emphasized that the percentages reported are of all adults, not percentages of households banked.
24
Clearly the numbers are rough but give a sense of the scope of the problem of financial
exclusion. In developing countries, somewhere near 70% of adults have no recourse to formal
savings. These figures present a challenge to policy makers and those in the development community.
4.6 Estimating the Number of Commercial Bank Loans in the World
We also made a limited attempt to predict the distribution of commercial bank loans around the
world for purposes of filling in the map in Figure 6. The prediction exercise is similar, but is
based on a single regression. The prediction of the number of loans per thousand adults was
carried out only for commercial banks, mainly because of data limitations. Furthermore, we did
not attempt to estimate an aggregated figure for the number of loans because, unlike deposit
service providers, there are a large number of non-regulated credit suppliers worldwide,
rendering the estimation of a meaningful figure through regulatory data practically impossible.
The rationale for these covariates is straightforward, as richer countries tend to have more
developed credit markets, and more and better information exchange among lenders is often
associated with deeper creditor markets. We model the number of loans per thousand adults (in
logarithms) as a quadratic function of (the logarithm of) GDP per capita and the 2009 Credit
Information Index of Doing Business.24 The results are presented in Table 18. Our preferred
specification is presented in column (1). We estimate this model using a sample of 69
observations. The adjusted R2 is 0.82. In column (2), we present an alternative specification
using the value of loans in commercial banks instead of income per-capita. The idea behind this
specification is that the average loan is relatively the same across different countries, although in
reality, this is not the case. The fit in terms of R2 is slightly better (Adjusted R2 = 0.86 vs. 0.82);
however we stayed with our previous specification as GDP per capita is available for a larger
number of countries.
Figure 6 shows the worldwide penetration of loans, measured by the number of commercial bank
loans per thousand adults, using a combination of the numbers collected through the Financial
24
We tried other specifications (using levels instead of logs), but the fit was worst. We also include as explanatory
variable the ratio of domestic credit to GDP; however, because of our reduced and selected sample of high-income
countries, including this covariate gives inaccurate estimations for developed economies.
25
Access survey and the estimates generated through the methodology already described. The
penetration of loans varies widely across countries and is closely correlated with economic
development. Developed economies have the largest number of loans per thousand adults.
What’s more, it is likely that this figure underestimates the true amount of loans in these
countries due to the presence of a large number of highly-developed unregulated lenders. In
Eastern Europe and Central Asia, there are, on average, 367 loans per thousand adults. Latin
America and South Asia follow with 314 and 268, respectively. Middle East and Africa are the
least develop markets.
5. Which Country Characteristics Are Correlated with Outreach of the Financial System?
This section explores the empirical relationships between our financial access indicators and an
array of country-level variables that intuition, theory, and previous empirical work suggest might
be relevant. Cross-country analyses such as these suffer from a number fundamental problems
(including omitted variables and reverse causality) which make it impossible to infer a causal
relationship between the independent and dependent variables, even when the regression
coefficients are statistically significant. In measuring the correlations and conditional
correlations with regressions we do not hope to imply any causal relationships, but simply to
map out the main relationships and features of the data. A similar exercise was carried out in
Beck, Demirgüç-Kunt, and Martinez-Peria (2007). In that paper, the authors correlate deposit
and loan penetration with various factors from the empirical literature to explore the
determinants of financial development. We improve on their effort by using a larger sample of
countries and newer data. As a rough guide, we rely on the framework outlined by Beck and de
la Torre (2006) to determine which factors are likely to impact financial outreach and thus
should be included in our analysis.
We divide the analysis into three areas: savings, credit, and physical outreach. Table 19 provides
correlations between all of our indicators of account penetration, loan penetration and
demographic penetration of branches and all explanatory variables. Tables 20-23 rely on a
theoretical framework and build simple OLS specifications to estimate the relationship between a
26
single country-level covariate, for example the quality of information in credit markets, and our
indicators of financial usage and banking sector outreach, controlling for the overall level of
economic development and/or population density as a catch of all for the many other factors
which may affect the outcome variable.25 Here, the coefficients are identified by the difference of
these covariates across countries. Throughout our analysis, we categorize the differences in
development, institutions, and economic policies across countries as arising as the result of
“market-developing” or “market-enabling” policies.26
5.1 Savings
We begin by studying the determinants of the number of deposit and saving accounts. Here, our
dependent variable is (the logarithm of) the number of accounts in commercial banks per
thousand adults. The results are presented in Table 20. We start by analyzing the role of factors
that generally call for market-developing policies.27 Specifically, we estimate the correlation
between the number of accounts per thousand adults and different variables reflecting macroeconomic and development conditions that are likely to affect the supply of deposit services in a
country. We consider first the role of economic development, income inequality, and population
density. Broadly speaking, these variables reflect the importance of income and market size on
the provision of financial services, although they might also be catchall proxies for other factors
which are correlated with the level of economic development. Consistent with previous research,
the first column in Table 20 shows that there are fewer accounts per thousand adults in countries
with low per capita income. In fact, as evidenced by the high R2 =0.64, GDP per capita accounts
for a large part of the cross-country variation in deposits penetration. The second column adds
population density, as measured by (the logarithm of) population divided by area. Even
controlling for per capita income, density is significantly correlated with the number of deposit
25
We use this specification because there is not enough variation in the data set to identify the coefficients when all
the variables are included at once. This raises the concern that despite controlling for per capita income, there may
be important unobserved heterogeneity among countries correlated with our covariate of interest. We cannot
overcome this problem.
26
Market-developing policies expand the possibilities frontier through structural reforms that improve institutions
and other state variables. In contrast, market enabling policies modify the incentives and constraints faced by
financial institutions in different regulatory environments
27
In Beck and de la Torre (2006) the authors describe factors that call for market developing policies as requiring
profound structural reforms and tend to change slowly over time.
27
accounts. Population density might be expected to correlate with financial access through a
number of channels. Among other things, more dense areas are easier to supply with
infrastructure and other services. Also, since one bank branch can serve more customers in a
dense area, banks may make greater investments in banking infrastructure in high population
density areas [see, e.g. Calem and Nakamura (1998)]. In the section on physical outreach, we
find that density is indeed related to greater branch penetration. Given these results, in what
follows, we include these two variables as controls for the level of development. In the third
column, we add the Gini coefficient to the previous model to test any relationship between
income inequality and account ownership. We find no significant association between the
income distribution and our dependent variable, after controlling for income per capita and
population density.28
Next, we study the relationship between variables that measure macro-economic stability and
deposit penetration. Previous research has shown there may be a two-way relationship between
stability and financial development. On the one hand, inflation and other forms of instability may
impair the growth of a robust banking sector [see e.g. Rousseau and Wachtel (2001)].29 On the
other hand, a strong and well regulated banking sector may dampen macro economic shocks or
prevent them from occurring thus implying a negative relationship between measures of the
development of the system and instability [see, e.g. Tornell and Westermann (2003)]. In column
(4), we add to our basic specification the average inflation (in terms of CPI) during the last ten
years to proxy for macroeconomic stability. As might be expected from the work cited above,
inflation has a negative and significant effect on deposit penetration. Yet given the many
channels by which these variables may affect each other, it is especially important not to impute
any causal interpretation to the relationships we find. Column (5) studies the association between
explicit deposit insurance and deposit account ownership. The column shows our basic
specification and a dummy variable coded one if the country had deposit insurance in 2003 and
28
In an unreported regression, we also include an interaction term of the Gini coefficient and the log of GDP per
capita. None of the covariates are statistically significant.
29
In theory, the effect of inflation on savings is ambiguous. In general, while inflation affects negatively the value of
savings by lowering the real value of wealth, it may also mobilize savings into the system as households might
prefer to save in banks provided that the nominal interest rate is sufficiently high, rather than in the form of money.
It is also possible that households might prefer to raise their savings rate in order to offset the negative wealth effect.
28
zero otherwise [data on deposit insurance comes from Demirgüç-Kunt, Karacaovali, and Laeven
(2005)]. The indicator variable is positive and significant; once we control for per capita income
and population density, countries with deposit insurance have more accounts per thousand adults
than countries without deposit insurance.30 Taken as a whole, there is some evidence that
macroeconomic stability is associated with a deeper penetration of deposits.
Columns (6) – (9) present the regression results of the impact of infrastructure and political
stability on deposit penetration. Deficiencies in infrastructure can drive up costs for financial
institutions to supply financial services; may impact business activity, reducing demand for
financial services; and (in the case of transport and information infrastructure) may imply higher
transaction costs for customers to access services. In fact, countries with higher electricity
consumption and more phones per capita have, on average, more accounts per adult. In contrast,
while a deficient transportation infrastructure might increase the cost of outreach into more
remote areas, the density of roads comes in negatively and not significantly. Although political
instability and violence can increase the cost of doing business, this variable is not significantly
associated with deeper penetration once per capita income is taken into account. Finally, column
(10) considers economies designated “offshore financial centers” by the IMF, whereas we would
expect to have greater numbers of accounts per resident, as they often hold accounts for large
numbers of non-residents. The sign of the indicator variable, however, is contrary to expectations
and not significant, which may reflect the fact that we do not include most of the small islands in
the Caribbean or the fact that many offshore centers do not specialize in banking services but
instead more advanced financial instruments.31
30
Our regression analysis uses an indicator variable that takes one if there is a deposit insurance scheme in place in
the country, and zero otherwise. This is for simplicity and thus the result should be taken with caution. On one hand,
the result does not provide any evidence of causality. On the other hand, there is a well established literature
showing that the moral hazard engendered by overly generous deposit insurance schemes leads to greater instability
and slower long-run financial development [Demirguc-Kunt, Kane, and Laeven (2008)].
31
In unreported regressions, we estimated separately the relationship between deposit penetration and two indices,
namely the Creditor Rights Index and the Credit Information Index, controlling for (the logarithm of) per capita
income and (the logarithm of) population density. Both indices enter with a positive sign and are statistically
significant. In theory, however, these indices should not affect directly the number of deposit accounts, so it is likely
that they are simply a proxy for unobserved variation in the development of financial systems or financial reforms.
29
The second panel in Table 20 reports the results of the impact of different market-enabling
policies on the number of deposit accounts [see Beck and de la Torre (2006)]. These policies
tend to provide incentives (or constraints) to financial institutions to operate more efficiently or
to broaden their customer base. The one measure of these policies is the level of competition in
the banking sector. Competitive pressures and the search for profits are key factors for
institutions to innovate, expand their customer base, and offer accounts to underserved clients.
Here, we follow Beck, Demirgüç-Kunt and Martinez-Peria (2008) and proxy competition with
bank concentration, which we approximate by the share of deposits in the five largest banks.
Column (12) presents the results. Consistent with the theory, higher concentration in the banking
sector is associated with lower deposit account penetration, although this result is only
significant at a 15% level. Next, we analyze the role of bank ownership which is a proxy for the
threat of entry as well as quality of management. These data are from Barth, Caprio, and Levine
(2004). Their government and foreign ownership measures are not statistically associated with
measures of deposit penetration. In columns (15) - (23), we study the relationship between price
and non-price barriers and financial outreach. We begin by analyzing the role of barriers to open
a bank account. Our focus here is on five variables: KYC requirements, exceptions to these
requirements, the existence of regulatory requirements for banks to offer basic or low fee
accounts, the number of document required, and the minimum balance to open a checking
account (as percentage of GDP per capita). The first three variables are from the Financial
Access Database (2009); the fourth is from World Bank (2008); the last one is from Beck,
Demirgüç-Kunt and Martinez-Peria (2008). Broadly speaking, the signs of the coefficients
support the idea that countries with lower barriers are associated with higher deposit penetration,
although none of the coefficients are statistically significant. In columns (20) and (21), we assess
whether policies to encourage people to save, such as tax incentives or annual fees, are
associated with deeper deposit penetration. Neither variable enter significantly. Finally, in the
last two columns we investigate whether regulatory policies aimed to foster financial inclusion
are associated with a deeper account penetration. Specifically, we study the role of agents and
postal banks. In the first case, we use and index of the extent to which agents are allowed to
30
perform savings operations.32 In the later case, we use a dummy variable that equals one if
financial services are offered in the post offices and are handled by a separate private operator
and zero otherwise. These variables are from the Financial Access Database (2009). Although
the signs are as expected, none of these policies are significantly associated with broader
access.33 This evidence, however, is not enough evidence to suggest that these policies do not
matter.
5.2 Credit
Table 21 presents the regressions of those factors associated with loan penetration, in which the
dependent variable is (the logarithm of) the number of outstanding loans in commercial banks
per thousand adults in 2008. In addition to costs, the outreach in the supply of credit (unlike the
supply of deposits) is constrained by risks, in particular default and agency risks. Therefore, the
penetration and quality of credit depends not only on the economic conditions, but also on the
financial infrastructure available to manage these risks.
We start by analyzing the relationship between measures of economic and development
conditions and loan penetration. In the first regressions, we only include (the logarithm of) per
capita income and (the logarithm of) population density as independent variables. Not
surprisingly, GDP per capita explains a large part of the cross-country variation. Since
population density is not statistically significant, in subsequent regression, we only include (the
logarithm of) GDP per capita as control.
In terms of structural factors, inequality and inflation appear in the regressions with a negative
sign, but neither is statistically significant. Columns (5) - (7) parallel the results of the previous
sub-section, namely that better physical infrastructure is positively and significantly associated
32
We tried an alternative specification using a dummy variable coded one if the country allows the use of agents.
The coefficient is not statistically significant. These variables, however, measure regulations at a very high level.
Interviews with policymakers show that these regulations often contain detailed provisions, making implementation
difficult and reducing uptake suggesting that more granular data on regulations are necessary to draw conclusions.
33
Allowing private entities to provide financial services through postal agencies enters significantly if we do not
control for population density.
31
with loan penetration. In column (8) we include contract enforcement days as a proxy for the
efficiency of the legal system. In principle, loans should be more prevalent in more efficient
legal systems. While this variable usually enters significantly in other work using private credit
to GDP as dependent variable34, the coefficient here is not statistically significant. A possible
explanation is that the quality of courts matters for business loans, but not so much for individual
loans which constitute a large proportion of loans granted by commercial banks. Column (9)
shows no significant relationship between political violence and number of loans. Finally,
columns (10) and (11) consider the role of financial infrastructure which creates mechanisms for
lenders to screen borrowers and enforce repayment. As proxies for such mechanisms, we use the
Creditor Right Index developed by Djankov, McLiesh, and Shleifer (2007) following La Porta,
Lopez-de-Silanes, Shleifer, and Vishny (1997, 1998) and the Credit Information Index
introduced by Djankov, McLiesh, and Shleifer (2007) and updated in “Doing Business 2009”
[World Bank (2009)]. The results show that the number of loans per thousand adults increases as
the contractual and informational frameworks, as measured by these two indices, improve.
The bottom part of Table 21 explores the relationship between policies that may broaden access
and the penetration of loans in the population across countries. In column (12), we include into
our baseline regression our proxy for bank concentration. Concentration tends to be negatively
and significantly associated with the supply of credit. The ownership structure of the banking
sector, in contrast, is insignificantly correlated with the loan penetration. Columns (15) – (18) of
Table 20 add to our baseline regression four measures of non-price barriers to accessing credit:
(1) The number of locations to submit a loan application, (2) the minimum amount of consumer
loan banks make expressed as a percent of GDP per capita, (3) the average fees banks charge on
consumer loans expressed as percentage of GDP per capita, and (4) the average number of days
banks take to process a typical consumer loan application. All these figures are from Beck,
Demirgüç-Kunt and Martinez-Peria (2008). Holding per capita income constant, these barriers
are not significantly correlated with the penetration of loans in the population. The last column
34
See for instance Djankov, McLiesh, and Shleifer (2007)
32
explores the role of extending credit through agents and uses an index of credit services that
agents are allowed to do. We find no significant association of our index and loan penetration.35
5.3 Outreach
Financial inclusion is unlikely to improve without sufficient physical access points where clients
can access the financial system. A client who must travel long distances to the nearest branch or
ATM to deposit or borrow a few dollars is likely to opt out of the formal financial system.
Evidence seems to support this story. In columns (11) and (12) in Table 20, for instance, we
show that there are more deposit accounts per thousand adults in countries with deeper
demographic and geographic branch penetration. And research indicates that banks also find it
difficult to lend to distant clients, especially to more “informationally opaque” clients such as
SMEs [Petersen and Rajan (1995), Mian (2006)]. Another line of research shows that developing
financial services at the local level can improve local GDP growth and other economic outcomes
[see Kendall (2008), Burgess and Pande (2005), Guiso, Sapienza, and Zingales (2004)]. Since
the extension of financial services to more localities is mostly a matter of increased branching,
these results show the importance of analyzing outreach as its own phenomenon.
To evaluate some of the factors associated with more intensive physical outreach, Table 22 and
Table 23 present OLS regressions using several measures of economic and institutional
conditions that, in theory, influence transaction costs and therefore, can limit access to financial
services. In these tables, we use the number of branches per 100,000 people and the number of
branches per squared km (both in logs) as measures of demographic and geographic branch
penetration, respectively. The first column in both tables presents our basic specification. Not
surprisingly, the evidence shows a strong positive correlation between both measures of outreach
and GDP per capita. In contrast, population density is statistically significant only in the
geographic specification. In the second column, we investigate the relationship between income
distribution and outreach. Contrary to the results in the previous sections, income inequality is
35
It is worth noting that both variables (operations savings and operations credit) reflect the regulations as written
rather than how they are implemented in practice and exercised by those who are subject to them.
33
negatively correlated with both, demographic and geographical outreach. Column (3) shows no
statistically significant relationship between the rate of inflation and any of our outreach
measures. In contrast, indicators of physical infrastructure are positively associated with banking
sector outreach, possibly because good infrastructure reduces the costs of opening and operating
branches (though there are many other unmeasured factors which may also drive the result).
Column (8) shows that, after controlling for the effects of income and population density, there is
no significant correlation between absence of violence and neither of our measures of outreach.
Contrary to our previous results, concentration is not significantly associated with either measure
of branch penetration. Finally, in column (10) we show that a regulatory barrier such as requiring
approval by the financial regulator to open a branch is not significantly associated with lower
branch penetration.36
In sum, the estimates in Tables 19 – 23 yield interesting results. The first one is obvious, GDP
per capita is strongly positively associated with all measures of financial inclusion; it explains a
large fraction of the cross-country variation. Second, after controlling for per capita income,
variables which reflect the institutional and development conditions and that are largely outside
of the control of financial regulators are also associated with a broader penetration of financial
services. More specifically, measures of physical infrastructure, such as phone line penetration
and electricity usage, are positively associated with all four measures of deposit penetration, loan
penetration, and geographic and population based branch penetration. Lower inflation is also
positively associated with deposit penetration. While these results probably confound the direct
impact of better infrastructure or inflation control with other aspects of economic development
and macroeconomic management, these results highlight the fact that some of the factors which
affect the ability of policy makers to set conditions for greater financial inclusion are outside of
their direct control. Third, there is also a significant relationship between measures of deposit
and loan penetration and variables which might plausibly be within the control of financial
regulators and policy makers. In particular, financial infrastructure, as measured by better
36
Cross-country analysis among richer countries shows that requiring branch approval is correlated with lower
branch penetration. This relationship is statistically significant even after controlling for income, population density,
and other factors.
34
contractual and informational environments in credit markets, is positively associated with
broader lending. Similarly, the presence of deposit insurance is also associated with more deposit
penetration37 while higher levels of concentration in the banking sector are positively associated
with more limited penetration of deposit and loans in the population, pointing to a possible role
for competition policy in determining the degree of financial inclusion. Finally, while significant
alone, once per capita income is taken into account, there is no significant association between
increased financial penetration and the presence of any of the policies which have financial
inclusion as a central goal. While cross-country OLS regressions are a very blunt tool and not
really appropriate for measuring the impact of specific policies, these results support previous
findings in the literature regarding the interaction between the macro/institutional environments
and individual policies. These results, however, should not be interpreted as evidence of failure
of any of these policies, but rather as a call for micro-studies to evaluate the effectiveness of
measures aiming to improve access to financial services.
6. Conclusion
This paper introduces a new set of financial access indicators for 139 countries across the globe
and describes the results of a preliminary analysis of the data. Despite its limitations, this data set
is one of the few sources of information which could be used to asses and compare the degree of
financial inclusion across countries. It also allows us to make back of the envelope estimates of
the total number of bank accounts in the world and the number of individuals who have access to
them. Despite the apparent overabundance of approximately 6.2 billion bank accounts in the
world - more than one per adult - a disproportionate amount of the accounts - 3.2 per adult - are
located in the developed world economies, while the equivalent figure in the developing world
reaches is only approximately 0.9 per adult, inclusive of accounts which are not owned by
individuals, such as those owned by government and business entities. In addition, our estimates
indicate that roughly 19% of developed world adults do not have bank accounts (though many
37
Although the causal chain could easily run in the opposite direction, if deposit insurance schemes are more often
adopted as more individuals begin to own accounts.
35
may live in households where other members have accounts), whereas nearer to 72% of adults in
the developing world do not have accounts. Even at the very low level of precision possible in
this type of exercise, these numbers indicate a major gap that has not yet even begun to be
addressed by the many policy initiatives currently underway or by the microfinance movement.
Finally, we investigate the relationship between deposit, loan, and bank branch penetration with
other variables and find significant associations of deposit and loan penetration with per capita
income, physical and financial infrastructure, and macro-economic stability, but no significant
association with policies which have financial inclusion as a central goal.
Having appropriate data is crucial to understanding and measuring financial inclusion. The data
introduced in the Financial Access Database should be viewed as an attempt to generate
consistent cross-country indicators of financial penetration around the world. Yet, as an effort to
document access to financial services worldwide, it faces many challenges. Many countries do
not collect information on key variables and most have incomplete data on the non-banking
sector. There is the need to improve the quality and availability of financial access data, both by
improving and extending cross-country indicators as well as employing country-specific-indepth diagnostics. Additionally, data are often collected on an ad-hoc basis. To be useful,
indicators must be collected on a repeated, regular basis, so that policymakers can set priorities
and track changes. Finally, supply-side data must be complemented with other efforts, mainly
through household surveys, to estimate accurately the characteristics of the population with
access to the financial system. Since this could be a time and resource-intensive exercise, an
alternative could be the use of quick financial access surveys or ‘snapshots’. These snapshots
should not be intended to replace deeper country-level surveys, but could provide a usefully
broad framework in which to set other sources of information. In the end, these efforts should
complement each other to provide a broader picture of financial inclusion in order to identify
obstacles and design policies to overcome them and ultimately expand access.
36
7. References
Banerjee, A., Duflo, E., Glennerster, R. and Kinnan, C. (2009), The miracle of microfinance?
Evidence from a randomized evaluation, Mimeo, May
Barth, J., Caprio, G., Levine, R., (2004), Bank regulation and supervision: What works best?
Journal of Financial Intermediation Vol. 13, pp. 205-248.
Beck, T. and de la Torre, A. (2006), The basic analytics of access to financial services, World
Bank Policy Research Working Paper No. 4026.
Beck, T., Demirgüç-Kunt, A. and Martinez-Peria, S, (2007), Reaching out: Access to and use of
banking services across countries, Journal of Financial Economics, Vol. 85 (1), pp. 234-266,
July.
Beck, T., Demirgüç-Kunt, A. and Martinez-Peria, S. (2008), Banking services for everyone?
Barriers to bank access and use around the World, The World Bank Economic Review, Vol. 22(3)
pp. 397-430
Bruhn, M. and Love, I. (2009), The Economic Impact of Banking the Unbanked: Evidence from
Mexico, World Bank Policy Research Working Paper No. 4981.
Burgess, R. and Pande, R. (2005), Do Rural Banks Matter? Evidence from the Indian Social
Banking Experiment, American Economic Review, vol. 95(3), pp. 780-795, June.
Calem, P. and Nakamura, L., (1998) Branch banking and the geography of bank pricing, The
Review of Economics and Statistics, MIT Press, vol. 80(4), pp. 600-610, November.
Christen, R. Jayadeva, V. and Rosenberg, R. (2004) Financial Institutions with a Double Bottom
Line: Implications for the future of microfinance, CGAP Occasional Paper No.8.
Collins, D., Morduch, J., Rutherford, S., and Ruthven, O. (2009), Portfolios of the Poor: How
the World's Poor Live on $2 a Day, Princeton University Press.
37
Demirgüç-Kunt, A., Karacaovali, B. and Laeven, L. (2005), Deposit insurance around the world:
A comprehensive database, World Bank Policy Research Working Paper No. 3628
Demirguc-Kunt, A., Kane, E. J. and Laeven, L., (2008), Deposit Insurance around the World:
Issues of Design and Implementation, vol. 1, 1 ed., The MIT Press.
Djankov, S., McLiesh, C. and Shleifer, A., (2007), Private credit in 129 countries, Journal of
Financial Economics, Vol. 84(2), pp. 299-329, May
Duan, N. (1983), Smearing Estimate: A Nonparametric Retransformation Method, Journal of the
American Statistical Association, Vol. 78, No. 383, pp. 605-610, September
Duflos, E. & Glisovic-Mézières, J. (2008), National Microfinance Strategies, CGAP Policy Brief.
Dupas, P. and Robinson, J. (2009), Savings Constraints and Micro-enterprise Development:
Evidence from a Field Experiment in Kenya, NBER Working Paper #14693.
European Commission (2008) Financial Services Provision and Prevention of Financial
Exclusion, European Commission, Directorate-General for Employment, Social Affairs and
Equal Opportunities Inclusion, Social Policy Aspects of Migration, Streamlining of Social
Policies, Brussels.
European Association of Cooperative Banks (EACB), Key Statistics, 2005.
FinScope (2006), FinScope Uganda [www.finscope.co.za/uganda.html]
FinScope (2008), FinScope Malawi
[www.finscope.co.za/documents/2009/Brochure_Malawi08.pdf ]
FinScope (2008), FinScope Rwanda .
[www.finscope.co.za/documents/2008/Brochure_Rwanda08.pdf ]
FinScope (2009), FinScope Pakistan
[www.finscope.co.za/documents/2009/Brochure_Pakistan.pdf ]
Guiso, L., Sapienza, P. and L. Zingales (2004), Does Local Financial Development Matter?
Quarterly Journal of Economic, Vol. 119(3), pp. 929-969, August
Karlan, D. and Zinman, J. (2009), Expanding Credit Access: Using randomized supply decisions
to estimate the impacts, Review of Financial Studies, forthcoming.
Karlan, D. and Zinman, J. (2009), Expanding Micro-enterprise Credit Access: Using
Randomized Supply Decisions to Estimate the Impacts in Manila, Mimeo, Dartmouth College.
38
Kendall, J. (2008), Local Financial Development and Growth, World Bank Policy Research
Working Paper No. 4838
La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998), Law and Finance, Journal
of Political Economy, Vol. 106(6), pp. 1113-1155, December.
Mian, A. (2006), Distance constraints: The limits of foreign lending in poor economies, Journal
of Finance, Vol. 61(3), pp. 1465-1505, June.
Petersen, M. and Rajan, R. (1995), The effect of credit market competition on lending
relationships, Quarterly Journal of Economics Vol. 110, pp. 407–443.
Rousseau, P.L. and P. Wachtel (2001), Inflation, Financial Development and Growth, Economic
Theory, Dynamics and Markets: Essays in the Honor of Ryuzo Sato, Deventer, Kluwer.
Rousseau, P.L. and Wachtel, P. (2002), Inflation thresholds and the finance–growth nexus,
Journal of International Money and Finance, Vol. 21(6), pp. 777-793, November
System of National Accounts (SNA, 1993)
Tejerina, L. and Westley, G. (2007), Financial Services for the Poor: Household Survey Sources
and Gaps in Borrowing and Saving, Sustainable Development Department Technical Papers
Series, Washington D.C. Inter-American Development Bank
Tornell, A. and Westermann F. (2003), Credit Market Imperfections in middle income countries,
NBER Working Paper 9737
World Bank (2008a), Banking the Poor: Measuring Banking Access in 54 Economies
World Bank (2008b), Payment Systems Worldwide: a Snapshot. Outcomes of the Global
Payment Systems Survey 2008
World Bank (2009), Doing Business
World Council of Credit Unions (2008), http://www.woccu.org/publications/statreport
39
Table 1: Definitions of different bank types: Commercial banks, MOFIs, SSFIs, MFIs, and
Other, as shown in the survey questionnaire.
For each type, respondents were asked: (1) For each category, provide the type of the institutions that you
supervise/regulate and (2) For each category, provide the type of the institutions that you do NOT supervise/regulate
and indicate who does.
DEFINITIONS
Commercial Banks
Banks with a full banking license. In some countries, the term
universal banks, or other terms may be used. Majority
government/State owned banks should be included in this category to
the extent that they perform a broad set of retail banking functions
Cooperatives, Credit Unions, &
Mutuals
These are financial institutions that are owned and controlled by their
members (customers).
Government Savings or
Development Banks
Specialized state owned institutions or extensions of the government
whose main purpose is to lend to support economic development
and/or to provide savings, payment, and deposit services to the
public. (Includes postal banks, government savings banks, SME
lending facilities, agriculture banks, development banks, etc.)
Microfinance Institution
Institutions whose primary business model is to lend to (and possibly
take deposits from) the poor, often using specialized methodologies
such as group lending.
Other institutions providing loans
and/or deposit services (institutions
which do not fall into other
categories)
Any other financial institutions which do not fall into the above
categories and provide standard loans and/or deposit services, for
example leasing and factoring companies
40
Table 2: Survey table requesting various data elements
Table 2: Statistics
Commercial
Banks
Cooperatives,
Credit Unions &
Mutuals
Government
Savings or
Development
Bank
Indicate the number of institutions in each category
[
]
[
]
[
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
[
[
[
[
Total
Microfinance
Institutions
Other institutions
providing loans
and/or deposit
services
]
[
]
[
]
[
]
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
[
[
[
[
[
[
[
[
]
]
]
]
]
]
]
]
]
]
]
]
[
[
[
[
]
]
]
]
[
[
[
[
]
]
]
]
[
[
[
[
]
]
]
]
[
[
[
[
]
]
]
]
(for the whole
financial system)
Deposits (please enter full numbers)
Total number of deposit accounts
Total number of checking deposit accounts
Total number of saving deposit accounts
Total number of time deposit accounts
Total value of deposit accounts
Total value of checking deposit accounts
Total value of saving deposit accounts
Total value of time deposit accounts
Total number of individuals with a deposit account
Loans (please enter full numbers)
Total number of outstanding loans
Total number of outstanding loans to nonTotal number of non-financial businesses with an
Total number outstanding loans to individuals
Total number of individuals with an outstanding
Total value of outstanding loans
Total value of outstanding loans to non-financial
Total value of outstanding loans to individuals
Retail Locations
Total number of bank branches
Number of bank branches in urban areas
Number of bank branches in rural areas
Banking Agents (Non-branch retail locations)
[
]
[
]
[
]
[
]
Payments - Please give system wide totals for the following:
Total number of debit/ATM cards in circulation
[
]
Total number of ATM machines
[
]
Total number of individuals with a debit/ATM card
[
]
Total number of Point of Service terminals
[
]
Total number of Point of Service transactions in the
[
]
[
] year
Year for which you provided the data
[
]
Comments
41
Table 3: Sample table for classifying bank and non-bank institutions into 5 main
categories.
DEFINITIONS
1. For each category, provide the
type of the institutions that you
supervise/regulate
2. For each category, provide the
type of the institutions that you do
NOT supervise/regulate and indicate
who does
[Commercial Banks, Universal
Banks]
[n/a]
Cooperatives, Credit Unions, &
Mutuals
[Credit Unions]
[Urban and Rural Cooperatives,
regulated by the Ministry of
Cooperatives]
Government Savings or
Development Bank
[Government Agricultural
Development Bank]
[Postal Bank, regulated by Ministry
of Commerce]
Microfinance Institution
[None]
[n/a]
Other institutions providing loans
and/or deposit services
[Municipal Savings and Loan
Institutions, Financing Companies
(deposit taking)]
[n/a]
Commercial Banks
42
Table 4: Supervisory arrangements for Banks and non-Banks
Commercial
Banks
MOFIs
SSOFIs
MFIs
100.0%
41.5%
41.1%
34.6%
Only some institutions in this category
supervised by bank regulator
0.0%
16.2%
7.0%
9.2%
Institutions in this category only
supervised by other regulator
0.0%
16.9%
14.0%
7.7%
No supervising authority reported for
Institutions in this category
0.0%
25.4%
38.0%
48.5%
Supervisory arrangement
Institutions in this category only
supervised by commercial bank regulator
Note: Frame is 129 respondents.
Table 5: Data available on the number of deposits and loans is limited, especially for nonbanks.
Data Available:
Data on Value of Deposits
Data on Number of Deposit Accounts
Data on Value of Loans
Data on Number of Loans
Data on Number of Individual Depositors
Data on Number of Individual Borrowers
Commercial
Banks
MOFIs
SSOFIs
MFIs
118
83
116
63
30
27
42
24
42
18
8
10
32
20
33
15
6
7
22
17
28
18
10
8
Note: The data counting the numbers of individual depositors likely reflects much double counting because most
respondents reported not being able to track individuals across multiple institutions even if they could ask banks for
their individual totals.
43
Table 6: Summary Statistics
This table presents summary statistics for our main indicators. All monetary values are expressed in 2008 US
dollars. The exact definition of each variable is shown in Appendix 1.
N
Mean
Std.
Dev.
25%
Median
75%
Commercial Banks
Accounts / Adult population
Av. Value of accounts1 / GDP per capita2
Loans / Adult population
Branches / Adult population
Branches / Sq km
85
72
65
119
119
996
2.6
299
16
33
995
5.6
304
15
138
270
0.4
42
5
1
702
0.9
214
12
6
1,498
2.4
430
23
21
Cooperatives3
Accounts / Adult population
Av. Value of accounts1 / GDP per capita2
Loans / Adult population
Branches / Adult population
Branches / Sq km
56
27
50
69
69
129
0.7
19
5
6
337
1.3
64
11
17
0
0.2
0
0
0
0.2
0.3
0
0.3
0.2
71
0.5
10
2.8
2.6
Government Banks3
Accounts / Adult population
Av. Value of accounts1 / GDP per capita2
Loans / Adult population
Branches / Adult population
Branches / Sq km
68
22
62
84
84
91
35
8.8
2.1
3.2
271
96.4
27.3
5.3
9.9
0
0.3
0
0
0
0
0.3
0
0
0
10
3.3
0.1
1.1
0.4
MFIs3
Accounts / Adult population
Av. Value of accounts1 / GDP per capita2
Loans / Adult population
Branches / Adult population
Branches / Sq km
70
15
70
77
77
9.3
0.8
7.9
1.1
1.6
28.2
1.7
18.5
3.1
8.7
0
0.1
0
0
0
0
0.2
0
0
0
0.3
0.5
1.8
0.7
0.2
1.
2.
3.
The average value of accounts is calculated as the (Total Value of Deposit Accounts / Number of Accounts)
To facilitate comparisons, the statistics are for the sample of countries with available data on the number of
accounts in commercial banks.
Includes countries where the cooperative sector (or government or microfinance sector, respectively) is not
regulated / supervised by any financial regulator (In those cases, the corresponding figures are zeros).
44
Table 7: Prediction models of the number of deposit accounts in commercial banks
This table shows the estimates of OLS regressions of the number of deposit accounts in commercial banks per
thousand adults on several covariates. Each column is represents a different regression. The dependent variable in
each regression is ln(Number of accounts in commercial banks/1000 adults).
ln(GDP/capita)
High Inc. OECD (dummy)
ln(Population Density)
ln(Private Credit/GDP)
ln(Branches/100k Adults)
ln(Value Deposits/Population)
Constant
Observations
R-squared
Model
(1)
0.41**
(0.16)
-0.59**
(0.29)
0.090
(0.062)
0.18
(0.13)
0.53***
(0.10)
0.041
(0.12)
1.48*
(0.87)
62
0.824
Model
(2)
0.44***
(0.084)
-0.55*
(0.28)
0.14**
(0.053)
0.17
(0.11)
0.56***
(0.093)
1.21
(0.77)
73
0.824
Model
(3)
0.63***
(0.18)
-0.67*
(0.34)
0.049
(0.073)
0.33**
(0.15)
Model
(4)
0.68***
(0.090)
-0.61*
(0.33)
0.12*
(0.064)
0.35***
(0.13)
Model
(5)
0.83***
(0.069)
-0.62*
(0.33)
0.17***
(0.058)
0.062
(0.14)
1.11
(1.04)
64
0.738
0.84
(0.94)
75
0.733
-1.06*
(0.60)
80
0.708
Note: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1 In these regressions and
subsequent predictions, we leave out countries with fewer than 100,000 adult populations and extreme
outliers.
45
Table 8: Data sources and models used in the estimation of the number of accounts in
commercial banks
This table shows the distribution of data sources and models used to estimate the number of accounts in commercial
banks around the world. The third column represents the percentage in terms of countries. The last column
represents the percentage in terms of accounts.
Data Source
FA Survey '09 (Reported Data)
BTP '08; BDM '07, CGAP
internal numbers*
Model 1 (branches volume)
Model 2 (branches)
Model 3 (volume)
Model 4 (with private credit)
Model 5 (without private credit)
Totals
# countries
91
% sample
59%
% final count
48%
25
16%
22%
17
2
6
7
6
154
11%
1%
4%
5%
4%
100%
9%
0%
5%
1%
14%
100%
* BTP = Banking the Poor (2008); BDM = Beck, Demirgüç-Kunt, and Martinez-Peria (2007), CGAP
numbers were gathered in email exchanges between other CGAP staff and national regulator staff.
Table 9: Prediction model of the number of deposit accounts in mutually owned institutions
This table shows the estimates of OLS regressions of the number of deposit accounts in mutually owned institutions
(cooperatives, credit unions, and Mutuals) per thousand adults on several covariates. The dependent variable in each
regression is ln(Number of accounts in mutually owned institutions/1000 adults).
ln(GDP/capita)
ln(Value Deps./Population)
Constant
Observations
R-squared
-0.93***
(0.20)
0.97***
(0.11)
7.44***
(1.49)
22
0.804
Note: Standard errors in parentheses; ***
p<0.01, ** p<0.05, * p<0.1
46
Table 10: Data sources used in the estimation of the number of MOFI accounts
This table shows the distribution of data sources used to estimate the number of accounts in mutually owned
financial institutions (MOFI) around the world. The third column represents the percentage in terms of countries.
The last column represents the percentage in terms of accounts.
Data Source
FA Survey '09
WOCCU
Predictions
Set to 25th percentile
EACB
CGAP internal numbers
Totals
# countries
56
52
15
21
13
1
158
% sample
35%
33%
9%
13%
8%
1%
100%
% final
count
56%
3%
6%
2%
13%
20%
100%
Notes: WOCCU is World Council of Credit Unions, ECBA is European Cooperative
Banking Association, 25th percentile used only in "preferred estimate" 50th
percentile used in "high estimate"; predictions based on regression, and CGAP
internal numbers are for Japan and come from direct communication with regulator.
Table 11: Data sources used in the estimation of the number of SSFI accounts
This table shows the distribution of data sources used to estimate the number of accounts in SSFIs around the world.
The third column represents the percentage in terms of countries. The last column represents the percentage in terms
of accounts.
Data Source
FA Survey '09
CGAP "Big Numbers" *
Totals
# countries
68
141
209
% sample
33%
67%
100%
% final
count
78%
22%
100%
Notes: CGAP "Big Numbers" refers to data gathered for CCGAP, Occasional Paper
no. 8, 2004
47
Table 12: Data sources used in the estimation of the number of MFI accounts
This table shows the distribution of data sources used to estimate the number of accounts in MFIs around the world.
The third column represents the percentage in terms of countries. The last column represents the percentage in terms
of accounts.
Data Source
FA Survey '09
Microfinance Exchange (MIX)
Totals
#
countries
14
75
89
% final
count
44%
56%
100%
% sample
16%
84%
100%
Note: MIX numbers are values voluntarily reported by MFIs and aggregated to
country level.
Table 13: Estimates for numbers of accounts by income group and by category of
providing instituion.
This table shows the estimation for the number of accounts by income group and category of providing institution.
Countries are divided according to the World Bank income group definition. The last column represents the
percentage of accounts in non-bank institutions (MOFIs + SSFIs + MFIs).
Commercial
Banks
Low Estimate
Preferred Estimate
High Estimate
Low Estimate
Preferred Estimate
High Estimate
Low Estimate
Preferred Estimate
High Estimate
MOFI
SSFI
MFIs
Panel A: World
4.75 Bn
0.68 Bn
0.34 Bn
0.01 Bn
5.01 Bn
0.71 Bn
0.44 Bn
0.02 Bn
5.27 Bn
0.82 Bn
n.a.
n.a.
Panel B: Middle income and developing countries
2.05 Bn
0.16 Bn
0.34 Bn
0.01 Bn
2.22 Bn
0.19 Bn
0.41 Bn
0.02 Bn
2.39 Bn
0.25 Bn
n.a.
n.a.
Panel C: High Income countries
2.67 Bn
0.52 Bn
0.00 Bn
0.00 Bn
2.79 Bn
0.53 Bn
0.03 Bn
0.00 Bn
2.91 Bn
0.57 Bn
n.a.
n.a.
Total
%
NonBank
5.78 Bn
6.18 Bn
6.55 Bn
18.99%
2.56 Bn
2.84 Bn
3.08 Bn
21.76%
3.19 Bn
3.35 Bn
3.51 Bn
16.67%
Note: Value per 1000 adults numbers have as base, population from all countries (including those which we have
no estimate of bank accounts).
48
Table 14: Reported rates of formal account ownership from household surveys and the
logarithm of deposit penetration.
This table shows the results of the OLS regressions of reported rates of formal account ownership from household
surveys and the logarithm of deposit accounts per thousand adults. Column (1) is surveys taken after 2003, (2) are
surveys after 2003 where the reported fraction is of all adults (rather than households) (3) is all available household
surveys (going back to 2000).
ln(Deposits./ 1000 Adults)
Constant
Observations
R-squared
(1)
25.2***
(7.51)
-112.2***
(-4.95)
25
0.7101
(2)
24.5***
(5.73)
-105.4***
(-3.51)
17
0.6861
Note: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
49
(3)
24.8***
(6.64)
-115.1***
(-4.65)
31
0.6034
Table 15: Household survey sources, years and countries
Some of these data points are not used in the calculation of the conversion factor because no reported or predicted
values for account numbers were available.
Country
Survey Source
Year
Austria
Belgium
Savings
Account
%
97
99
Base
Conversion
factor
(see note)
R, PR, CB
R, PR, CB
European Commission
2008
All Adults
European Commission
2008
All Adults
Claessens (2006) Botswana
FINSCOPE
2003
47
Household
Cyprus
European Commission
2008
82
All Adults
Czech Republic
European Commission
2008
83
All Adults
R, PR, CB
Denmark
European Commission
2008
99
All Adults
PR
Estonia
European Commission
2008
84
All Adults
R, PR, CB
Finland
European Commission
2008
94
All Adults
PR
France
European Commission
2008
98
All Adults
PR
Germany
European Commission
2008
97
All Adults
PR
Greece
European Commission
2008
72
All Adults
R, PR, CB
Hungary
European Commission
2008
66
All Adults
PR, CB
India (UP and AP)
Claessens (2006) - RFAS
2003
47.5
Household
Ireland
European Commission
2008
88
All Adults
PR
Italy
European Commission
2008
84
All Adults
R, PR, CB
Latvia
European Commission
2008
52
All Adults
R, PR, CB
Claessens (2006) Lesotho
FINSCOPE
2003
17
Household
Lithuania
European Commission
2008
59
All Adults
R, PR, CB
Luxembourg
European Commission
2008
99
All Adults
PR
Malawi
FINSCOPE
2008
17
All Adults
PR, CB
Malta
European Commission
2008
79
All Adults
Claessens (2006) Namibia
FINSCOPE
2003
28.4
Household
Netherlands
European Commission
2008
99
All Adults
R, PR, CB
Nigeria
FINSCOPE
2008
13.9
All Adults
PR
Pakistan
FINSCOPE
2008
3
All Adults
R, PR, CB
Panama
Tejerina and Westley (2007)
2003
35.2
Household
Poland
European Commission
2008
60
All Adults
R, PR, CB
Portugal
European Commission
2008
83
All Adults
PR
Rwanda
FINSCOPE
2008
14
All Adults
R, PR, CB
Slovakia
European Commission
2008
74
All Adults
PR
Slovenia
European Commission
2008
94
All Adults
R, PR, CB
Claessens (2006) South Africa
FINSCOPE
2004
46
Household
Spain
European Commission
2008
92
All Adults
R, PR, CB
Claessens (2006) Swaziland
FINSCOPE
2003
35.3
Household
Sweden
European Commission
2008
98
All Adults
PR
Uganda
FINSCOPE
2007
19
All Adults
R, PR, CB
UK
European Commission
2008
94
All Adults
PR
*Note: R = used in "reported totals only" conversion factor calculation, RP = "reported and predicted" calculation,
CB = "commercial banks only" calculation
50
Table 16: Computation methods of the “conversion factor”
This table shows the results of the so-called conversion factor using three different methodologies. The conversion
factor is calculated, for each country for which household data are available, as the total number of deposit accounts
divided by the number of individuals who have savings/deposit accounts in formal financial institutions. This factor
is a proxy for the average number of accounts per account holder. Such proxy is not perfect, since the figure for the
total number of accounts include not only individual accounts, but also business and government accounts. For
reported totals only, the sample was restricted to countries with reported account data and household survey data. In
the reported and predicted row, we use countries with household survey data and account data, regardless of
whether it were reported or predicted data. Finally, in the last row, labeled comm. bank accounts, we use a sample of
countries with household survey data and commercial bank reported account data. The columns present the mean, as
well as the 90% confidence interval.
Conversion
Factor
Conversion Factor Calculation
Method
Using reported totals only
Using reported and predicted
Using reported comm. bank accts.
# Obs.
15
28
17
Mean
3.0
3.2
2.5*
90% C.I.
2.0
2.6
1.6
+
3.9
3.9
3.4
Notes: *The conversion factor from reported commercial bank accounts is not necessarily comparable to the
others since it will use generate the number of banked adults from the number of commercial bank accounts
only.
Table 17: Estimation of the number of banked people using the three different conversion
factors
This table shows the estimation of the banked population using three different methodologies to estimate the socalled conversion factor. The conversion factor is calculated, for each country for which household data are
available, as the total number of deposit accounts divided by the number of individuals who have savings/deposit
accounts in formal financial institutions. For reported totals only, the sample was restricted to countries with
reported account data and household survey data. In the reported and predicted row, we use countries with
household survey data and account data, regardless of whether it were reported or predicted data. Finally, in the last
row, labeled comm. bank accounts, we use a sample of countries with household survey data and commercial bank
reported account data.
Conversion Factor Calculation
Method
Using reported totals only
Using reported and predicted
Using reported comm. bank accounts
Region
World
High Income
Developing
World
High Income
Developing
World
High Income
Developing
51
Estimate for
Number Banked
1.78 Bn
0.70 Bn
1.08 Bn
1.91 Bn
0.74 Bn
1.17 Bn
1.71 Bn
0.65 Bn
1.06 Bn
% Adult Pop.
Banked
38%
81%
28%
41%
86%
31%
37%
75%
28%
Table 18: Prediction model for the number of loans per thousand adults
This table shows the estimates of OLS regressions of the number of loans per thousand adults on a quadratic
function of the natural logarithm of GDP per capita and the 2008 Credit Information Index of Doing Business. Our
preferred specification is presented in column (1).
(1)
3.18***
(0.65)
-0.14***
(0.04)
ln(GPD per capita)
ln(GPD per capita)2
ln(Value of Loans per 1000 adults)
ln(Value of Loans per 1000 adults) 2
Credit Information Index
0.09**
(0.04)
-11.57***
(2.50)
69
0.817
Constant
Observations
Adjusted R-squared
(2)
2.31***
(0.29)
--0.10***
(0.02)
0.07*
(0.04)
-6.04***
(1.04)
58
0.867
Note: Standard errors in parentheses; *** p<0.01, ** p<0.05, * p<0.1
52
Table 19: Cross-country correlations between measures of deposit account penetration,
loan penetration, and demographic penetration of branches and country characteristics
This table summarizes the pairwise correlations between measures of deposit account penetration, loan penetration,
and demographic penetration of branches and different country characteristics. Each cell displays the pairwise
correlation between the variable at the top of the column and the covariate in the corresponding row. The definition
of each variable is shown in Appendix 1. ***= significant at the 1% level; ** = significant at the 5% level; * =
significant at the 10% level.
GDP per capita (log)
Population Density (log)
Gini
Inflation
Deposit Insurance
Electricity consumption
Phone lines
Road density
Absence of violence
Offshore
Concentration
Government bank share
Foreign bank share
KYC requirements
Basic Accounts (Exc.)
Basic Accounts (Req.)
Documents required
Min. Bal. Open checking
Checking annual fee
Tax incentives
Operations agents (savings)
Postal Network
Contract enforcement
Creditor Rights Index
Creditor Inf. Index
Places to submit loan
Minimum loan
Fee consumer loan
Days process loan
Operations agents (credit)
Branch approval
Accounts per
thousand adults
in commercial
banks (logs)
0.80***
0.25**
-0.16
-0.46***
0.46***
0.74***
0.84***
0.53***
0.49***
0.21**
-0.28**
-0.18
-0.02
-0.18*
-0.17
0.11
-0.22
-0.44***
-0.50***
0.28***
0.20*
0.35***
-0.47***
0.37***
0.48***
0.51***
-0.32**
-0.17
-0.27*
0.17
-0.08
Loans per
thousand adults in
commercial banks
(logs)
0.86***
-0.04
0.16
-0.24**
0.51***
0.80***
0.85***
0.42***
0.46***
0.20
-0.35***
-0.19
-0.03
-0.05
-0.03
0.13
-0.05
-0.41***
-0.44**
0.25**
0.20
0.28**
-0.32***
0.31**
0.60***
0.63***
-0.34*
0.06
-0.24
0.22*
-0.09
53
Branches per
thousand adults in
commercial banks
(logs)
0.77***
0.08
-0.25**
-0.30***
0.38***
0.70***
0.79***
0.49***
0.53***
0.18**
-0.18*
-0.21*
-0.03
-0.24***
-0.16
0.18**
-0.07
-0.46***
-0.58***
0.25***
0.11
0.29***
-0.43***
0.25***
0.52***
0.52***
-0.36**
-0.02
-0.21
0.14
-0.28***
Branches per
squared km in
commercial banks
(logs)
0.53***
0.79***
-0.45***
-0.31***
0.30***
0.44***
0.59***
0.82***
0.32***
0.37***
-0.23**
-0.11
-0.04
-0.15*
0.07
0.10
-0.28*
-0.33***
-0.41***
0.12
0.11
0.31***
-0.28***
0.28***
0.32***
0.25*
-0.14
-0.18
-0.06
0.10
-0.08
Table 20: Cross-country covariates associated with deposit account penetration
This table summarizes the results of OLS regressions of the (logarithm of) number of accounts per thousand adults in commercial banks on different country
characteristics. Each column in each panel represents the result of one regression. The definition of each variable is shown in Appendix 1. Robust standard errors
are in parentheses. ***= significant at the 1% level; ** = significant at the 5% level; * = significant at the 10% level.
GDP per capita (log)
(1)
(2)
(3)
(4)
(5)
(6)
(7)
(8)
(9)
(10)
(11)
(12)
0.68***
(0.06)
0.66***
(0.06)
0.15*
(0.08)
0.75***
(0.08)
0.35***
(0.10)
0.00
(0.01)
0.63***
(0.06)
0.14
(0.08)
0.62***
(0.06)
0.16*
(0.08)
0.40***
(0.12)
0.13*
(0.07)
0.21*
(0.10)
0.13*
(0.07)
0.69***
(0.11)
0.16
(0.12)
0.63***
(0.06)
0.15*
(0.08)
0.66***
(0.06)
0.14
(0.08)
0.42***
(0.08)
0.14**
(0.06)
0.39***
(0.09)
-0.30*
(0.15)
Population Density (log)
Gini
Inflation
-0.02***
(0.01)
Deposit Insurance
0.34**
(0.16)
Electricity consumption
0.25**
(0.13)
Phone lines
0.47***
(0.11)
Road density
-0.06
(0.13)
Absence of violence
0.09
(0.10)
Offshore
0.20
(0.31)
Branches per adult (log)
0.45***
(0.14)
Branches per km2 (log)
Constant
N
R2
0.90
(0.54)
85
0.64
0.37
(0.69)
85
0.67
-1.13
(1.34)
38
0.67
0.81
(0.71)
81
0.70
0.49
(0.66)
84
0.68
54
0.75
(0.75)
72
0.64
3.05***
(0.80)
84
0.75
0.29
(1.02)
59
0.71
0.63
(0.69)
85
0.67
0.44
(0.72)
85
0.67
1.33**
(0.59)
83
0.72
0.44***
(0.13)
3.78***
(1.08)
83
0.72
Table 20 (Cont): Cross-country covariates associated with deposit account penetration
GDP per capita (log)
Population Density (log)
Concentration
(12)
0.61***
(0.05)
0.13*
(0.08)
-0.63
(0.43)
Government bank share
(13)
0.58***
(0.07)
0.13**
(0.06)
(14)
0.58***
(0.07)
0.19***
(0.06)
(15)
0.66***
(0.07)
0.15*
(0.08)
(16)
0.70***
(0.08)
0.09
(0.09)
(17)
0.66***
(0.06)
0.15*
(0.08)
(18)
0.81***
(0.15)
0.19*
(0.11)
(19)
0.59***
(0.06)
0.15***
(0.05)
(20)
0.61***
(0.08)
0.24***
(0.08)
(21)
0.68***
(0.07)
0.15*
(0.08)
(22)
0.66***
(0.06)
0.15*
(0.08)
0.16
(0.35)
Foreign bank share
0.19
(0.26)
KYC requirements
-0.02
(0.04)
Basic Accounts (Exc.)
0.18
(0.21)
Basic Accounts (Req.)
0.12
(0.17)
Documents required
-0.00
(0.13)
Min. Bal. Open checking
0.00
(0.01)
Checking annual fee
-0.01
(0.01)
Tax incentives
-0.17
(0.16)
Operations agents (savings)
0.09
(0.21)
Postal Network
Constant
N
R2
(23)
0.64***
(0.07)
0.13
(0.08)
1.31*
(0.69)
82
0.67
1.16*
(0.61)
56
0.72
0.84
(0.62)
56
0.72
0.49
(0.83)
85
0.67
0.23
(0.85)
66
0.67
0.37*
(0.69)
85
0.67
55
-0.92
(1.64)
32
0.73
0.97
(0.57)
51
0.77
0.47
(0.90)
36
0.78
0.24
(0.70)
85
0.67
0.40
(0.70)
85
0.67
0.21
(0.17)
0.53
(0.73)
85
0.67
Table 21: Cross-country covariates associated with the loan penetration
This table summarizes the results of OLS regressions of the (logarithm of) number of loans per thousand adults in commercial banks on different country
characteristics. Each column in each panel represents the result of one regression. The definition of each variable is shown in Appendix 1. Robust standard errors are in
parentheses. ***= significant at the 1% level; ** = significant at the 5% level; * = significant at the 10% level.
GDP per capita (log)
(1)
1.05***
(0.09)
Pop. Density (log)
(2)
1.05***
(0.09)
-0.05
(0.08)
Gini
(3)
0.99***
(0.09)
(4)
1.04***
(0.09)
(5)
0.57**
(0.21)
(6)
0.60***
(0.19)
(7)
1.01***
(0.13)
(8)
1.08***
(0.11)
(9)
1.04***
(0.10)
(10)
1.02***
(0.09)
(11)
0.93***
(0.10)
(12)
0.66***
(0.10)
0.01
(0.01)
Inflation
-0.01
(0.01)
Elect. consumption
0.40*
(0.22)
Phone lines
0.46***
(0.17)
Road density
0.06
(0.16)
Contract enforcement
0.00
(0.00)
Absence of violence
0.06
(0.16)
Creditor Rights Index
0.07**
(0.04)
Creditor Inf. Index
0.14***
(0.05)
Branches per adult (log)
0.66***
(0.13)
Branches per km2 (log)
Constant
N
R2
(13)
0.97***
(0.11)
-3.4***
(0.77)
65
0.73
-3.2***
(0.78)
65
0.74
-3.0***
(0.84)
33
0.75
-3.2***
(0.77)
64
0.74
-2.3**
(0.91)
53
0.69
-0.7
(1.30)
64
0.77
56
-3.4***
(0.96)
42
0.74
-3.8***
(1.00)
65
0.74
-3.2***
(0.82)
65
0.74
-3.6***
(0.75)
65
0.75
-2.9***
(0.79)
65
0.76
-1.81**
(0.71)
64
0.80
0.10
(0.08)
-2.53*
(0.99)
64
0.74
Table 21 (Cont): Cross-country covariates associated with the loan penetration
GDP per capita (log)
Concentration
(12)
1.00***
(0.08)
-1.40***
(0.46)
Government bank share
(13)
0.98***
(0.12)
(14)
0.93***
(0.13)
(15)
0.96***
(0.15)
(16)
0.92***
(0.15)
(17)
0.92***
(0.15)
(18)
0.92***
(0.14)
0.03
(0.91)
Foreign bank share
-0.28
(0.43)
Places to submit loan
-0.05
(0.10)
Minimum loan
0.00
(0.00)
Fee consumer loan
-0.06
(0.09)
Days process loan
0.01
(0.02)
Operations agents (credit)
Constant
N
R2
(19)
1.04***
(0.09)
-2.1***
(0.72)
64
0.76
-2.7***
(1.01)
42
0.73
-2.1*
(1.12)
39
0.74
-2.4***
(1.07)
32
0.76
-2.2*
(1.25)
30
0.75
57
-2.1
(1.32)
27
0.73
-2.2*
(1.15)
31
0.76
0.38
(0.39)
-3.4***
(0.76)
65
0.74
Table 22: Cross-country covariates associated with the demographic penetration of branches
This table summarizes the results of OLS regressions of the (logarithm of) number of branches per thousand adults in commercial banks on different country
characteristics. Each column in each panel represents the result of one regression. The definition of each variable is shown in Appendix 1. Robust standard errors
are in parentheses. ***= significant at the 1% level; ** = significant at the 5% level; * = significant at the 10% level.
GDP per capita (log)
Population Density (log)
(1)
0.51***
(0.03)
0.02
(0.05)
Gini
(2)
0.58***
(0.07)
0.06
(0.07)
-0.02*
(0.01)
Inflation
(3)
0.50***
(0.04)
0.01
(0.05)
(4)
0.30***
(0.08)
0.02
(0.05)
(5)
0.22***
(0.07)
-0.00
(0.05)
(6)
0.48***
(0.06)
-0.04
(0.08)
(7)
0.48***
(0.03)
0.02
(0.05)
(8)
0.49***
(0.05)
0.02
(0.05)
(9)
0.49***
(0.03)
0.01
(0.05)
-0.01
(0.01)
Electricity consumption
0.22**
(0.09)
Phone lines
0.34***
(0.07)
Road density
0.13
(0.09)
Contract enforcement
-0.00
(0.00)
Absence of violence
0.04
(0.09)
Concentration
-0.16
(0.40)
Branch approval
Constant
N
R2
(10)
0.49***
(0.04)
0.02
(0.05)
-2.0***
(0.41)
119
0.59
-1.6
(1.01)
49
0.56
-1.8***
(0.44)
115
0.60
-1.8***
(0.50)
102
0.54
-0.3
(0.53)
118
0.66
58
-2.0***
(0.56)
84
0.68
-1.7***
(0.47)
119
0.60
-1.9***
(0.49)
118
0.60
-1.7***
(0.57)
115
0.58
-0.02
(0.16)
-1.8***
(0.46)
116
0.57
Table 23: Cross-country covariates associated with the geographic penetration of branches
This table summarizes the results of OLS regressions of the (logarithm of) number of branches per squared km in commercial banks on different country
characteristics. Each column in each panel represents the result of one regression. The definition of each variable is shown in Appendix 1. Robust standard errors
are in parentheses. ***= significant at the 1% level; ** = significant at the 5% level; * = significant at the 10% level.
GDP per capita (log)
Population Density (log)
(1)
0.58***
(0.03)
1.03***
(0.05)
Gini
(2)
0.67***
(0.08)
1.06***
(0.07)
-0.02**
(0.01)
Inflation
(3)
0.57***
(0.04)
1.02***
(0.05)
(4)
0.31***
(0.09)
1.03***
(0.06)
(5)
0.24***
(0.07)
1.00***
(0.05)
(6)
0.54***
(0.06)
0.95***
(0.08)
(7)
0.55***
(0.04)
1.03***
(0.05)
(8)
0.56***
(0.05)
1.03***
(0.05)
(9)
0.56***
(0.03)
1.02***
(0.05)
-0.01
(0.01)
Electricity consumption
0.28***
(0.10)
Phone lines
0.41***
(0.07)
Road density
0.16
(0.10)
Contract enforcement
-0.00
(0.00)
Absence of violence
0.04
(0.10)
Concentration
-0.23
(0.40)
Branch approval
Constant
N
R2
(10)
0.56***
(0.04)
1.02***
(0.05)
-7.6***
(0.44)
119
0.86
-7.0***
(1.10)
49
0.77
-7.4***
(0.48)
115
0.86
-7.4***
(0.52)
102
0.86
-5.6***
(0.55)
118
0.89
59
-7.5***
(0.60)
84
0.87
-7.2***
(0.50)
119
0.86
-7.5***
(0.52)
118
0.87
-7.2***
(0.59)
115
0.86
-0.03
(0.17)
-7.4***
(0.49)
116
0.86
Figure 1: Added variable plot from Model (1) in Table 8.
100
This figure shows the line of best fit from Model (1) in Table 8, namely, the OLS regression between the rates of
formal account ownership from household surveys and the logarithm of deposit accounts per thousand adults. The
estimation only takes into accounts surveys taken after 2003.
NLD
ESP
% Households w Savings
40
60
80
ITA
BEL
AUT
EST
CZE
GRC
HUN
POL
IDN
BWA
SWZ
LTU
LVA
IND ZAF
PAN
NAM
MWI
UGA
LSO
RWA
PAK
0
20
SVN
5
6
7
Ln(Deposits/1000 Adults)
60
8
Figure 2: Numbers of commercial bank accounts from Financial Access Survey against
numbers of accounts in deposit money banks collected in 2008 and 2003
This figure shows the correlation between the number of accounts per thousand adults collected in the Financial
Access Survey and previous data. The left panel shows the correlation with the same figure as collected by the
World Bank (2008). The right panel presents the correlation with the numbers collected by Beck, Demirgüç-Kunt
and Martinez-Peria (2008). The line represents the OLS regression.
61
Figure 3: Deposit penetration and income per capita (Actual and predicted values)
This figure shows the actual and predicted values (Model 1) of accounts in commercial banks per thousand adults
and income per capita. The blue dots represent the actual values. The red crosses represent the predicted values.
Figure 4: Predicted and actual values using Model 1.
This figure shows the actual and predicted values (Model 1) of accounts in commercial banks per thousand adults.
Spain and Italy are outliers due to their large cooperative sectors, which function as banks but are not counted in
commercial bank data.
62
Figure 5: Number of deposit accounts in banks and regulated non-bank financial
institutions per thousand adults
This figure shows the worldwide distribution of deposit accounts in banks and regulated non-bank financial
institutions per thousand adults. Predicted values are used when data are not available.
Figure 6: Number of bank loans per thousand adults in commercial banks
This figure shows the worldwide distribution of deposit accounts in banks and regulated non-bank financial
institutions per thousand adults. Predicted values are used when data are not available.
63
Appendix 1: Variables
Variable
Description
Source
GDP per capita
Population Density
Adult Population
Population
Branches per adult
Branches per sq km
Gini
Inflation
Gross Domestic Product per capita in current dollars of 2007
Total number of people in 2008 divided by the area of land measured in km²
Adult population in 2008. When the 2008 data are not available, we use the most recent
Total population in 2008. When the 2008 data are not available, we use the most recent
Number of branches per 100,000 adults
Number of branches per squared km
Scored on zero to six scale; scores increasing with scope, access, and quality of credit
information
Gini coefficient for income inequality in each country (5-year average 2003-2007)
10 year average (1998-2008) of the change in the CPI in each country
Deposit Insurance
Equals one if the country had an explicit deposit insurance in 2003 and zero otherwise.
Electricity Consumption
Phone Lines
Road Density
Absence of Violence
Offshore
Concentration
Government Bank Share
Foreign Bank Share
Watts-hour consumption per capita in 2006
Total telephone mainlines per thousand people (5-year average 2003-2007)
Kilometers of roads per 100 km² of land area (5-year average 2003-2007)
Sub-Index of Political Stability / No Violence, 2008
Equals one if the country was defined by the IMF as an offshore center in 2008
Share of deposits in the five largest banks
Percentage of banking system assets in banks 50% + owned by government
Percentage of banking system assets in banks 50% + owned by foreign entities
An index aggregating the documentation required to open a checking account. This
includes: (1) Government issued ID, (2) Any ID, (3) Proof of nationality or legal status, (4)
Proof of address, (5) Proof of income, (6) Proof of employment, and (7) Other
Equals one if there were regulatory exceptions for low income people in 2008 to the
documentation requirements for opening a bank checking account
Equals one if there are, among the policies to promote access to finance, regulatory
requirements for banks to offer a basic or low fee account.
The average number of documents (in the 5 largest banks) required to open a checking
account in 2008.
Minimum balance required to open a checking account expressed as a percent of GDP. It
combines two databases. When data were available in both data sets, the variable takes the
latest value.
World Development Indicators
World Development Indicators
World Development Indicators
World Development Indicators
Financial Access Database (2009)
Financial Access Database (2009)
World Bank Doing Business
Indicators, 2009
World Development Indicators
International Financial Statistics
Demirgüç-Kunt., Karacaovali, and
Laeven (2005)
World Development Indicators
World Development Indicators
World Development Indicators
World Bank Governance Indicators
IMF
Barth, Caprio, and Levine (2004)
Barth, Caprio, and Levine (2004)
Barth, Caprio, and Levine (2004)
Credit Information Index
KYC Requirements
Basic Accounts (Exc.)
Basic Accounts (Req.)
Documents Required
Min. Bal. open checking
64
Financial Access Database (2009)
Financial Access Database (2009)
Financial Access Database (2009)
World Bank, Banking the Poor
(2008)
Beck, Demirgüç-Kunt., and
Martinez-Peria (2008) and World
Bank, Banking the Poor (2008)
Appendix 1 (Cont): Variables
Variable
Checking annual fee
Tax incentives
Operations agents (savings)
Postal Network
Contract Enforcement
Description
Fees associated with maintaining a checking account expressed as percent of GDP per
capita in 2004
Equals one if a country claims that it has pursued tax incentive savings schemes (such as
tax incentives for retirement, education, or medical savings) to promote financial access
and zero otherwise.
This index is an equally weighted mean of responses to the following survey questions: (1)
Agents are allowed to receive and forward applications to open accounts; (2) Agents are
allowed to open accounts on behalf of the bank; (3) Agents to accept funds for deposit in
the client's bank account and; (4) Agents to pay withdrawals from client's bank account.
Equals one if financial services are offered in the post offices and these services are
handled by a separate private operator (typically a bank) and zero otherwise.
Total enforcement cost, including legal fees, assessment and court fees expressed as
percentage of total debt in 2008
Creditor Rights Index
Index of creditor rights following La Porta, Lopez-de-Silanes, Shleifer and Vishny (1998)
Minimum loan
Lowest amount of consumer loan banks make expressed as a percent of GDP per capita in
2004
Fee consumer loan
Fee banks charge on consumer loans expressed as percent of GDP per capita in 2004
Days process loan
Number of days banks take to process a typical consumer loan application in 2004
Operations agents (credit)
Branch Approval
This index is an equally weighted mean of responses to the following survey questions: (1)
Agents to receive and forward loan requests to the bank; (2) Agents to conduct credit
evaluations and to approve loans on behalf of the bank; (3) Agents to collect loan payments
on behalf of the bank.
Equals one if the Supervisor/Regulator approval is required to open each bank branch and
zero otherwise in 2008.
65
Source
Beck, Demirgüç-Kunt., and
Martinez-Peria (2008)
Financial Access Database (2009)
Financial Access Database (2009)
Financial Access Database (2009)
World Bank Doing Business
Indicators, 2008
World Bank Doing Business
Indicators, 2008
Beck, Demirgüç-Kunt., and
Martinez-Peria (2008)
Beck, Demirgüç-Kunt., and
Martinez-Peria (2008)
Beck, Demirgüç-Kunt., and
Martinez-Peria (2008)
Financial Access Database (2009)
Financial Access Database (2009)