Access to Financial Services and the Financial Inclusion Agenda Around the World: A Cross-Country Analysis with a New Data Set

Recent empirical evidence highlights that access to basic financial services can make a substantial positive difference in improving poor people's lives. Accordingly, financial sector reforms that promote financial inclusion are increasingly at the core of policymakers? agendas. The Consultative Group to Assist the Poor and the World Bank Group, in response, launched the Financial Access project, including a cross-country database on financial inclusion topics and an annual report to inform the policy debate. Using this database, this paper (i) counts the number of unbanked adults around the world at 56 percent, (ii) analyzes the state of access to deposit and loan services as well as the extent of retail networks, and (iii) discusses the state of financial inclusion mandates around the world.


Introduction
The strong relationship between financial development and economic growth is well documented in the literature (see, for example, King and Levine, 1993;Beck et al., 2000;Demirgüç-Kunt and Maksimovic, 1998;Beck et al., 2004;Levine, 2005;Klapper et al., 2006;Demirgüç-Kunt et al., 2008). In more recent years the debate expanded to include the notion of financial "exclusion" as a barrier to economic development and the need to build inclusive financial systems (Beck et al., 2008).
Recent empirical evidence using household data indicates that access to basic financial services such as savings, payments and credit can make a substantial positive difference in improving poor people's lives (Caskey et al., 2006;Dupas and Robinson 2009). For firms, especially small and medium enterprises (SMEs), access to finance is often the main obstacle to growth (Schiffer and Weder, 2001;Cressy, 2002;IADB, 2004;and Beck et al., 2005and Beck et al., , 2006and Beck et al., , and 2008. Accordingly, financial sector reforms that promote financial inclusion are increasingly at the core of the international development agenda for policy makers and development institutions at the global level. The United Nations (UN) has declared 2005 the Year of Microfinance, and the recent Pittsburg and Korea G-20 communiqués increasingly underscore the importance that this topic has gained in the international arena. The Nobel Institute awarded the Nobel Peace Prize to the founders of microfinance, Muhammad Yunus and the Grameen Bank, in 2006. New international bodies, such as the Alliance for Financial Inclusion (AFI) have emerged whose primary objective is to advance financial inclusion for the world's poor. The International Monetary Fund (IMF) and the International Finance Corporation (IFC) also increasingly pay attention to this debate. The IMF has launched a new database on financial inclusion, 1 and the IFC together with Consultative Group to Assist the Poor (CGAP) and AFI have been leading the G-20 discussion around financial inclusion for households and SMEs. The Bill & Melinda Gates Foundation has just pledged $500 million per year over the next five years to expand access to saving services.
Against this backdrop, the World Bank Group (WBG) has initiated the Financial Access indicators and reports to respond to a higher demand for data and measurement of financial inclusion. This initiative aims to fill the gap in the data landscape by collecting supply-side data on financial access as well as data on financial inclusion policies and regulations from financial regulators around the world. The pilot round of surveys was implemented and published as Banking the Poor in 2008, which was based on data from both regulator and bank surveys, covering 54 countries around the world with a strong focus on Sub-Saharan African countries. Subsequently, CGAP and the WBG launched Financial Access 2009, the first in an annual series of reports to inform policy debate and monitor statistics of financial access worldwide. Financial Access 2009 introduced statistics on the use of financial services in 139 countries and mapped a broad range of policies and initiatives supporting financial inclusion, focusing on data collection and measurement challenges. Building on this, Financial Access 2010 updated statistics on the use of financial services and analyzed changes that took place in 2009 -a turbulent year for the financial sector in most countries around the world. As more countries provided more and higher quality data after the 2009 report, the focus shifted to the analysis of financial inclusion policies around the world (see World Bank, 2008;CGAP andthe World Bank, 2009 and2010).
This paper uses the Financial Access database to count the number of unbanked individuals in the world, and analyze the change of access to formal financial services around the world. Next, it reviews the role of the main financial regulator in relation to the implementation of financial inclusion reforms and mandates to give a broad global picture of the state of financial inclusion policies.
Our findings indicate that 56 percent of adults worldwide are unbanked, although the numbers differ across high-income and developing countries (17 percent and 64 percent, respectively). Overall, we predict a slight improvement in access to financial services in 2009: approximately 50 million more adults have accounts, although the adult population increased by approximately 79 million. The results of the econometric analysis confirm the earlier findings in the literature that higher deposit and/or loan correlation is associated with higher economic and financial development as measured by GDP per capita, the amount of electricity use, the availability of explicit deposit insurance and better credit environment, etc. Although access to deposit services has improved in 2009, the global financial crisis took its toll: volume of deposits and loans shrank. The world as a whole added 65 deposit accounts per 1,000 adults but the number of outstanding loans remained more or less the same. 2 At the same time, global retail networks, consisting of financial institution branches, ATMs, and POS terminals expanded. Per 100,000 adults, there are 167 new POS terminals, five ATMs, and one bank branch.
The analysis of data on financial inclusion mandates under the purview of financial regulators indicates that financial inclusion is high on policymakers' agendas and reforms are widespread. However, implementation capacity is often limited. Countries that have financial inclusion strategy documents also have more topics under the purview of financial regulators with more resources. Low-income countries have more topics under the purview of financial regulators, but fewer resources.
The rest of this paper is organized as follows. Section 2 describes the methodology and results of counting the number of unbanked adults across the world. The state of and changes in access to deposit and credit services, as well as in the extent of retail networks, are analyzed in Section 3.
4 Data on financial inclusion mandates and reforms are introduced and analyzed in section 4. Section 5 concludes with a discussion.

Counting the unbanked
We follow Beck et al. (2007) to predict the extent of access to formal deposit services offered by regulated financial institutions by households according to the following model: where HH share i is the percentage of adults with a bank account in country i based on various household surveys collected on or after 2006. This data are compiled from various sources: recent Living Standard Measurement Surveys (World Bank, various years) where available, as well as regional sources: for the European Union, the European Commission's Eurobarometer, Special Barometer 260 (2007); for Africa, FinMark Trust's FinScope; for Latin America, Tejerina and Westley (2007), the MECOVI database, and Barr et al. (2007); and Nenova et al. (2007). These data are referenced and expanded upon in Claessens (2006), Honohan (2008), Gasparini et al. (2005) and Beck et al. (2007). See Table 1 for further details. Note that although some of these surveys are at the household level and some are at the individual level, Honohan (2008) argues that they can be used interchangeably. Beck et al. (2007) point out that the logarithmic specification on the right-hand-side of (1) is due to outreach indicators having fat tails. Note also that the dependent variable takes values between zero and one only, and to avoid the predicted values from falling outside this range, it is possible to estimate equation (1) using a Tobit specification. Beck et al. also state that the coefficients and significance levels are similar under Tobit and OLS, which is also confirmed by our estimation results. Nevertheless, for the predicted values to lie within zero and one as well, Tobit specification is preferred. Table 2 provides the results from estimating equation (1) by both OLS and Tobit using 46 observations. As expected, a larger number of accounts and a larger number of branches per adult are associated with a greater percentage of banked households. The model is then used to predict the percentage of banked households for the rest of the world for which household survey data do not exist. The correlation between the actual and predicted values is 87 percent. See Figure 1 for the fit of the estimated model.
Given the data constraints for the explanatory variables, it is possible to construct out-of-sample predictions for 55 countries. Combined with the actual data available for 46 countries, this enables us to pin down the percentage of banked households for a total of 101 countries. Kendall et al. (2010) develop a different methodology and predict the number of accounts for each type of regulated financial institution in the Financial Access database. They then convert the number of accounts into the number of banked adults. With this methodology, they are able to predict the 5 percentage of banked adults for a total of 154 countries. We use their predictions for those countries for which data on explanatory variables in equation (1) were not available, and end up with 154 countries for which there are predictions or actual data on the percentage of formally banked households. The results indicate that, on average, half the world is unbanked. The regional breakdown of these predictions shows that Sub-Saharan Africa (SSA) and South Asia  Table 3) and for the same sample of 191 countries.
Kendall et al. (2010) use a variety of econometric models to predict the number of commercial bank deposit accounts around the world. Table 4 provides the estimation results for the two sets of predictions mentioned above, which are almost identical to the results reported by Kendall et al. using Financial Access 2009 data. We then use these models to predict the number of commercial bank deposit accounts for those countries for which data are unavailable. For the rest of the non-bank financial institution types, no estimation was made as the number of comparable observations at the country level is not sufficient. In addition, not all countries have different types of non-bank financial institutions. Instead, Financial Access data for both years are complemented by other data sources (see Table 3), and the number of deposit accounts for each institution type for each reporting country is counted. The total number of deposit accounts across all institutional types within each country is then calculated by adding up the numbers for each institutional type. This prediction exercise yields the total number of deposit accounts in each country. However, the indicator of interest is the total number of banked (or unbanked) adults (see Table 3). Hence, following Kendall et al., we convert the number of deposit accounts 6 to the number of individuals with an account by using their estimate that, on average, each individual has 3 deposit accounts. 4 As a result of this exercise we revise the estimate by Kendall et al. (2010) for the number of unbanked: as of December 2008, we predict that approximately 2.8 billion adults are unbanked, 175 million of which are in high-income countries, and 2.63 billion are in the developing world. These numbers imply that 58 percent of the world is unbanked (21 percent of adults in highincome countries, and 66 percent in developing countries). The numbers slightly improve by the end of 2009. A total of approximately 2.75 billion adults around the world are predicted to be unbanked, out of which 138 million live in high-income countries and 2.61 billion live in developing countries. Overall, 56 percent of the world is unbanked by the end of 2009, with 17 percent of adults unbanked in high-income countries and 64 percent in developing countries. Hence, about 50 million more adults in the world have bank accounts, and most of the improvement is seen in high-income countries. Note that, at the same time, the total adult population in the world increased by approximately 79 million.

Access to financial services
This section presents the state of access to financial services in 2009. We consider deposit and credit services only, as these are the basics. For both deposit services and credit services, we provide an overview of the global picture and summarize the results of econometric analyses. We also compare banks to non-bank financial institutions (NBFIs) in terms of deposit and loan penetration, although data on NBFIs are limited. Next, we consider the outreach of the financial system. An overview of the state of retail networks, including branches, ATMs, and POS terminals as well as urban vs. rural, and bank vs. non-bank divides are presented.
Note that due to limitations in data availability, especially in time dimension, it is not possible to control for potential endogeneity problems between financial service penetration and other indicators of growth and development. Hence, the econometric results reported below should be interpreted with caution, as they are likely not to point to causal effects but are rather partial associations. Furthermore, the data are from a crisis year, which may bias the results.

Deposits
In 2009 as the global financial crisis unfolded, about 60 percent of countries experienced a decline in real per capita GDP, and for those that went through expansions, median growth was only 2.1 percent. Deteriorating macroeconomic conditions affected the deposit volume around the globe as individuals and firms had to tap into their savings. Seventy-seven percent of countries in our sample experienced a decline in deposit-to-GDP ratio, with an average decline 7 of 11.8 percent. The overall world deposit-to-GDP ratio decreased from 72 percent at the end of 2008 to 66 percent at the end of 2009.
Yet even in the midst of the crisis, the use of financial services continued to expand. The number of deposit accounts per 1,000 adults increased in 69 percent of the countries reporting data. The world as a whole added 65 accounts per 1,000 adults in 2009, which is roughly a 10 percent increase in the median number of accounts per 1,000 adults. Growth has been uneven across countries, however, and the median change was only 4.3 percent. Table 5 presents the summary statistics of the main indicators used in this part of the analysis (commercial banks). The table also shows substantial variance among countries in the different indicators. Changes in volume of deposits and number of deposit accounts differ substantially across regions. Europe and Central Asia, where the financial system was severely affected by the crisis, is the region that experienced the largest declines in deposit volumes.
Europe and Central Asia is also one of the regions that performed significantly worse than the rest of the world, with a decline in the number of deposit accounts of 1 percent. Sub-Saharan Africa, the region with the lowest level of deposit account penetration, experienced the second largest median increase in the number of deposit accounts per 1,000 adults, surpassed only by Latin America and the Caribbean. Both regions have experienced increases significantly larger than the rest of the world.
The variation in changes in the number of deposit accounts and the volume of deposits exhibit similar patterns across income groups. Middle-income and high-income countries on average experienced a significantly larger decline in volume of deposits as a percentage of GDP than did the low end of the world income distribution, mostly because of the effect of the financial crisis. Though the largest median increase in deposit account penetration is observed in the poorest 20 percent of countries, the richest 20 percent have seen only a slight expansion.
A thorough analysis of the cross-country covariates of deposit account penetration implies that the level of economic development is of great consequence. Table 6 reports the results of various models estimated by OLS with the number of deposit accounts per 1,000 adults in commercial banks as the dependent variable. GDP per capita and population density are both significantly and positively associated with deposit account penetration, confirming the results by Kendall et al. (2010). Hence, in what follows, we control for GDP per capita and population density and discuss partial associations with the rest of the variables.
Indicators of overall macroeconomic conditions apart from GDP per capita are not significant. Inflation rate, which was significant with a negative coefficient in Kendall et al. is not statistically significant with 2009 data. To control for the effects of the global financial crisis, we used a variable that we named as "expected real economic growth," which is an indicator variable that takes a value of one if the actual real GDP growth in 2008 was negative and the expected real GDP growth for 2009 is positive, to capture a positive economic outlook for a 8 country. The source of the expected real GDP growth data is World Economic Outlook by the IMF (IMF, April 2010). According to our results, this variable is not statistically significant. Deposit rate, on the other hand, is significantly positively correlated with the number of deposit accounts.
Among the infrastructure indicators, electricity consumption does not have a significant relationship with deposit penetration, whereas the number of land line and cell phone users is positively correlated with the number of deposit accounts. Financial sector infrastructure indicators, such as the number of commercial bank branches per 100,000 adults and per 1,000 square kilometer that measure the geographic and demographic outreach of the financial system, are positively and significantly related to deposit penetration.
The legal environment, measured by the index of the strength of legal rights for borrowers and lenders, is also closely related to deposit penetration. A favorable legal environment for lending may enable banks to operate more profitably through lending and to grow, eventually leading to an expansion of deposit services. The analysis of the correlation between deposit account penetration and other indicators of favorable credit environment also produces similar findings. Moreover, the availability of explicit deposit insurance scheme suggests higher deposit penetration.
A greater degree of competition, as proxied by the concentration ratio in the banking sector, is associated with greater deposit penetration. This is in line with earlier studies (see, for example, Demirgüç-Kunt and Huizinga, 1999) arguing that competition in the banking sector would increase efficiency, and in turn, would lead to a larger variety of products and services offered to a larger depositor base. 5 The analysis of the correlation between the interest rate spread (lending rate minus deposit rate) and deposit account penetration lends support to this argument (not shown in Table 6). Countries with lower interest rate spreads also show a higher number of deposit accounts per 1,000 adults.
Analysis of the volume of deposits normalized by GDP offers further insight into the factors affecting financial access (Table 7). Although inflation, as a proxy of macroeconomic stability, is not significantly correlated with the number of deposit accounts, it is negatively related to deposit volume. In other words, macroeconomic stability seems to matter for the decision of how much money to put in the bank account, but not for having an account at all. This result is consistent with the evidence of a simultaneous decrease in deposit volume and increase in number of accounts despite the financial crisis. It also supports the view that a deposit account is a basic service, and having one is inelastic with respect to macroeconomic disturbances.
The analysis of the changes in the number of deposit accounts per 1,000 adults between 2008 and 2009 highlights the importance of macroeconomic stability and growth for improving financial access. The change in the number of accounts per 1,000 adults is strongly positively correlated with macroeconomic outlook, measured by short-run forecasts of real per capita GDP growth. In addition, lower deposit penetration at the end of 2008 is positively associated with a greater change in the number of deposit accounts per 1,000 adults.
A number of other macro and policy environment variables were tested and showed weak correlations after controlling for income per capita (Table 8). Overall, the results indicate that access to basic deposit services continues to improve, but macroeconomic stability and growth are essential for sustainable improvement in financial access.

Banks vs. non-banks in access to deposit services
Robust analysis of the changes in the number and volume of deposits between banks and non banks is not feasible due to data limitations. 6 Only 73 countries were able to provide data on the breakdown of deposits by the type of depositors and financial institutions. Estimates from the smaller subset of countries with comparable data suggest that commercial banks experienced a larger decline on average in deposit-to-GDP ratio than non-banks, while also enjoying a larger median increase in the number of accounts per 1,000 adults.
Banks are the main providers of deposit services around the world. The majority of the deposit volume by both individuals and non-financial businesses are in commercial banks. 91 percent of total deposit volume and 94 percent of total deposit accounts are held in commercial banks. These numbers should be interpreted with caution, as they are likely to underestimate the share of deposits held by non-banks due to limited reporting on deposit volumes and especially on the number of accounts for non-bank financial institutions (NBFIs).
Deposits by individuals account for the vast majority of deposit volume. Deposits held by individuals account for 71 percent of the total volume of deposits and 96 percent of the number of accounts. Individual deposits represent a greater share of bank deposit volume (76 percent) as well as a greater share of deposit volume in NBFIs (86 percent).

Loans
Credit services fared much worse than deposits in 2009. Indeed, the global financial crisis took its toll on access to credit services, with the value of loans as a percentage of world GDP declining from 74 percent to 65 percent. In about 85 percent of countries, loan volume as a share of GDP declined in 2009. At the same time throughout the year, the number of outstanding loans per 1,000 adults remained more or less unchanged. The number of outstanding loans per 1,000 adults decreased in 57 percent of the countries. Table 5 presents the summary statistics for these indicators aggregated at the world level and also at the regional level. The differences in median changes in commercial bank loan volume (as a percentage of GDP) show less variation across 10 countries than median changes in the number of loans. Loan volume as a percentage of GDP dropped in all regions.
Although all regions have seen contractions in lending volumes, there is significant variance amongst regions. Europe and Central Asia, where the financial system was severely affected by the crisis, experienced the largest decline in lending volumes. All the countries in that region for which data are available experienced drops in loans-to-GDP ratios ranging between 19 percent and 52 percent. All countries in Latin America and the Caribbean, but one, also experienced declines in lending volumes. For the number of outstanding loan accounts, the picture differs. Some regions have even seen improvements in loan account penetration. For instance, all countries in the Middle East and North Africa, except for Israel, underwent expansions in the number of outstanding loans. Conversely, all countries in Europe and Central Asia, except for Albania and Turkey, have had contractions in the number of outstanding loans. Table 9 presents the regression results where the dependent variable is the number of outstanding commercial bank loans per 1,000 adults. Confirming the findings by Kendall et al. (2010), we find that while GDP per capita is significantly positively associated with loan penetration, population density is insignificant. On the other hand, the number of outstanding loans is negatively correlated with inflation. Branch penetration and physical infrastructure indicators such as phone lines per capita are significantly positively related to loan penetration.
The link between loan penetration and banking sector concentration is negative, indicating that more competitive banking markets have higher levels of credit access. Better creditor rights and comprehensive credit information systems are associated with greater access to financial services. The minimum amount of consumer loans that banks make as a percentage of GSDP per capita, as a non-price barrier to accessing credit services, is significantly negatively related to number of outstanding loans.
There is also a positive link between the existence of explicit deposit insurance and loan penetration. Similar to the effect of credit rights on deposits, this indicates that to improve access to financial services, a favorable business climate that supports both deposit and loan services is essential. Sustainable improvement in access to credit is possible only if sustainable financial institutions are able to effectively manage both the asset and liability sides of their balance sheets.
Loan volume as a ratio to GDP is significantly positively related to GDP per capita and population density, and negatively related to inflation and lending rate (Table 10). Offshore financial centers have larger loan volumes as a percentage of GDP, after controlling for income per capita and population density. Branch penetration and allowing agents to provide credit related services are positively related to loan volume, in addition to a favorable credit environment measured by legal rights of borrowers and lenders, and better credit information systems.
Countries with higher expected GDP levels are more likely to experience an increase in the number of loans per 1,000 adults (Table 11). This also means that countries where macroeconomic expectations are negative are more likely to experience decreases in loan numbers as demand shrinks and banks tighten supply as credit quality deteriorates. Increase in loan penetration is larger for countries with a smaller number of outstanding loans to start with, indicating that even after controlling for expected GDP-a proxy for macroeconomic stabilityaccess to credit continues to expand. The implications of these results are twofold. First, similar to deposit services, there is evidence that access to financial services is improving, albeit slowly. Second, macroeconomic stability is fundamental for access to credit services.

Banks vs. non-banks in access to credit services
Comparison of the data provided by regulators on loans granted by banks and nonbanks is difficult for the same reasons highlighted earlier with regard to deposit services. The subset of countries with comparable data on the number of loans includes only 19 countries. The numbers are even smaller for specialized state financial institutions and microfinance institutions: 15 and 7 countries, respectively. In all cases, the average percentage change in the number of loans is not statistically significantly different from zero, similar to commercial banks.
The decline in loan volume is essentially even across different financial institutions. Nevertheless, NBFIs have experienced an increase, on average, in the number of loans per 1,000 adults, the median increase being 9 percent. It is important to note, however, that a number of countries in the sample which did not report any data on some or all types of NBFIs in 2009 did so in 2010, leading to a possible overestimation of the changes in loan volume.
Among the regulated financial institutions, banks hold 87 percent of loans-to-GDP ratio and 81 percent of number of outstanding loans per 1,000 adults. Among NBFIs, cooperatives and credit unions are the major source of credit by volume as well as by number, though the structure of the NBFI sector varies greatly from country to country. For example, in Spain and France, cooperatives account for at least half the credit volume while in South Korea, specialized state financial institutions account for nearly 67 percent of the volume.
Loans to firms represent 43.5 percent of the credit portfolio by value, but only 17.4 percent in terms of number. Banks are the main providers of credit to firms, while NBFIs are more often geared towards serving individuals in terms of volume. Nevertheless, the number of outstanding loans to individuals by banks constitutes the bulk of bank loans by number, if not by volume.

Outreach
Global retail networks, consisting of financial institution branches, ATMs, and POS terminals, expanded in 2009. The world on average added about one bank branch, five ATMs, and 167 POS terminals per 100,000 adults. However, this growth was not universal. The number of bank branches decreased in 43 percent of countries, about half of which are high-income and Eastern European countries. The number of ATMs decreased in 16 out of 104 countries, and the number of POS terminals decreased in 13 out of 77.
Growth in the retail network varies across regions and income groups. Low-income countries show the highest rates of growth in the number of bank branches, ATMs, and POS terminals, which is another sign of improved access to financial services. Africa, South Asia, and the Middle East, the regions with the lowest levels of retail network outreach, show higher rates of growth in the number of bank branches and ATMs.
The growth in low-income countries starts from a low base, especially for ATM and POS numbers, and the increase in coverage is less pronounced. For example, a 27 percent increase in the number of ATMs in Malaysia translates into an increase in coverage by more than 10 ATMs per 100,000 adults, but an equivalent percentage change in Kenya adds only 1.6 ATMs per 100,000 adults. At the extreme, Burundi doubled the number of ATMs but still has only about 0.08 ATMs per 100,000 adults-a total of four ATMs in the entire country. Less dramatic examples are Syria (with 366 ATMs) and Malawi (with 203 ATMs)-both doubled the number of ATMs, resulting in coverage of 2.6 ATMs per 100,000 adults. Overall, patterns of retail network outreach are broadly unchanged.
Tables 12 and 13 show the regression results where demographic penetration of branchesmeasured by number of branches per 100,000 adults-and geographic penetration of branchesmeasured by number of branches per square kilometer-are used as dependent variables, respectively. While GDP per capita, and number of phone users are significantly positively associated with both measures of branch penetration, population density is significantly positively associated only with geographic branch penetration. No other variable is found to have a statistically significant relationship with branch penetration. Table 14 reports the partial association between changes in branch penetration and relevant indicators. The statistically significant results indicate that countries with lower branch penetration to start with have larger increases in branch penetration. Moreover, change in the number of banks is positively related to change in the number of branches. Regional dummies for East Asia and Pacific, Latin America and the Caribbean, high income, and offshore financial center dummies all have significant coefficient estimates with negative signs, indicating countries in those groups experienced decreases in branch penetration.

Branches vs. non-branch retail locations
Despite the fast growth, ATM and POS networks remain small relative to branch networks in low and middle-income countries. In 2009 the average number of ATMs barely exceeded the number of bank branches in low-income countries for the first time. There are two ATMs per bank branch in middle-income countries and three per bank branch in high-income countries. In South Asia the ratio is the lowest -on average there is one ATM for two bank branches. It is important to note that a greater number of ATMs does not automatically mean better access. For 13 example, lack of interoperability in the system requiring each bank to build its own ATM network raises overall cost. 7 Still, the magnitude of variation between regions and income groups is indicative of the physical reach of the existing retail network.
As new technologies evolve, the trends in the use of ATMs and POS are changing. Data show that the number of ATMs relative to the number of branches slightly declined in high-income countries. At the same time the number of POS increased, reflecting a growing reliance on noncash payments. A greater use of electronic transactions through internet and cell phones may also reduce the need for ATMs going forward. For now though, as cash remains the main medium of exchange for retail transactions in most countries, cash-in-cash-out points, whether it be branches, ATMs or POS allowing cash-back, are essential elements of financial access.

Banks vs. nonbanks in retail networks
Worldwide, banks have the largest branch network representing two-thirds of all branches.
Cooperatives are the second largest with 23 percent of branches worldwide ( Figure 2). In over 60 percent of countries (52 out of 83) with the data, non-banks had half the number of branches of commercial banks. The number of cooperative branches exceeds the number of bank branches only in a small number of countries, namely Austria, Burundi, Germany, Hungary, Korea, and Spain. Figure 2 likely underestimates the size of the nonbank branch network due to data limitations for non-banks.
Commercial banks mainly target urban areas. Most bank branches are located in urban areas, representing 90 percent of all financial institutions in urban areas. On average, only 24 percent of all bank branches are in rural areas, compared to 40 percent in cooperatives and specialized state financial institutions and 50 percent in microfinance institutions ( Figure 2). Even though a smaller share of bank branches is located in rural areas, they still provide the bulk of rural coverage.
Data limitations do not allow for a robust change analysis in bank versus non-bank branch network. The challenge is threefold. First, in a number of countries information on a larger number of NBFIs is not available. Second, financial institutions changed their status by becoming banks in a number of countries. Third, only partial data are available, distorting comparison of banks and non-banks within the system. For a small subset of countries with comparable data for banks and non-banks (45 countries), there is no statistically significant difference in the growth rate of bank and non-bank branch networks.

Financial inclusion agenda
In addition to collecting fundamental indicators on financial access, the survey also explores the extent to which financial regulators are mandated to oversee financial inclusion. Historically, the main role of the financial regulator has been to ensure the stability of the financial system, focusing on regulation and supervision for the safety and soundness of financial institutions. With many countries embracing financial inclusion as a reform objective and putting in place programs to expand financial access, some regulators are playing a more promotional role as well. However, recent financial crises underscored the dilemma of choice between encouraging financial sector growth, especially credit, and maintaining the stability of the financial system.

Financial inclusion mandates and reforms
The survey asked financial regulators which of the following topics relevant for a financial inclusion agenda were under the purview of their agency: consumer protection, financial capability, regulation of microfinance, promotion of savings, promotion of access to finance for SMEs, and promotion of rural finance.
We find that 86 percent of countries in the sample have at least one element of financial inclusion under their responsibility. Seventy-one percent of countries stated they have at least two elements of financial inclusion under their responsibility. About half have at least three, and only 17 percent of the countries in our sample have all elements of financial inclusion under their mandate.
Regulators from developing countries tend to be more concerned with final inclusion as a policy objective. They have, on average, more topics under their mandate than developed countries do (3.34 versus 2.13 out of a maximum of 6). Consumer protection and financial literacy are the topics that seem to be important across all regions but developing countries also seem to place a lot of emphasis on the rest of the topics as well. A plausible explanation could be that richer countries have already achieved higher levels of inclusion and do not have the need to actively pursue policy goals in this respect. However, consumer protection and financial literacy are topics that are always relevant especially in the light of the recent financial crisis.
The survey also asked regulators whether there were reforms in the following ten areas in 2009: (1) consumer protection, (2) financial literacy, (3) basic accounts, (4) government-to-person transfers, (5) access to finance in rural areas, (6) microfinance, (7) know-your-customer (KYC) requirements, (8) access to finance by SMEs, (9) branchless banking, and (10) overindebtedness. Consumer protection is the most popular area of reform, followed by KYC requirements, access to finance by SMEs, microfinance, access to finance in rural areas, and financial literacy. The remaining four areas are not as popular. Consumer protection reforms are especially popular in high-income countries. Reforms are widespread in South Asia. East Asia and Pacific, Sub-Saharan Africa, Latin America and the Caribbean, and Middle East and North Africa also have, on average, reforms in 4-5 areas. 15

Strategy documents
Not only is financial inclusion high on regulators' agenda but we also find that many regulators draft comprehensive financial inclusion strategy documents in order to help them advance their policy objectives. Forty-five percent of regulators in the sample have produced such documents, and have done so recently. Two-thirds of countries that report the existence of financial inclusion strategy documents have drafted them in the last three years and 91 percent have done so after 2004. In addition, we find that countries that have strategy documents also have more topics under their responsibility and more resources allocated to achieving their financial inclusion objectives.

Resource gap
Low-income countries, in addition to more pressing financial inclusion needs with lower levels of access and with more financial inclusion topics under the responsibility of the regulators, need to manage with fewer resources available to overcome their challenges. One possible measure of the resource gap is to count the existing teams or units within regulators that work on topics in the financial inclusion agenda which are under the purview of their agencies. According to the survey, many regulators have more topics under their responsibility than teams and staff working on these topics. The problem is even more severe for low-income countries. Table 15 summarizes the output of selected econometric analyses on the partial association between the existence of various types of financial inclusion mandates and reforms, and deposit penetration. The results show a negative correlation, suggesting the possibility that the direction of causality may, in fact, run from the extent of financial access to the number of items in the financial inclusion agenda and reforms. We henceforth take this approach and report the results of these analyses in Tables 16 to 21. We form two indices using data on the six topics of financial inclusion under the purview of financial regulators. One index, named as FI index, assigns a value between 0 and 1 to each country based on the number of topics under the regulator's mandate, 0 for no topics and 1 for all six topics simultaneously. However, this may not be an ideal measure for the comprehensiveness of the regulators' financial inclusion agenda because it does not provide us with any information on the implementation capacity of the regulators. Therefore, we also form a second index, named FI effective index, which is also between 0 and 1, that takes into account the topic under purview together with the availability of designated staff or teams that work on that topic. In this case, a country can have a value of 0 for this index even if it has a value of 1 in the FI index, simply because there are no teams designated to work on any one of the topics. A value of 1 for the FI effective index means that the regulator has all six topics under its mandate, and each topic has a designated team. Although this second index is not an exact measure of if and how the financial inclusion agenda is implemented, it is a proxy of implementation capacity. Countries with lower income per capita levels, after controlling for deposit and loan penetration, have more topics under the mandate of the regulators. This result also holds for countries with a positive economic outlook. The existence of explicit deposit insurance is negatively related to both indices, and loan penetration is positively related to the FI effective index. Hence, poorer countries seem to have a more loaded financial inclusion agenda.

Cross-country covariates of financial inclusion mandates and reforms
The extent of reforms, measured by a similar index constructed using the ten areas of financial inclusion reforms surveyed, is also negatively related to GDP per capita, and not related to the degree of deposit or loan penetration. However, once the items in the financial inclusion agenda are controlled for using the two indices described above, the effect of GDP per capita is no longer observed. Countries that have a more loaded financial inclusion agenda seem to have a higher number of reforms in 2009 (Table 17), although it is not possible to comment on the effectiveness of the reforms.
When examining specific areas of financial inclusion topics, we find that countries with lower GDP per capita and lower deposit penetration are more likely to have promotion of basic accounts as a mandate of the regulator with a designated team (Table 18), whereas regulators in countries with lower GDP per capita and a positive economic outlook tend to have promoting access to finance in rural areas as a mandate (Table 19). Financial consumer protection and financial literacy are the two topics that are likely to receive more attention in countries where loan penetration is higher (Table 20), but countries with explicit deposit insurance tend to be less likely to have financial literacy as a mandate for the regulator with a designated team. Lastly, countries with lower GDP per capita levels are more likely to have their regulators work effectively on promoting access to finance by SMEs (Table 21).

Conclusion
Using the Financial Access database by CGAP and the World Bank Group, this paper (i) counts the number of unbanked adults around the world, (ii) analyzes the state of access to deposit and loan services as well as the extent of retail networks, and (iii) discusses the state of financial inclusion mandates around the world.
The findings indicate that there is yet much to be done in the financial inclusion arena. Fifty-six percent of adults in the world do not have access to formal financial services. The situation is even worse in the developing world with 64 percent of adults unbanked. Nevertheless, highincome countries also have to worry because approximately one in every five adults is unbanked.
On the contrary to conventional wisdom, poor people indeed need and use financial services, albeit in small amounts and usually from informal sources as it is costly for formal providers to provide services for such small amounts. Anecdotal evidence suggests that informal financial services are at least 5-10 times more costly and also less reliable than formal ones. Hence, making formal and affordable financial services available for the unbanked would definitely have positive consequences on the lives of these people.
Fortunately, the need for improving access to financial services and building inclusive financial systems are increasingly at the core of policymakers' agendas. The data indicates that access to deposit services have improved in 2009 despite the crisis. In addition, financial regulators are increasingly assuming the role of promoting financial access besides their traditional roles of regulating and supervising financial institutions for the soundness of the financial system and to ensure financial stability.

Table 8: Cross-country covariates associated with changes in deposit accounts penetration
The results of OLS regressions of the changes in the natural logarithm of number of deposit accounts per 1,000 adults in commercial banks on different country characteristics. (1) (2)

Table 11: Cross-country covariates associated with changes in loan penetration
The results of OLS regressions of the changes in the natural logarithm of number of loans per 1,000 adults in commercial banks on different country characteristics. (1) (2) (3) (4) (5) (6) (7) (8)