The Effect of Global Banking on Financial Inclusion in Emerging and Developing Economies

(1)

The Effect of Global Banking on

Financial Inclusion in Emerging and

Developing Economies

June 12, 2019 MSc Finance Thesis

(2)

1

Abstract: Increasing financial inclusion might be important to decrease inequality and alleviate

poverty in developing and emerging economies. At the same time, foreign bank presence increases in these countries. This paper will investigate the relationship between foreign bank presence and financial inclusion in 103 developing and emerging economies in the period 2004-2013. The results suggest that there is no direct relationship, but foreign banks might be able to increase financial inclusion at higher levels of GDP per capita.

Keywords: foreign banks, financial inclusion, financial development, financial globalization,

financial liberalization

(3)

2

1. Introduction

Since the positive link between financial development and economic growth was established in the 1990s, financial development has been promoted as a tool to boost growth (King and Levine, 1993; Rajan and Zingales, 1998; Levine and Zervos, 1998). At the same time, developing economies started to liberalize their financial sectors to increase competition and efficiency. This allowed foreign banks to increase their presence in these countries.

However, after the global financial crisis in 2007-2008, the weaknesses in the global financial system were exposed. The international banks were not only exporting risk, but also importing it during the crisis. The reversal of cash flows caused the financial shock to transmit across borders, while countries with capital controls were more resilient during the crisis (Ostry, Ghosh, Habermeier, Mahvash Qureshi, and Reinhardt, 2010).

Meanwhile, the research interest has shifted the focus from general financial development towards financial inclusion. Financial development literature concentrated more on increasing financial sector depth, a concept that was often measured as the ratio of private sector credit to GDP (Beck, Demirgüç-Kunt, and Martinez Peria, 2007). As more data has become available, financial inclusion, or the breadth of the financial sector has gained more interest. The aim is to especially include the people that were previously excluded from financial services, to increase growth and reduce poverty (Ahamed, Mallick, and Matousek, 2018). Financial inclusion can be measured as the access and the use of formal financial services, such as bank accounts and digital payments (Demirgüç-Kunt, Klapper, Singer, Ansar, and Hess, 2018).

This thesis will study the effect of foreign bank presence on financial inclusion in developing economies. Global banking might have an effect on financial inclusion through multiple channels. It is possible that international banks bring technologies and capital that enable access to finance. For example, international banks may invest more in mobile banking. Introducing financial technologies can be beneficial, as it can enhance the income earning potential, and allow people to manage their risk (Demirgüç-Kunt et al., 2018). On the other hand, international banks may decrease access to finance if they cherry-pick the best borrowers, which decreases the competitiveness of domestic banks (Claessens and van Horen, 2012). This could also cause the domestic banks to shift their focus towards the financially less transparent clients, since they may be better able to overcome the information asymmetries than foreign banks (Beck and Brown, 2015).

(5)

4 The contribution of this thesis is the inclusion of more recent data compared to the existing literature. Data from the IMF Financial Access Survey (FAS) has been used to study the relationship of foreign banking and financial inclusion in at least three other papers. Detragiache Tressel, and Gupta (2008) conducted a similar study, but they used a pooled OLS model with a sample including only the years 2003-2004. The sample in Beck and Martinez Peria (2010) only studied the effect of foreign banks in Mexico. The most recent study is the one by Gopalan and Rajan (2018), but their sample is limited to the years 2004-2009. This thesis includes updated data from the FAS until the year 2013.

The results of this paper suggest that there is no relationship between foreign bank presence and financial inclusion. There is some evidence that the relationship is conditional on level of GDP per capita. Above a threshold level of GDP per capita, increasing the share of foreign banks in a country had a positive effect on some of the measures of financial inclusion.

The paper will be structured as follows. Section 2 has an overview of the relevant literature: it explains what financial inclusion is, and what is the role of foreign banks in host countries. Section 3 presents the methodology of this paper. Section 4 discusses the data and presents some descriptive statistics. Section 5 shows the results of the main specification and the robustness tests. The last section concludes and presents the limitations of this study.

2. Literature review

This section will start with a subsection on financial inclusion: the definition, and how to measure it. After this there is a discussion on the role of foreign banks. This includes a review of some empirical evidence from the literature, and a formulation of the hypothesis of this study. In the end, other factors that could affect the main relationship in this study are identified.

2.1 What is financial inclusion?

(6)

5 Increasing financial inclusion can have similar effects as financial development, but the benefits are more equally distributed. Based on the neoclassical theory, we expect that poor households and small businesses with low levels of capital will have a higher return on investment. At the same time, they might not have the necessary collateral or credit history to mitigate asymmetric information problems. As a consequence, they will not receive credit for profitable investments. This can lead to poverty traps and persistent income inequality (World Bank, 2014). Constrained credit access can also lead to a less efficient resource allocation at the aggregate level, which can decrease growth. Schumpeter pointed out the importance of creative destruction, where new, more productive firms enter, and less productive firms exit. This process will be impaired if credit is not available for new firms (Beck et al., 2007).

(7)

6 Financial inclusion can also be beneficial for banks. A more varied customer base means that also the risks of the bank are better diversified, which could have a stabilizing effect on the system (Mehrotra and Yetman, 2015).

2.2 What is the role of foreign banks?

The literature has studied three main consequences of foreign bank entry: the impact on competition, stability and access to credit.1

In general, foreign banks have a competitive advantage over domestic banks as they have a better access to funding in their home economy, economies of scale, access to better technologies and management experience (Beck, 2015). These factors allow the bank to offer services with lower costs, innovate new products, and attract more deposits. Previous research has studied the effect of foreign bank entry on the market structure of the banking sector, and especially the effect on the level of competition and concentration. Most papers found that foreign banks are associated with higher competition in the banking sector, which can make it more efficient (Cull and Martinez Peria, 2012; Claessens and Laeven, 2004). Similar results are found in a sample including African countries (Kanga, Murinde, Senbet, and Soumaré, 2018; Leon, 2016). The method of entry seems to be relevant: foreign bank entry through a merger or acquisition increases the concentration of the banking sector, while greenfield investment does not (Delis, Kokas, and Ongena, 2016).

Another advantage of cross-border banking is that it allows for banks to diversify their risks across regions and enables risk sharing. In addition, they may improve the global allocation of capital. Foreign banks may also improve regulation and supervision by breaking the political ties between the regulators and the domestic banks. Alternatively, weak regulatory environments may increase excessive risk-taking of foreign banks (World Bank, 2018).

Foreign banks can also benefit the host economy: if the foreign banks have different home countries, there will be natural risk diversification. If foreign and domestic banks are substitutes, local customers can build multiple lending relationships and shift their lending between foreign and domestic banks if either of them decreases its lending due to a shock (Berger, Klapper, and Udell, 2001). Having foreign banks in the banking system can also decrease volatility for non-customers, as only domestic banks would be affected by a local credit shocks (Beck, 2015). This relates back to the financial inclusion discussion: it can be argued that both financial

(8)

7 inclusion and foreign banks have a stabilizing effect on the financial system: both can increase the diversification of risk.

Anyhow, foreign banks can also be a source of instability. Foreign banks have contagion risk: if there is a shock in the foreign bank’s home economy, the bank might decrease its lending in the host economy (Arena, Reinhart, and Vazquez, 2007). Foreign capital is also more mobile than domestic capital, which can cause sudden stops and financial flow reversals. To decrease the risks of cross-border banking regulatory cooperation is needed, since a bank failure will create negative externalities that are transmitted through the borders (Beck, 2015). This would be especially important in developing countries with weak regulatory systems.

Foreign banks also have a disadvantage caused by a larger distance and the lack of information on the local market (Beck, 2015). Since the foreign bank has less information than the domestic bank, to decrease the information asymmetry it will rely more on financial information and demand collateral from the borrowers. Because foreign banks are able to offer their services at a lower price than domestic banks, the foreign banks are able to cherry-pick the customers that are able to offer more information or collateral (Detragiache et al., 2008; Mian, 2006; Berger et al., 2001). This could potentially be harmful to the banking sector, if the customers not served by foreign banks are less profitable, which will lead to domestic banks exiting the market and leaving the informationally more opaque customers without banking services (Beck and Brown, 2015).

Empirical evidence shows that foreign banks mainly lend to large firms (Sengupta, 2007). This can impact the performance of small firms negatively if they are dependent on external financing (Gormley, 2010). This effect is not limited to firms: comparison between foreign and domestic bank customers has shown that foreign banks will cherry-pick individuals with higher education, income and wealth, and who are formally employed (Beck and Brown, 2015; Beck and Brown, 2011). A study in Mexico shows that foreign bank entry coincided with the decrease in deposit and loan accounts, while the number of bank branches increased in rich and urban municipalities (Beck and Martinez Peria, 2010). Other studies have confirmed the result and found that foreign bank presence had a negative effect on the number of loan and deposit accounts and borrowers (Detragiache et al., 2008; Beck et al., 2007; Gopalan and Rajan, 2018). However, there seems to be a positive relation with the number of ATMs (Gopalan and Rajan, 2018).

(9)

8 There are also other factors that could influence the outcome of this study. The level of economic development is connected to higher financial outreach, while it can also make a country more attractive for foreign bank entry (Beck et al., 2007). Low population density will increase transaction costs, which decreases financial outreach, and decreases foreign bank entry (Beck, 2015). It can be expected that the quality of legal institutions, and the informational environment will increase foreign bank entry and financial inclusion (Beck et al., 2007). A high level of international banks means that the institutional structure in the country is strong enough that a foreign bank would decide to enter. Good institutions are probably correlated with higher financial inclusion. However, weak institutions could also be perceived as a benefit for an entering bank, since there are more possibilities for rent seeking behavior. A banking crisis in an economy can have effects on financial inclusion and the number of foreign banks, as some banks may exit the market because of the crisis. It is also possible that banks decrease their lending to informationally opaque customers during crisis (Berger et al., 2001). Physical infrastructure can play a role: a country with good infrastructure will decrease transaction costs and is likely to increase both foreign banking and financial inclusion (Beck et al., 2007). To mitigate the omitted variable bias, these factors will be controlled for in the regression model. As a consequence, the results will show the effect of foreign banks on financial inclusion while keeping the control variables constant.

3. Methodology

Following the methodology in Gopalan and Rajan (2018), we use the Equation (1) to estimate the relationship between foreign bank presence and financial inclusion in country i and time t:2

Yit = δi +β1 Fit +β2 X it + v it (1) where Yit is financial inclusion measure in country i at time t; δi is the country fixed effect, Fit is the share of foreign banks; Xit is a matrix of control variables; and vit is the idiosyncratic error term. The parameter of interest is β1, which is the coefficient on the foreign bank share. The results of the diagnostic tests used to choose the specification of the model are in Appendix A. The proxies for financial inclusion are the number of borrowers, deposit accounts, loan accounts, ATMs and branches. The variables are scaled by adult population in the country to get an indicator of demographic access. The number of ATMs is also scaled by the area of the country to get a view on the geographic access (Beck et al., 2007). Table 1 shows a summary

(10)

9 of the results in the articles that used the same variables to measure financial inclusion as this paper. Based on these results, we can expect to find a positive effect on the number of deposit accounts and borrowers, but it is not clear what the effect will be on the other variables of interest.

Table 1 Summary of previous findings in the literature: the sign of the coefficient for foreign bank presence

Dependent variable

More foreign bank presence has a negative effect

No significant effect More foreign bank

presence has a positive effect Access to

financial services

ATMs /adults Beck et al., 2007 Gopalan & Rajan,

2018

ATMs /km2 _{Beck et al., 2007} _{Gopalan & Rajan,}

2018 Bank branches

/adults

Detragiache et al., 2008 Beck et al., 2007 Gopalan & Rajan, 2018

Usage of financial services

Borrowers Gopalan & Rajan, 2018 Deposit

accounts

Beck et al., 2007; Detragiache et al 2008; Gopalan & Rajan, 2018

Loan accounts Beck et al., 2007 Detragiache et al., 2008; Gopalan & Rajan, 2018

The measurement of financial inclusion is concentrated in three dimensions: outreach, usage and quality of financial services (Mialou et al., 2017).

(11)

10 (Beck et al., 2007). In addition, we are not able to separate the access points of foreign banks and the access points of domestic banks, which could point out they are distributed differently. Usage of financial services is measured as how many individuals are using financial services, for example the number of deposit or loan accounts per 1,000 adults. The limitation of this measure is that one individual may have more than one account (Beck et al., 2007). It is also possible that accounts are created but not actively used (World Bank, 2014).

The quality of financial services is measured as the extent to which the financial services providers are able to offer products that fit the needs of the customers. Proxies for this are the cost of usage, financial literacy or disclosure requirements, but this data is not readily available (Mialou et al., 2017). This is why this measure is left out from this study.

The share of foreign banks is measured in two ways: as the number of foreign banks over all banks, and the share of foreign bank assets of all bank assets. The share of foreign banks relative to all banks is intuitively the preferred measure, because it takes into account the size effect of the bank.

The control variables are selected based on the existing literature and try to capture the effect of macroeconomic and institutional factors that could change the relationship between foreign banking and financial inclusion.

First, we control for GDP per capita, which measures the general level of development in a country (Detragiache et al., 2008). It is expected that a country with a higher GDP per capita also has a higher level of financial inclusion, therefore its coefficient should have a positive sign.

Second, there are differences between the institutions in different countries. Therefore, we control for the credit information and the cost of contract enforcement (Beck et al., 2007). In addition, the extent of legal rights is controlled for, although we can expect that in developing countries contract enforcement is more important than legal rights, if law enforcement is not effective (Detragiache et al., 2008). It is expected that the sign of credit information and legal rights will be positive, as better institutions are likely connected to more financial inclusion. Since the cost of contract enforcement will increase transaction costs, we expect that it has a negative coefficient.

(12)

11 Fourth, the physical infrastructure is likely to have a positive effect on financial inclusion. Better infrastructure makes it easier and cheaper to increase banking sector outreach (Beck et al., 2007). We use rail lines divided by land area as a proxy for infrastructure (Beck et al., 2007). The expected sign of the coefficient is positive.

Lastly, higher population density decreases the transaction costs. Therefore, it is expected that the effect on financial inclusion is positive.

Since foreign bank presence can be correlated with financial development, there is a risk that the results suffer from endogeneity bias. It is hard to say if foreign banks increase financial inclusion or if countries with more financial development also have more financial inclusion, and also attract more foreign banks (Claessens and van Horen, 2012). Countries with lower financial development might not have good business prospects which makes them less desirable to enter. This would cause the coefficient of foreign banks to be biased upwards. On the other hand, foreign banks may find it more attractive to enter countries with low financial development, since these markets may grow faster. This would cause a downward bias in the coefficient of interest (Detragiache et al., 2008).

A preferred solution for endogeneity would be an instrumental variable model. A good instrument would be a variable that is correlated with foreign bank presence but not with financial inclusion. Detragiache et al. (2008) use three instruments: the share of large international banks located in a former colonizing country of the host country, population as a proxy for market size, and language as a proxy for cultural distance. However, it is unclear if these instruments are valid and not indirectly correlated with financial inclusion. The data on the share of banks located in a former colonizing country is limited to the year 2002. Population does not necessarily capture the size of the market, because of heterogeneity in the characteristics of countries with the same population. As discussed in the literature review, some financial exclusion is voluntary, which decreases the demand for financial services. This could decrease the potential market size as perceived by the foreign bank and decrease financial inclusion at the same time. It is not clear how only measuring population would capture the attractiveness of a market to a foreign bank. Cultural distance, measured as the number of European languages as an official language in the host country, also does not seem appropriate. We already know that a larger distance will decrease financial inclusion because it increases information asymmetries (Mian, 2006). Therefore, cultural distance is likely connected to the variables of financial inclusion. Because of these reasons, we do not perform instrumental variable regressions in this paper.

(13)

12 Instead, the endogeneity problem is addressed by using a fixed effects model, which controls for time-invariant entity-fixed effects. In a fixed effects model, it is studied if a higher foreign bank presence has an effect on financial inclusion within countries. A fixed effects model assumes that the entity-fixed effect has the same slope and variance over time. This means that country-specific attributes that change over time may bias the results. In addition, a fixed effects estimator assumes that there is strict exogeneity, which means that the error term is not correlated with the covariates. Anyhow, the fixed effect model allows the regressors to be correlated with the entity-fixed effect.

To test the robustness of the fixed effects model, this paper presents the results of a specification using first differences. If the fixed effects and the first differences model give similar results, we can assume that there are only time-invariant factors and the strict exogeneity assumption holds. If the results are different, we can conclude that there may be endogeneity bias, which should be solved using instrumental variables. The first differences estimation is preferred if the dataset is a balanced panel, because an unbalanced panel will cause too many observations to be dropped.

4. Data

This paper measures the relationship between foreign bank presence and financial inclusion in 103 developing and emerging economies, for the years 2004-2013.3 The main variables of interest are foreign bank presence and financial inclusion. The final sample is smaller, depending on the availability of data for each specification.4 A summary of the definitions and sources of the variables is in Appendix B.

To measure foreign bank presence, this paper uses the bank ownership database of Claessens and van Horen (2015). The complete database covers the years 1995-2013, and 138 countries of which 107 are classified as developing or emerging economies. In total, there is ownership information on 5,498 banks that reported to Bankscope during the time period. A limitation of the dataset is that the Bankscope data does not capture all banks, and this is especially a problem for the developing countries. The dataset classifies a bank as foreign if at least 50% of the banks’ shares are foreign owned.

The literature measures foreign bank presence either as the share of foreign banks of all banks or using a measure that takes into consideration the size of the foreign banks: the share of

3_{See Appendix D for the list of countries. Final sample may vary based on the availability of data on the control} variables.

(14)

13 foreign bank assets to all bank assets. Claessens and van Horen (2012) measure foreign bank presence as the ratio of foreign bank assets to all bank assets in an economy. Foreign bank presence is measured as the number of foreign banks over all banks in Kanga et al. (2018) and Delis et al. (2016). This paper will use both measures for robustness. Table 2 shows that the summary statistics between the two measures are quite similar: the share of foreign banks of bank assets has a lower mean, but the standard deviation is higher in the overall sample, but also within countries. However, the differences are not large, and it is expected that these measures will lead to similar results.

Table 2: Comparison between the summary statistics of the two measures of foreign bank presence: share of assets and share of banks. N refers to number of observations, n is the number of countries and T is the average number of years per country. Sample is developing and emerging economies in years 2004-2013. Full summary table with all variables is in Appendix C.

Variable Mean Std. Dev. Min Max Observations

Foreign banks % assets overall 43.70692 33.13112 0 100 N = 911 between 32.43632 0 100 n = 102 within 7.826879 -13.6264 75.26247 T = 8.93137 % banks overall 44.86214 27.20559 0 100 N = 1030 between 26.54674 0 100 n = 103 within 6.448164 6.862136 62.86214 T = 10

Financial inclusion will be measured using variables from the IMF Financial Access Survey data. The dataset is constructed using data supplied by the countries’ officials. The variables belong in two groups: access to and use of financial services. The complete dataset covers 65 indicators and 180 countries, and it has annual data from the years 2004-2016. Since bank ownership information is only available until 2013, data after this year is excluded from the final sample.

(15)

14 All financial inclusion variables are scaled either using the size of the population or the land area in a country.5 The variables can be divided in two groups: outreach and usage variables. The outreach variables refer to the physical access points that are available: the ATMs and bank branches. Because of the rise of branchless banking, measuring bank branches may not capture the newest trends (Čihák, Demirgüç-Kunt, Feyen, and Levine, 2012). In branchless banking an official bank account is not necessary, as the money transactions use technologies such as internet, ATMs, debit cards and mobile phones. Since only limited data is available on mobile banking, the number of ATMs works as a proxy for branchless banking. The use of financial services is measured as the number of borrowers, number of deposit accounts and number of loan accounts. These variables are scaled using the size of the adult population.

The Financial Access Survey measures the use of financial services in two ways: as the customers that are in the nonfinancial corporations, and only including the household sector. As can be seen in the summary statistics in Table 3, the household measures have only about half the observations compared to the measures that include the whole nonfinancial sector. Including firms does not seem to change the values a lot as can be seen from the means and ranges of these variables. In addition, the correlation table shows that the variables measuring the nonfinancial sector are highly correlated with the corresponding variables that only include households (Table 4). Therefore, we will use the variable that includes nonfinancial firms will be beneficial as it has fewer missing data, while it is closely related to the measure that only includes households. The regression results show some differences between these two measures, but these differences are likely caused by the changes in the size of the sample. The Findex variables measure the percentage of adult population that reported having a bank account or taking a loan in the past year. The correlation table (Table 4) shows that both are positively correlated with the financial inclusion variables from the FAS, although the variable Account has higher and more significant correlation coefficients. Since the FAS variables have observations for more years, they are preferred in the regression, and the Findex variables are only used to test robustness (see Table 3).

(16)

15 Table 3

Summary statistics for financial inclusion variables, only countries in the sample and years 2004-2013. The full summary statistics for all variables are in Appendix C.

Variable Obs. Mean S.D. Min Max

ATMs /adults 892 29.98395 34.73844 0 288.632 ATMs /km2

907 32.93685 113.5352 0 1274.66 Bank branches /adults 993 14.16432 14.36571 0.391847 142.192 Borrowers /adults 536 151.3242 158.4932 0.018254 872.8073 Borrowers only HH 276 174.3086 151.8093 2.000781 786.3641 Deposit acc /adults 652 938.4139 876.0125 1.300208 5144.291 Deposit acc only HH 257 1060.1 928.2417 0.892341 4634.326 Loan acc/adults 486 288.5145 275.3828 0.406705 2280.154 Loan acc only HH 288 308.9881 285.482 0.21693 1915.196 Account (% adults) 99 38.13131 24.61441 1.5217 96.8243 Loan (% adults) 99 9.24562 5.786871 0.92778 30.6537

Data for GDP per capita is from the WDI. GDP per capita is log-transformed to decrease outliers. The banking crisis dummy is from the World Bank Global Financial Development Database (GFDD).

The data for legal rights index, credit information index, and cost of enforcing contracts are from the World Bank Doing Business Database. The legal rights index measures how well the rights of lenders and borrowers are protected. The credit information index measures how accessible credit information is, and of what quality and scope. The cost of enforcing contracts measures the official costs of court procedures, such as the use of attorneys and administrative debt recovery procedure as a percentage of the debt value. The Doing Business database uses different methodology for different years, which means that they were aggregated to one variable to construct a complete time series. This should not be a problem in the regressions, since these variables are used as controls.

To take into account differences in physical infrastructure and demographics, we use rail density and population density, both from the World Bank World Development Indicators (WDI).

The variables used in the robustness test, urban population and bank overhead cost, are from the WDI and the GFDD, respectively.

(17)

16 Table 4

Pairwise correlations, including only the final sample: developing and emerging economies for the years 2004-2013. * p<0.05

Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13)

(1) Year 1.000

(2) Foreign banks % assets -0.013 1.000

(3) Foreign banks % banks 0.105* 0.867* 1.000

(4) ATMs /adults 0.182* 0.015 -0.022 1.000

(5) Bank branches /adults 0.103* 0.137* 0.017 0.372* 1.000

(6) ATMs /km2 _0.040 _-0.051 _-0.059 _0.737* _0.103* _1.000

(7) Borrowers /adults 0.243* 0.067 0.087* 0.709* 0.387* 0.458* 1.000

(8) Borrowers only HH 0.336* 0.062 0.166* 0.642* 0.256* 0.535* 0.992* 1.000

(9) Deposit acc /adults 0.114* 0.020 0.026 0.776* 0.341* 0.583* 0.756* 0.780* 1.000

(10) Deposit acc only HH 0.257* -0.056 -0.006 0.821* 0.430* 0.462* 0.800* 0.814* 0.997* 1.000

(11) Loan acc/adults 0.175* 0.123* 0.187* 0.758* 0.364* 0.306* 0.836* 0.801* 0.723* 0.840* 1.000

(12) Loan acc only HH 0.237* 0.140* 0.225* 0.731* 0.336* 0.419* 0.836* 0.831* 0.705* 0.840* 0.993* 1.000

(18)

17 Continuation of Table 4 Variables (14) (15) (16) (17) (18) (19) (20) (21) (22) (23) (24) (25) (26) (14) Loan (% adults) 1.000 (15) Banking crisis 0.046 1.000 (16) GDP per capita 0.176 0.081* 1.000 (17) Population -0.058 -0.007 -0.095* 1.000 (18) Private credit to GDP 0.303* 0.098* 0.419* 0.234* 1.000 (19) offshore dummy 0.186 -0.047 0.325* -0.072* 0.255* 1.000

(20) Legal Rights Index 0.232* 0.130* 0.168* -0.066 0.280* 0.011 1.000

(21) Credit info. index 0.364* 0.071 0.432* -0.040 0.417* -0.027 0.129* 1.000

(22) Cost of contract enf. -0.243* -0.060 -0.358* 0.006 -0.357* -0.090* -0.082* -0.278* 1.000

(23) Rail density 0.028 0.066 0.577* -0.110* 0.157* . 0.450* 0.281* -0.240* 1.000

(24) Population density 0.335* -0.072* 0.194* 0.109* 0.206* 0.466* -0.004 -0.027 -0.082* 0.135* 1.000

(25) Urban population 0.153 0.131* 0.617* -0.095* 0.295* -0.018 -0.062 0.535* -0.379* 0.234* -0.084* 1.000

(19)

18

5. Results

The preferred estimation method for this paper is the fixed effects regression, because it controls for time-invariant country-fixed effects. The results are in Table 5, where foreign bank presence is measured as the ratio of foreign-owned bank assets to all bank assets. Holding all other variables constant, the coefficient of the variable of interest is not significant. This is the case across all the specifications, which suggests that there is no connection between foreign bank presence and financial inclusion.

To make sure that the results are not caused by a misspecification of the estimation, additional estimations were made, using OLS, random effects, year-fixed effects and first differences. The first differences estimation is used as a robustness test for the fixed effects model: if the results differ significantly, the estimation has probably a strong endogeneity problem. In addition, more measures for the variables were used.6

Table 6 and Table 7 show the summary of the results, with both measures for foreign presence, and different proxies for financial inclusion.7 In Table 6 foreign banks are measured using their asset share. The results show that only the OLS and RE estimation had some significant coefficients with a positive sign. The interpretation would be that the larger the share of foreign assets of all assets, the more there is financial inclusion. However, the econometric diagnostic tests show that we are breaking the assumptions of pooled OLS and RE estimation, because there is significant variance across entities and the country-fixed effect is correlated with the regressors.8 Therefore, the results of the pooled OLS and RE estimation are not valid. When controlling for country- and year-fixed effects, or using first differences, the main explanatory variable is no longer significant.

In Table 7, foreign bank presence is measured as the share of banks, regardless of their size. This changes the results: now also the fixed effects regression is significant if measuring financial inclusion as the number of ATMs, borrowers, deposit accounts or loan accounts. The other models do not have significant results.

The reason for this could be that bank size matters. When a country has more foreign banks, there will be more financial inclusion, but when the foreign banks are larger and fewer, the effect is not significant. Another explanation could be that foreign bank presence as the share of banks has more within-country variance than foreign bank presence as the share of assets. The reason could be that there have been mergers and acquisitions among the domestic banks,

6_{The model using Findex variables Account and Loan were also estimated with lagged explanatory variables to} make a panel of 2 years, but the results were not significant. Output can be requested from the author.

(20)

19 which will change the total number of banks, but not necessarily the size of the banking sector measured in assets. In an additional robustness test, banking sector concentration was used as a control variable, but this did not change the results.9

Table 5: Regression results of fixed effects panel specification. The sample includes developing and emerging economies in years 2004-2013. First row shows the dependent variable. Foreign bank share is the ratio of bank assets of foreign-owned banks to total bank assets in the country. GDP per capita is GDP measured in constant US$, divided by population. Banking crisis dummy is 1 in years when the country has a systemic banking crisis. Legal rights index measures the extent that collateral and bankruptcy laws protect the rights of borrowers and lenders. Credit information index measures how accessible credit data is for creditors. Cost of enforcing contracts measures the cost of attorney, court and enforcement. Rail density measures the development of physical infrastructure. Population density is the population divided by the land area. Robust standard errors in parentheses. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4) (5) (6)

ATMs/adult s

ATMs/km2 _Branches _Borrowers _Deposit

(21)

20 Table 6. Summary of results 1: Foreign bank share measured as % of foreign assets to all assets. Dependent variable in the first column; first row shows the estimation method. Stars refer to the significance of the main explanatory variable, and + or - refers to the sign. Robust standard errors are used in all specifications. Sample only includes emerging and developing economies in years 2004-2013. All regressions include the following control variables: ln (GDP/capita), banking crisis dummy, legal rights index, credit information index, cost of enforcing contracts, rail density, and population density. *** p<0.01, ** p<0.05, * p<0.1

OLS RE FE FE and year

(22)

21 Table 7. Summary of results 2: Foreign bank share measured as % of foreign banks to all banks. Dependent variable in the first column; first row shows the estimation method. Stars refer to the significance of the main explanatory variable, and + or - refers to the sign. Robust standard errors are used in all specifications. Sample only includes emerging and developing economies in years 2004-2013. All regressions include the following control variables: ln (GDP/capita), banking crisis dummy, legal rights index, credit information index, cost of enforcing contracts, rail density, and population density. *** p<0.01, ** p<0.05, * p<0.1

OLS RE FE FE, year FE First Diff.

1. ATMs/adults +*** +*** +* + +** 2. ATMs / km2 _+** ₊ _- _- _- 3. Branches - +* + + +* 4. Borrowers + +** +* + + 5. Deposit accounts + +*** +** + + 6. Loan accounts +** +*** +* + +* 7. Borrowers (only households) +** +** + + + 8. Deposit accounts (only households) - +* - - - 9. Loan accounts (only households) +*** +** + - + 10. Account (% adults) (Findex) - only OLS 11. Loan (% adults) (Findex) -* only OLS

5.1 Robustness tests

In addition to the main specification, some robustness tests were performed. First, we report the results of a model including an interaction term, and then we show a replication of the model of Gopalan and Rajan (2018).

Interaction term with GDP per capita

We add an interaction term, GDP per capita multiplied by foreign bank share, to the model presented in the previous section.10 A specification with an interaction term between GDP per capita and foreign bank share measured in bank assets was never significant. Interaction using

(23)

22 the foreign bank presence as a share of banks, was significant in some specifications. Table 8 presents the output of the regressions where the interaction term was significant. The results show a clear pattern: the coefficient of foreign bank share is significant and negative, while the coefficient of the interaction is significant and positive. This points to a non-linear effect: the slope of one variable changes when the value of the other variable changes. Simply put, the effect of foreign banks on the number of ATMs becomes larger as the GDP per capita increases. The interpretation of the interaction is easier using a graph of the effect (see Figure 1). The figure shows the estimated effect of the share of foreign banks on the number of ATMs at each level of GDP per capita. Every point where the confidence interval is above zero, the increasing the share of foreign banks will result in a significant increase in the number of ATMs. The cutoff point is when GDP per capita is above 3,600.11 A similar interpretation can be given for the other interaction terms: after a cutoff point, the having more foreign banks will increase financial inclusion.

Figure 1: This figure shows the average marginal effect of foreign banks on the number of ATMs per 100,000 adults, conditional on the GDP per capita. GDP per capita is transformed with a natural logarithm. The rest of the model is the same as in Table 8, column 1.

(24)

23 Table 8

Interaction between foreign bank presence and GDP per capita. First row shows the dependent variable. Foreign bank share is the ratio of bank assets of foreign-owned banks to total bank assets in the country. All models use country-fixed effects. Robust standard errors in parentheses. The sample includes developing and emerging economies in years 2004-2013. *** p<0.01, ** p<0.05, * p<0.1

(1) (2) (3) (4) (5) (6) (7)

ATMs/adults Borrowers Borrowers (only households) Borrowers (only households) Deposit accounts (only households) Deposit accounts (only households) Loan accounts (only households) Foreign banks (% banks) -1.979* -20.60** -36.25** -32.76** -46.69* -77.28* -50.36*** (1.075) (9.701) (15.33) (14.89) (25.94) (37.07) (14.12) Foreign banks * ln (GDP/capita) 0.284** 2.943** 4.847** 4.313** 5.691** 8.486** 6.511*** (0.131) (1.208) (1.965) (1.928) (2.690) (4.019) (1.840) ln (GDP/capita) 28.59* 10.45 1.228 -58.05 1,755*** 695.9 -403.3 (14.67) (48.80) (47.65) (47.24) (554.2) (659.5) (246.8) Banking crisis dummy 18.89*** 4.494 -147.7** -149.4** (4.802) (19.90) (56.88) (59.05)

Legal Rights Index 1.552 -1.408 -5.597 -8.351 -137.1 -60.35 0.331

(25)

24

Replication of Gopalan and Rajan (2018)

Since the main regression results in this paper were not significant, we turn to a published paper that used the same variables, to check whether we are able to replicate their results. Gopalan and Rajan (2018) use a smaller panel: only 50 emerging and developing economies, for the years 2004-2009. They find that ATMs and branches had a significant and positive relationship with the share of foreign bank assets, while borrowers and deposit accounts had a significant negative relationship. The variable loan accounts was not significant in their specification. We estimate the same regression as they do, adjusting the sample to include the same countries and years. The results are in Table 9.

The number of ATMs, both scaled by land area and adult population is significant and positive. column (1) shows that one percentage point increase in the share of foreign bank assets in a country is related to an increase of 0.35 ATMs per 100,000 adults, keeping all other variables constant. Similarly, we see in column (2) that one percentage point increase in foreign banks assets is correlated with 0.4 more ATMs per 1,000km2. This is in line with the results in Gopalan and Rajan (2018).

Anyhow, the coefficient of share of foreign banks is not significant in the specifications using the other dependent variables, which was not expected, since Gopalan and Rajan found a significant effect using the same sample. One explanation for the different results is the number of countries in the specification. In Table 9, this paper has 20-39 countries, whereas Gopalan and Rajan have a sample with 25-45 countries in the same regressions. It is not clear, why this is the case, since all the variables have been downloaded from the same data sources. Gopalan and Rajan also do not report trimming or winsorizing the data. Possibly some of the data was updated which has caused the missing data. This suggests that the six countries that are missing in this dataset, might be driving the result in Gopalan and Rajan.

Since the Gopalan and Rajan sample ends in 2009, their model was also estimated with a sample including years until 2013.12 In this case, all coefficients for foreign banks became insignificant. This seems to imply that their results were driven by the specific time period and sample of countries, which means that we should be cautious when interpreting their results.

(26)

25 Table 9: Gopalan & Rajan (2018) Tables 2 and 3 reproduced. Dependent variables are in the first row: variables measuring access: number of ATMs and branches, and variables measuring usage: number of borrowers, deposit accounts, and loan accounts. Foreign bank presence is measured as the share of foreign owned bank assets to all bank assets. Sample includes 50 emerging and developing countries listed in Gopalan and Rajan, and the years 2004-2009. All regressions use country-fixed effects and all errors are clustered by country. *** p<0.01, ** p<0.05, * p<0.1

Access indicators Usage indicators

(1) (2) (3) (4) (5) (6) ATMs/ adults ATMs/ km2 Bank branches Borrowers Deposit accounts Loan accounts Foreign banks (% assets) 0.352** 0.405*** 0.0120 0.297 1.699 1.479 (0.157) (0.137) (0.0361) (0.720) (2.416) (0.974) GDP per capita 0.000566 -0.000692 0.000319* 0.0130 0.0790* 0.103*** (0.000940) (0.000956) (0.000188) (0.00821) (0.0430) (0.0119) Urban population 3.526* 1.994 -0.217 1.095 11.89 3.582 (1.812) (1.848) (0.184) (5.356) (18.90) (4.968) Credit information index 2.267** 2.419*** 0.366 18.95 15.24 -2.561 (1.059) (0.731) (0.249) (15.23) (16.49) (3.470) Legal Rights Index 2.208 0.720 0.582 3.139 49.72** 4.475

(1.310) (0.888) (0.409) (1.835) (22.20) (3.734) Bank overhead costs -0.116 0.262 0.109 -0.210 -2.065 2.257

(27)

26

6. Conclusion

This paper studied the effect of foreign bank presence on financial inclusion in emerging and developing economies. Higher financial inclusion can have various positive effects on the society. Most importantly, it has the potential to alleviate poverty and decrease income inequality. Foreign bank presence has been increasing in the developing and emerging economies, and it is important to study what effects they might have on the economy.

The data was analyzed using a panel regression with data from 103 emerging and developing economies for the years 2004-2013. Financial inclusion was captured using outreach and usage indicators, and foreign banking was measured as the foreign share of all bank assets and the foreign share of all banks. The findings seem to suggest that there is no significant linear relationship. This was the case across nearly all specifications. However, there is some evidence that for some of the indicators the relationship might be conditional on the level of economic development: the relationship between foreign banks and financial inclusion was significant and positive at higher levels of GDP per capita. The significant results were obtained when measuring foreign bank presence as a share of the total number of banks, but they became insignificant when measuring it as the share of banking sector assets. The differences point to the direction that the size of the bank matters.

The results point towards some tentative policy implications. In general, promoting the entrance of foreign banks might not have an effect on the levels of usage and access of financial services. No effect was found between the share of foreign banks and financial inclusion at lower levels of GDP per capita. Therefore, foreign bank entry should not be promoted with the aim of increasing financial inclusion. Instead, these policies could be useful for countries with a higher level of economic development.

6.1 Limitations

Endogeneity could be biasing the results of the regressions. Especially the significant results in the OLS and RE regressions could be caused by endogeneity. The summaries of the regressions show that the specifications using pooled OLS or random effects panel regression had significant coefficients, but they became insignificant when controlling for the country or time-fixed effect. A possible explanation is that the share of foreign banks does not vary enough over time. In this case, the country-fixed effect captures the whole effect. This can especially be a problem if each country does not have enough observations. A solution could be to increase the sample size, which in effect increases the variation within a country.

(28)

27 such as profitable firms. On the other hand, a low level of financial inclusion could also attract foreign banks if the country experiences high demand but low supple of financial services. It is also necessary to point out the limitations in the availability of the data. This study does not measure financial inclusion in the strict sense: banks serving more people who were previously unbanked. In addition, the financial inclusion of different income groups was not studied. The Global Findex database has variables that could be used to measure this, but the problem is the availability of bank ownership data. This could be a possible research path in the future, when the bank ownership database is updated to include more years.

In addition, the focus of this thesis was on bank lending and deposit taking, and not on other financial services such as insurance and payments. For example, mobile payments could be an interesting research path for the future. The variable “bank branches” only includes the data from commercial banks, because the data on microfinance institutions, credit unions and financial cooperatives is largely missing. International banks do not necessarily set up a branch in the foreign country but might have other cross-border activities (World Bank, 2018). Measurement error might also bias the results. The bank ownership database only includes banks that are in Bankscope, which does not include all banks: especially the coverage in poor countries is less than perfect (Kanga et al., 2018). Similarly, the other variables may have less observations or less accurate values in less developed countries.

The variables of this study are outcomes, which means that we cannot be sure, how forces of supply and demand have impacted the equilibrium (Beck et al., 2007).

(29)

28

References

Ahamed, M., Ho, S., Mallick, S., Matousek, R. 2018. Inclusive Banking, Financial Regulation and Bank Performance: Cross-Country Evidence. DECRG Kuala Lumpur Seminar Series working paper. https://drive.google.com/file/d/134PAkB8209qnRfPydIOkVBxUe-9edtXM/view.

Arena, M., Reinhart, C., Vazquez F. 2007. The Lending Channel in Emerging Economics: Are Foreign Banks Different? IMF Working Paper 07/48, International Monetary Fund, Washington, DC. https://www.imf.org/external/pubs/ft/wp/2007/wp0748.pdf.

Beck, T. 2015. Cross-Border Banking and Financial Deepening: The African Experience. Journal of African Economies 24, AERC Supplement 1, i32–i45. doi: 10.1093/jae/eju028. Beck, T., Brown M. 2015. Foreign Bank Ownership and Household Credit. Journal of Financial

Intermediation 24, 466–486. http://dx.doi.org/10.1016/j.jfi.2013.10.002.

Beck, T., Brown, M. 2011. Which Households Use Banks? Evidence from the Transition Economies. ECB Working Paper No. 1295. https://ssrn.com/abstract=1761435.

Beck, T., Demirgüç-Kunt, A., Martinez Peria, M.S. 2007. Reaching out: Access to and use of banking services across countries. Journal of Financial Economics 85, 234–266. doi:10.1016/j.jfineco.2006.07.002.

Beck, T., Martinez Peria, M.S. 2010. Foreign bank participation and outreach: Evidence from Mexico. Journal of Financial Intermediation 19, 52–73. doi:10.1016/j.jfi.2009.03.002. Berger, A. N., Klapper, L.F., Udell G.F. 2001. The Ability of Banks to Lend to Informationally

Opaque Small Businesses. Journal of Banking and Finance 25, 2127–67. https://doi.org/10.1016/S0378-4266(01)00189-3.

Čihák, M., Demirgüç-Kunt, A., Feyen, E., Levine, R. 2012. Benchmarking Financial Systems around the World. Policy Research Working Paper 6175. World Bank, Washington, DC. https://openknowledge.worldbank.org/handle/10986/12031

Claessens, S., Laeven, L. 2004. What Drives Bank Competition? Some International Evidence.

Journal of Money, Credit, and Banking, 36, 563-583.

https://www.jstor.org/stable/3838954.

Claessens, S., van Horen, N. 2012. Foreign Banks: Trends, Impact and Financial Stability. IMF Working Paper No. 12/10. https://www.imf.org/en/Publications/WP/Issues/2016/12/31/ Foreign-Banks-Trends-Impact-and-Financial-Stability-25618.

Claessens, S., van Horen, N. 2014. Foreign Banks: Trends and Impact. Journal of Money, Credit and Banking 46, 295–326.

Claessens, S., van Horen, N. 2015. The impact of the global financial crisis on banking globalization. IMF Economic Review 63, 868–918. doi:10.1057/imfer.2015.38.

Cull, R., Martínez Pería M.S. 2012. Foreign Bank Participation in Developing Countries. In: Caprio, G. (Ed.), The Evidence and Impact of Financial Globalization. Elsevier Science

& Technology. ProQuest Ebook Central, 213-222.

(30)

29 Delis, M.D., Kokas, S., Ongena, S. 2016. Foreign Ownership and Market Power in Banking:

Evidence from a World Sample. Journal of Money, Credit and Banking 48, 449-483. Demirgüç-Kunt, A., Klapper, L., Singer, D., Ansar, S., Hess, J. 2018. The global findex

database 2017: Measuring financial inclusion and the fintech revolution. Washington, DC: World Bank.

Detragiache, E., Tressel, T., Gupta, P. 2008. Foreign Banks in Poor Countries: Theory and Evidence. The Journal of Finance 63, 2123-2160. https://doi.org/10.1111/j.1540-6261.2008.01392.x.

Gopalan, S., Rajan, R. 2018. Foreign banks and financial inclusion in emerging and developing economies: An empirical investigation. Journal of International Development, 30, 559-583. doi:10.1002/jid.3354.

Gormley, T. A. 2010. The Impact of Foreign Bank Entry in Emerging Markets: Evidence From India. Journal of Financial Intermediation, 19, 26-51. http://dx.doi.org/10.1016/j.jfi.2009.01.003.

International Monetary Fund (IMF). Financial Access Survey (FAS). http://data.imf.org/FAS. Kanga, D., Murinde, V., Senbet, L., Soumaré, I. 2018. Pan-African Banks on the Rise: Does

Cross-Border Banking Increase Firms' Access to Finance in WAEMU? Centre for Global Finance Working Paper Series no. 13/2018.

King, R., Levine, R. 1993. Finance and growth: Schumpeter might be right. Quarterly Journal of Economics, 108, 717-737.

Leon, F. 2016. Does the expansion of regional cross-border banks affect competition in Africa? Indirect evidence. Research in International Business and Finance 37, 66–77. http://dx.doi.org/10.1016/j.ribaf.2015.10.015.

Levine, R. 1996. Foreign Banks, Financial Development and Economic Growth. In: Barfield, C.E. (Ed.), International Financial Markets: Harmonization versus Competition. AEI Press, Washington, DC, pp. 224-255.

Levine, R., Zervos, S. 1998. Stock Markets, Banks, and Economic Growth. The American Economic Review, 88 (3): 537-558. https://www.jstor.org/stable/116848.

Mehrotra, A., Yetman, J. 2015. Financial inclusion – issues for central banks. BIS Quarterly Review.

Mialou, A., Amidzic, G., Massara, A. 2017. Assessing Countries’ Financial Inclusion Standing – A New Composite Index. Journal of Banking and Financial Economics 2, 105–126. doi: 10.7172/2353-6845.jbfe.2017.2.5.

Mian, A. 2006. Distance Constraints: The Limits of Foreign Lending in Poor Economies. The Journal of Finance 61, 1465-1505. https://www.jstor.org/stable/3699329.

Nickell, S. 1981. Biases in dynamic models with fixed effects. Econometrica 49, 1417-1426. https://www.jstor.org/stable/1911408.

(31)

30 Rajan, R.G., Zingales, L. 1998. Financial Dependence and Growth. The American Economic

Review, 88, 559-586. http://www.jstor.org/stable/116849.

Sengupta, R. 2007. Foreign entry and bank competition. Journal of Financial Economics 84, 502–528. doi:10.1016/j.jfineco.2006.04.002.

World Bank. 2014. Global Financial Development Report 2014: Financial Inclusion. Washington, DC: World Bank. doi:10.1596/978-0-8213-9985-9.

World Bank. 2018. Global Financial Development Report 2017/2018: Bankers without Borders. Washington, DC: World Bank. doi:10.1596/978-1-4648-1148-7.

World Bank. Doing Business Database. http://www.doingbusiness.org. World Bank. Global Financial Development Database (GFDD).

World Bank. Global Financial Inclusion (Global Findex) Database. https://globalfindex.worldbank.org/.

(32)

31

Appendices

A. Diagnostic tests

Table 10: The p-values of diagnostic tests. The independent variables are listed in the first column and the explanatory variables are the same as in the main regression: share of foreign bank assets, ln (GDP/capita), banking crisis dummy, legal rights index, credit information index, cost of enforcing contracts, rail density, and population density. The sample is the final sample of the study.

(1) LM test p-value (2) Hausman test p-value (3) Time FE p-value (4) Heteroscedasticity test p-value (5) Serial correlation p-value 1. ATMs / capita 0.0000 0.0015 0.0000 0.0000 0.0000 2. ATMs / km2 _0.0000 _0.0000 _0.1131 _0.0000 _0.0000 3. Branches 0.0000 0.2529 0.0258 0.0000 0.0000 4. Borrowers 0.0000 0.0108 0.0031 0.0000 0.0000 5. Deposit accounts 0.0000 0.0301 0.0850 0.0000 0.0000 6. Loan accounts 0.0000 0.0000 0.0044 0.0000 0.0000

The p-values of the diagnostic tests with different dependent variables are summarized in Table 10.

First, using the Breusch-Pagan Lagrange multiplier (LM), we test if pooled OLS regression is preferred or a random effects regression. Random effects model is able to deal with heterogeneity better than the pooled OLS model. The null hypothesis is that the variances across entities is zero and there is no panel effect. As a consequence, a pooled OLS regression is appropriate. The results are in Table 10, column (1). Since the null hypothesis is rejected in all specifications, we conclude that RE regression is preferred over OLS.

Second, we do the Hausman test to find out if fixed effects regression is preferred over the random effects model. The null hypothesis is that the individual effects are not correlated with any of the regressors. If the null hypothesis is rejected, the estimates of a random effects model are biased and inconsistent. If the null hypothesis is not rejected, both FE and RE regression should produce similar estimates. The results are in Table 10, column (2). We reject the null hypothesis at 0.05 significance level in all specifications except in the specification where the number of branches is the dependent variable. As a conclusion, fixed effects model should be used, because the individual effects are significantly correlated with at least one of the regressors. Also in the specification where the null was not rejected, fixed effects model will be used because it is still unbiased and consistent.

(33)

32 with and without time-fixed effects. This does not change the significance of the main variable of interest.

The heteroscedasticity of the fixed effects model was tested using the Modified Wald test. The resulting p-value was 0.0000 in all specification, which means that we reject the null hypothesis and conclude heteroscedasticity was found and it will be solved using robust standard errors. Serial correlation of the error term was tested using the Lagram-Multiplier test for serial correlation. The p-value of the test was 0.0000, which means that the dataset has serial correlation. This will be solved by clustering the standard errors by country.

B. Variable definitions and sources

Table 11

Definitions and sources of all variables. Source is the original source of the data.

Variable Definition Source

Foreign banks % assets

The percentage to the total banking assets that are held by foreign banks. Bank is classified as foreign owned if more than 50% of its assets are held by foreigners

Claessens and van Horen, (2014); Claessens and van Horen (2015) Foreign banks %

banks

The percentage of the number of foreign banks to the total number of banks in an economy.

Claessens and van Horen, (2014); Claessens and van Horen (2015) ATMs /adults The number of automated teller machines (ATMs) of

all financial institutions for every 100,000 adults in the reporting country

IMF, Financial Access Survey (FAS)

ATMs /km2 _{The number of all ATMs of all financial institutions}

for every 1,000 km2 _{land area in the country}

IMF, FAS

Bank branches /adults

The number of commercial banks and their branches per 100,000 adults in the country

IMF, FAS

Borrowers /adults The total number of resident customers that are nonfinancial corporations (public and private) and individuals from the household sector who obtained loans from commercial banks for every 1,000 adults in the reporting country

IMF, FAS

Borrowers only HH Total number of resident customers that are from the household sector who obtained loans from

commercial banks for every 1,00 adults in the reporting jurisdiction

(34)

33

Deposit accounts /adults

Total number of deposit account holders that are nonfinancial corporations and individuals from the household sector at commercial banks for every 1,000 adults in the reporting jurisdiction

IMF, FAS

Deposit accounts only HH

Total number of deposit account holders that are individuals from the household sector at commercial banks for every 1,000 adults in the reporting

jurisdiction

IMF, FAS

Loan

accounts/adults

The total number of loan accounts of resident nonfinancial corporations (public and private) and individuals from the household sector that have obtained credit from commercial banks for every 1,000 adults in the reporting jurisdiction

IMF, FAS

Loan accounts (only HH)

The total number of loan accounts of individuals from the household sector that have obtained credit from commercial banks for every 1,000 adults in the reporting jurisdiction

IMF, FAS

Account (% adults) The percentage of respondents who reported having an account in formal financial institution (bank, credit union or another financial institution, such as cooperative or microfinance institution) or the post office, including the ones who reported having a debit card (%, age 15+)

World Bank, Global Financial Inclusion (Global Findex) Database

Loan (% adults) The percentage of respondents who reported

borrowing money from a formal financial institution in the past year (%, age 15+)

World Bank, Global Findex

GDP per capita GDP per capita measured in constant 2010 U.S. dollars

World Bank, World Development Indicators (WDI) Banking crisis

dummy (1=banking crisis, 0=none)

A dummy that is 1 when there is a systemic banking crisis: there are significant signs of financial distress in the banking system and significant banking policy measures in response to losses.

World Bank, Global Financial

Development Database (GFDD) Legal rights index Strength of legal rights index measures the degree to

which collateral and bankruptcy laws protect the rights of borrowers and lenders. A higher score indicates better protection.

World Bank, Doing Business database

Credit information index

The credit information index uses six components to measure the accessibility of credit data

World Bank, Doing Business database Cost of contract

enforcement

The cost of enforcing contracts as a percentage of claim, including average attorney fees, court costs and enforcement costs.

(35)

34

Rail density The total route of rail lines divided by the land areas of the country

World Bank, WDI

Population density Population in a country divided by land area. World Bank, WDI Bank overhead cost Bank overhead costs to total assets (%) World Bank, GFDD Urban population The percentage of people living in urban areas World Bank, WDI

C. Summary statistics

Table A 1

Variables: Summary statistics only including the sample: emerging and developing countries in year 2003-2014.

(1) (2) (3) (4) (5)

Variables N mean sd min max

Foreign banks % assets 930 43.46022 33.12252 0 100 Foreign banks % banks 1,050 44.61048 27.36143 0 100 ATMs /adults 892 29.98395 34.73844 0 288.632 Bank branches /adults 993 14.16432 14.36571 0.391847 142.192 ATMs /km2 _{907 32.93685 113.5352} ₀ _1274.66

(36)

35

D. List of countries

We use the countries that have data on bank ownership. The countries are grouped in income groups following the methodology of Claessens and van Horen (2015). Emerging economies include countries that are included in the S&P’s Emerging Market and Frontier Markets indices but were not classified as high-income countries by the World Bank in 2000. Developing countries include countries that are not emerging economies nor classified as high-income economies. Bahrain, Democratic Rep. Of Congo, Cuba and Romania are dropped from the final sample, because the Financial Access Survey does not have data on them. In total there are 103 countries: 43 emerging economies and 60 developing countries.

Emerging economies

Argentina, Botswana, Brazil, Bulgaria, Chile, China-People's Rep., Colombia, Croatia, Czech Republic, Ecuador, Egypt, Estonia, Hungary, India, Indonesia, Jamaica, Jordan, Korea Rep. Of, Latvia, Lebanon, Lithuania, Malaysia, Mauritius, Mexico, Morocco, Namibia, Nigeria, Oman, Pakistan, Peru, Philippines, Poland, Russian Federation, Saudi Arabia, Slovakia, South Africa, Sri Lanka, Thailand, Trinidad And Tobago, Tunisia, Turkey, Ukraine, Venezuela

Developing countries

The Effect of Global Banking on Financial Inclusion in Emerging and Developing Economies