1
Analyzing the Effect of Foreign Bank Entry: Heterogeneity
Across Countries and Banks
Edwin Höltke1
Supervisors: Dr. M.J. Gerritse & Prof. Dr. W. Bessler
Abstract
This paper analyzes the short-term effect of foreign bank entry on the profit-, cost- and income efficiency of incumbent domestic bank in developing and emerging countries. Foreign bank entry only affects the least efficient banks in terms of profits before taxes and non-interest income. Moreover, taking into account the home country regulatory quality is important, while the home country use of hard information does not make a significant difference.
Key Words: Foreign bank entry; Spillover effects; Economic development JEL Classifications: E44; F21; G21
1
2
1 Introduction
In recent decades the world globalized at an increasing pace and the banking sector was no exception. Causes for increased globalization include decreasing costs of information technologies, financial deregulation, and further economic integration with the world economy. From 1995 to 2009 the foreign bank share increased from 20% to 34% throughout the world. More interestingly, the share of foreign banks in developing countries increased from 24% to 45% in the same period (Claessens and Van Horen, 2014). Beck, Levine and Loayza (2000) argue that especially developing countries can benefit from foreign banks by liberalizing their financial sectors. This would lead to an increase in the credit provided to the domestic market. Moreover, they argue that foreign banks induce restructuring of the domestic financial sector.
This paper analyzes the effects of foreign bank entry in developing and emerging countries empirically. By using data on 3277 banks across 62 developing and 44 emerging countries between 1995 and 2013, the short-term effects of foreign bank entry is analyzed.2 More specifically, the short-term effects of foreign bank entry on the efficiency of domestic banks is considered in terms of profits, income and costs. This paper is an extension to Lensink and Hermes (2004) and Claessens, Demirgüç-Kunt and Huizinga (2001). Whereas Lensink and Hermes show that the short-term effects of foreign bank entry depends on the development level of the market entered, this paper allows for heterogeneity across (i) the markets entered, (ii) the countries of origin of the foreign banks and (iii) the efficiency of domestic banks. By allowing for heterogeneity across these three lines, this paper adds to the existing literature by analyzing if the effect of foreign entry depends on the country of origin and if banks with different degrees of efficiencies are affected differently. This leads to the following research question:
Does the short-term impact of foreign bank entry on the efficiency of domestic banks depend on (i) the country of origin of the foreign bank and (ii) the efficiency of the incumbent domestic bank?
This paper finds that taking into account the host country level of development is important. Foreign bank entry does not have a significant impact on the efficiency of domestic banks in developing countries. In emerging countries, the profits before taxes and non-interest incomes are significantly affected. Moreover, the effects are significant only for the least
3
efficient banks, while the most efficient banks are not affected by foreign entry. Turning to home country characteristics, taking into account the home country regulatory quality makes a significant difference, while the home country use of hard information does not matter.
This paper is structured as follows. The next section will discuss the relevant literature. Section 3 will discuss the data and methodology used. The results and discussion, comparison with previous studies, and robustness checks will be provided in section 4. Finally, this paper will be concluded in section 5.
2 Literature review
This part is structured as follows. Section 2.1 will introduce the channels through which foreign bank entry affects the domestic banks and market. The literature concerning the heterogeneity will be discussed in section 2.2, while the heterogeneity of the host market will be considered in section 2.3. The last section, 2.4, will introduce why it is important to consider the home country of a foreign bank as well.
2.1 Transmission channels of foreign bank entry
The impact of foreign bank entry on the domestic banking sector is transmitted via two channels, namely knowledge spillovers and competition (Goldberg, 2007). The first channel acts through a foreign bank,which holds superior knowledge to the incumbent banks, entering developing markets. By interaction between the foreign bank, the host market, and domestic banks, this superior knowledge can spill over from the foreign bank to the domestic banks. The effect of spillovers has been widely discussed in the literature on research and development (R&D). First, Coe and Helpman (1995) found that foreign R&D expenditure has a positive impact on the host market productivity. Second, Coe, Helpman and Hoffmaister (2009) argue that the ease of doing business and tertiary educational attainment both stimulate knowledge spillovers. A study by Engelbrecht (1997) supports this by stating that human capital acts as a vehicle in the process of transferring (international) knowledge spillovers.
4
either case, spillovers or investment, the screening abilities of banks increase. The authors argue that the result of investment in screening technology is superior to spillovers.
The second channel through which foreign bank entry affects the domestic market is through competition. Melitz (2003) shows that an exposure to international trade can improve the overall productivity in a market. Important in his analysis is that firms are heterogeneous. Applying this model to the banking sector, the most productive banks are able to expand abroad, while the least productive banks will be forced out of the market. The banks with an average productivity will remain active only in the domestic market.
In the empirical work, Drakos (2003) and Poghosyan and Poghosyan (2010) analyze the effects of foreign bank entry in Central and Eastern Europe. They find that the foreign banks increase the competitive pressure and generate positive externalities in the domestic banking market, both in line with the transmission channels mentioned above.
2.2 Taking into account the heterogeneity of banks
There are a large number of studies concerning the performance of foreign banks. Berger, Hasan and Zhou (2009) find that foreign banks outperform domestic banks in China when looking at cost- and profit efficiency. Bonin, Hasan and Wachtel (2005) analyze 11 Eastern European countries and conclude that foreign banks outperform domestic banks in efficiency measures, but not in terms of return on assets. In a study performed by Claessens et al. (2001), profits before taxes to total assets is used as a measure of efficiency. They find that foreign banks perform better in developing countries. However, when turning to developed countries, foreign banks are performing worse compared to domestic banks. The latter result is consistent with findings in the USA (Chang, Hasan and Hunter, 1998; DeYoung and Nolle, 1996). For a full review of previous literature on foreign bank performance, see Claessens and Van Horen (2012).
5
performance. Nevertheless, this study only analyzes the performance of foreign banks. It does not attempt to find the impact of foreign bank entry on the host country financial sector.
Foreign banks have a positive impact on the financial sectors they are active in. Especially developing and emerging economies are positively impacted by foreign bank entries (Beck et al., 2000). This may be channeled through spillovers or due to competition. Moreover, the performance of foreign banks depends on several factors (Claessens and van Horen, 2012).
However, it is not unlikely that the effect of foreign entry is heterogeneous across the domestic banking sector. First, the least efficient banks are being forced out of the market, increasing the overall efficiency across the sample of the least efficient banks. Second, the marginal returns coming from spillovers are likely to the largest (smallest) for the least (most) efficient banks, simply because they have the most (least) ground to gain. This leads to the following hypothesis:
(H1) Foreign bank entry has a significant positive impact on the efficiency of the least efficient banks, while the effect on the most efficient banks is insignificant or significantly smaller.
2.3 Taking into account the heterogeneity of host market
Previous research has taken into account several host country characteristics. More specifically, Lensink & Hermes (2004) argue that foreign banks can induce spillovers to the host market via various channels. First, foreign banks can introduce new financial products in the market, inducing domestic banks to develop new products as well. Second, domestic banks can copy more efficient banking practices brought in the market by foreign banks, stimulating the sector wide efficiency. Third, foreign banks can demand better supervision and regulation. This will lead to the financial system3 as a whole to become more sound and prudent. Finally, the entrance of foreign direct investment (FDI) may reduce the involvement of the government in the financial sector. Because government banks tend to operate inefficiently in emerging markets, reducing government involvement improves the overall efficiency (Main, 2003).
3
6
In their empirical work, Lensink and Hermes (2004) analyze the short-term effects of foreign bank entry using two different measures, namely the ratio of the number of foreign banks to the total amount of banks and the foreign assets to the total assets in a country. The impact of foreign bank entry is measured on the efficiency of income, profits, and costs. They show that, when foreign banks enter the market, domestic banks increase interest rate margins to finance the new investments required. They argue that banking markets in developing countries benefit more from spillovers in banking techniques because their financial systems are less developed. This view is consistent with King and Levine (1993), who show that developing countries generally have a lower developed financial system. It thus shows that taking into account the host country characteristics is vital when analyzing foreign bank entry.
2.4 Taking into account the heterogeneity of home market
To illustrate the importance of taking into account the home country of a foreign bank, consider the following example. The world consists of three countries: North, South One and South Two. North has a highly developed financial sector and South One and South Two both have a poorly developed financial sectors. Now imagine that a bank from North expands to South One. The expanding bank brings new banking techniques for screening borrowers, introduces new financial products for firms to hedge their risks, stimulates improvements in regulation and supervision and so on. Domestic banks in South One undergo investments to copy the new banking techniques and introduce new financial products. The supervising and regulatory authorities will improve their standards. Overall, the banking sector is likely to improve in efficiency and stability due to the entrance of a bank from a more developed country.
Now consider a bank expanding from South Two into South One (or vice versa), both equally developed. Even though only the most productive participants in an industry generally move abroad (Melitz, 2003), the expanding bank is only expected to be only marginally more efficient than (most) banks in the host (and home) country. Moreover, Claessens and Van Horen (2012) show that banks from developing countries generally perform worse when expanding abroad. This implies that the spillover effects from banks originating in developing countries are small or non-existent. The only way that the bank from South Two is able to increase the efficiency in the market, is via increasing the competitive pressure.
7
foreign entry on the domestic market. One way in which home countries can be different is the use of hard and soft information. Hard information refers to quantitative information which is easy to store and transmit and is independent of the collection procedure (e.g. balance sheet data, payment history). Information which is harder to store in a single numeric score is referred to as soft information (e.g. entrepreneurial skills). Hard information is associated with lower collection cost, longer durability and easier communication within and across organizations. However, in the collection process it is inevitable for some information to become lost. Nevertheless, the hardening of information is associated with reduced costs for banks and possibilities to expand beyond their usual scope of customers (Petersen, 2004).
Another aspect influencing bank performance is the regulatory quality of the host and home country (Claessens and Van Horen, 2012). Good regulation allows banks to focus on their main tasks and stimulates the commitment a country is willing to take in a certain country. For instance, good contract law and enforcement stimulates banks to lend more as it reduces uncertainty about future repayments. Foreign banks can stimulate good (financial) regulation in their host country by stimulating local governments to improve their regulation in return for their commitment.
This example and the home country characteristics – use of hard information and regulatory quality – mentioned above result in the following two hypotheses:
(H2) First, Foreign banks with a home country characteristic above average have a significant positive impact on the efficiency of domestic banks upon entry.
(H3) Second, Foreign banks with a home country characteristic below average have a significantly smaller or insignificant impact on the efficiency of domestic banks upon entry.
8
3 Data & Methodology 3.1 Data
In order to analyze the effect of foreign entry on the performance of incumbent domestic banks, data provided by Claessens and Van Horen (2015)4 is used. This database tracks the ownership status of banks in 139 countries with a total of 5498 banks between 1995 and 2013. It covers at least 90% of each banking system in terms of assets (Claessens and Van Horen, 2014) and states the home country for all foreign banks. This allows the incorporation of home country characteristics. To ensure comparability across the sample, the analysis is restricted to commercial banks only, reducing the sample to 4327 individual banks.5 Because the main interest of this paper is to find the effect of foreign entry on developing and emerging countries, the sample is further restricted to the 1150 banks active in developing countries and the 2127 banks in emerging countries.
The database has an ownership indicator for every year that a bank is active. A bank is defined as foreign if 50% or more of its shares are owned by foreign investors. For these banks, Claessens and Van Horen (2015) sum the percentages of foreign shares held by country of residence. The home country is the country with the largest stake in the bank. Note that it is thus possible that, if the majority of the shares is foreign owned, the shares held in the registered home country are below 50%. It is now possible to calculate the number of foreign banks to the total number of banks active in each country (FBNUM).
An important facet of the database is that it provides unique index numbers for every bank in the database to be combined with Bankscope database, provided by Bureau van Dijk6. This allows me to match the banks without encountering double accounting. By combining the Bankscope database with the ownership database, it is possible to calculate the relative size of foreign banks. This is done by dividing all foreign bank assets in each country by the total bank assets (FBSHR). The methods to calculate FBNUM and FBSHR are in line with Claessens et al. (2001) and Lensink & Hermes (2004).
4
Bank Ownership Database, http://www.dnb.nl/en/onderzoek-2/databases/bank.jsp.
5
Actually, there are 4521 commercial banks in the sample. However, after removing banks with an unknown ownership status, 4405 banks remain. Several other observations are removed, as will be discussed later. Banks in Ecuador are removed from the database due to significant outliers across multiple banks and multiple years. For instance, it is not unusual for banks in Ecuador to have deposits and short-term funding over 10 times as much as total assets. In the end, there are 4327 banks in the database.
9
The reason for using two different measures of foreign bank entry, FBNUM and FBSHR, is as follows. On the one hand, foreign bank presence as measured in the number of banks captures true entry into a market. However, it might also be the case that the size of an entry matters. A large entry is likely to put both more pressure on the domestic market and increase the possibility of spillovers between the foreign and domestic banks, compared to a small entry. Therefore, measuring entry in terms of assets seems more sensible. However, the drawback is that foreign bank presence in terms of assets does not only measure the entry of foreign banks. If a foreign bank, which is already present in the market, expands its market share in terms of assets, it will show up in the first difference of the foreign entry measure in assets. Moreover, when a bank is losing market share, it will show up in the data as a foreign bank exit. Therefore, the measure in terms of assets could be measuring more than entry alone. To see if there are any differences between the two measures, both are included in the initial analysis in the empirical section.
All host countries and their respective income group, foreign bank presence and the number of foreign banks in 1995 and 2013 are reported in Table 1. The income group is as defined in Claessens and Van Horen (2014). The abbreviations DEV and EM stand for developing economy and emerging economy, respectively. The foreign bank presence in Table 1 is the average FBNUM over the years 1995-2013. Table 2 summarizes the foreign bank share per income group and per region. Only developing and emerging countries are reported since this research is only concerned with these countries.
These two tables reveal several interesting insights. First, the foreign bank presence is, on average, similar for developing- and emerging countries. Second, there seems to be some variation within each income group. For instance, within the developing countries the foreign presence ranges from 3% in Bangladesh to 95% in Madagascar. This is further illustrated by the standard deviations, which are are about half the size of the average for both groups. Third, the number foreign banks in each country between 1995 and 2013 has increased for almost every country. Turning to the regions, the foreign bank presence differs as well. It ranges from 11% in South Asia to 51% in Sub-Saharan Africa, albeit that the number of countries within regions also differs from 5 to 31.
10
this has increased to 78 banks. Note that it is possible for a country to be a home country in the
Table 1: Foreign bank shares in host countries, as measured by the number of foreign banks
Country Income Group Share of foreign banks No. of foreign banks in 1995 No. of foreign banks in 2013 Country Income Group Share of foreign banks No. of foreign banks in 1995 No. of foreign banks in 2013
Albania DEV 0.76 1 10 Tanzania DEV 0.57 6 17
Algeria DEV 0.44 1 9 Togo DEV 0.17 0 1
Angola DEV 0.47 2 6 Uganda DEV 0.73 8 15
Antigua and Barbuda DEV 0.24 1 2 Uruguay DEV 0.88 33 18
Armenia DEV 0.47 2 11 Uzbekistan DEV 0.22 2 3
Azerbaijan DEV 0.11 1 3 Vietnam DEV 0.14 3 9
Bangladesh DEV 0.03 0 1 Yemen DEV 0.00 0 0
Barbados DEV 0.67 1 3 Zambia DEV 0.71 5 15
Belarus DEV 0.40 2 13 Zimbabwe DEV 0.48 3 5
Benin DEV 0.81 5 7 Argentina EM 0.35 27 20
Bolivia DEV 0.38 4 3 Bahrain EM 0.59 5 9
Bosnia-Herzegovina DEV 0.40 2 12 Botswana EM 0.77 4 6
Burkina Faso DEV 0.88 3 7 Brazil EM 0.34 38 50
Burundi DEV 0.30 1 2 Bulgaria EM 0.58 6 16
Cambodia DEV 0.46 3 11 Chile EM 0.49 14 12
Cameroon DEV 0.60 3 8 China-People's Rep. EM 0.13 5 29
Congo, Democratic Rep. of DEV 0.61 2 10 Colombia EM 0.32 8 8
Costa Rica DEV 0.43 10 7 Croatia EM 0.33 3 15
Cote d'Ivoire DEV 0.75 6 10 Czech Republic EM 0.52 11 11
Cuba DEV 0.00 0 0 Egypt EM 0.32 2 13
Dominican Republic DEV 0.10 2 3 Estonia EM 0.51 1 6
El Salvador DEV 0.66 2 10 Hungary EM 0.80 19 19
Ethiopia DEV 0.00 0 0 India EM 0.10 5 7
Georgia Rep. of DEV 0.36 0 10 Indonesia EM 0.36 23 25
Ghana DEV 0.52 5 11 Jamaica EM 0.59 3 6
Guatemala DEV 0.29 4 8 Jordan EM 0.23 1 4
Haiti DEV 0.00 0 0 Korea Rep. of EM 0.12 0 2
Honduras DEV 0.36 4 9 Latvia EM 0.41 2 11
Iran DEV 0.00 0 0 Lebanon EM 0.34 18 16
Kazakhstan DEV 0.34 5 11 Lithuania EM 0.51 0 6
Kenya DEV 0.24 8 10 Malaysia EM 0.43 14 18
Kyrgyzstan DEV 0.60 0 4 Mauritius EM 0.70 6 9
Libya DEV 0.00 0 0 Mexico EM 0.48 13 15
Macedonia (Fyrom) DEV 0.40 0 7 Morocco EM 0.36 5 4
Madagascar DEV 0.95 3 6 Namibia EM 0.44 2 2
Malawi DEV 0.34 2 2 Nigeria EM 0.11 3 3
Mali DEV 0.37 1 5 Oman EM 0.02 0 1
Mauritania DEV 0.24 1 3 Pakistan EM 0.22 1 9
Moldova Rep. of DEV 0.25 0 6 Peru EM 0.61 7 11
Mongolia DEV 0.10 0 1 Philippines EM 0.18 4 5
Montenegro DEV 0.48 0 7 Poland EM 0.67 14 31
Mozambique DEV 0.82 1 10 Romania EM 0.63 4 21
Nepal DEV 0.18 4 3 Russian Federation EM 0.13 16 29
Nicaragua DEV 0.49 2 3 Saudi Arabia EM 0.00 0 0
Niger DEV 0.79 3 4 Slovakia EM 0.66 8 8
Panama DEV 0.67 45 40 South Africa EM 0.29 7 6
Paraguay DEV 0.60 12 7 Sri Lanka EM 0.00 0 0
Rwanda DEV 0.33 2 4 Thailand EM 0.17 1 5
Senegal DEV 0.61 3 7 Trinidad and Tobago EM 0.55 3 6
Serbia DEV 0.35 1 19 Tunisia EM 0.43 5 8
Seychelles DEV 0.43 1 2 Turkey EM 0.27 7 14
Sudan DEV 0.15 1 3 Ukraine EM 0.28 2 22
Swaziland DEV 0.74 4 3 Venezuela EM 0.26 5 7
11
Table 2: Foreign bank presence per income group and per region Mean of foreign bank share Std. Dev. Of foreign bank share Number of countries Income Groups Developing 0.42 0.26 62 Emerging 0.38 0.21 44 Regions
Eastern Europe and Central Asia 0.44 0.18 26
Middle East and North Africa 0.23 0.21 12
Sub-Saharan Africa 0.51 0.25 31
Latin America and Caribbean 0.42 0.22 23
South Asia 0.11 0.09 5
East Asia and Pacific 0.23 0.14 9
Note: This Table shows the unweighted average and standard deviation.
Table 3: The number of foreign banks per home country, which are active in developing and emerging countries.
Country No. of foreign banks in 1995 No. of foreign banks in 2013 Country No. of foreign banks in 1995 No. of foreign banks in 2013 Country No. of foreign banks in 1995 No. of foreign banks in 2013
United States 74 78 Mauritius 1 7 Liechtenstein 2 2
France 60 72 Nicaragua 3 7 Bangladesh 1 1
United Kingdom 40 70 Slovenia 1 7 Barbados 0 1
Germany 23 37 Cyprus 2 6 Botswana 0 1
Russian Federation 3 33 Norway 0 6 Cuba 0 1
Austria 14 30 Taiwan 3 6 Czech Republic 0 1
Italy 14 30 Hong Kong 3 5 Estonia 0 1
South Africa 15 24 Israel 5 5 Iceland 0 1
Netherlands 32 23 Saudi Arabia 5 5 Ireland 1 1
Nigeria 2 23 Ukraine 0 5 Jamaica 0 1
Canada 10 22 Venezuela 8 5 Paraguay 0 1
Colombia 9 22 Bahrain 2 4 Slovakia 0 1
Greece 3 22 Belgium 7 4 Trinidad and Tobago 0 1
Spain 23 22 Denmark 0 4 Uzbekistan 0 1
Togo 5 22 Egypt 3 4 Vietnam 0 1
Japan 18 21 Guatemala 0 4 Gabon 0 1
Switzerland 11 18 Mexico 1 4 Gambia 0 1
Libya 9 17 Pakistan 3 4 Iraq 1 1
Portugal 5 17 Peru 3 4 San Marino 0 1
Korea Rep. of 9 15 Thailand 1 4 Syria 1 1
Brazil 19 14 Australia 4 3 Virgin Islands, British 0 1
Turkey 5 13 Costa Rica 3 3 Belarus 1 0
Singapore 7 12 Jordan 1 3 Bolivia 1 0
India 7 11 Luxembourg 1 3 Burkina Faso 0 0
Kenya 1 11 Malawi 0 3 Croatia 1 0
Lebanon 0 11 Panama 10 3 El Salvador 2 0
Mali 2 11 Qatar 0 3 Finland 0 0
Sweden 2 10 Serbia 0 3 Indonesia 1 0
United Arab Emirates 0 10 Bulgaria 0 2 Lithuania 0 0
Argentina 13 9 Dominican Republic 2 2 Romania 0 0
China-People'S Rep. 1 9 Georgia Rep. of 0 2 Uganda 1 0
Kazakhstan 0 9 Honduras 0 2 Zambia 1 0
Malaysia 6 9 Iran 1 2 Andorra 0 0
Morocco 0 9 Latvia 0 2 Bermuda 0 0
Hungary 1 8 Oman 0 2 Cayman Islands 1 0
Chile 1 7 Philippines 1 2 Malta 0 0
Ecuador 7 7 Poland 0 2 Netherlands Antilles 0 0
Kuwait 3 7 Uruguay 5 2
12
database but not a host country. This is because the banking sector has been analyzed for 139 countries only, while foreign banks can come from other countries as well.7
To account for the first hypothesis, I split the domestic banks in each country into two groups. Banks are, in every year and country, assigned to a group based on its profits before taxes to total assets. The 50% most profitable banks are placed in the group top 50% banks and the 50% least profitable banks are grouped into the bottom 50% banks. This method allows me to analyze hypothesis 1.
Turning to the second and third hypothesis, to measure the home country effects, foreign banks will be divided into two groups. The first group includes all foreign banks which have a home country characteristic (i.e. use of hard and soft information or regulatory quality) above the mean for a specific year. The second group includes all foreign banks which have a home country characteristic measured below the mean for a specific year. Important to note is that foreign banks can move from one group to the other in subsequent years. Next, for each host country and year, the number of banks in both groups is divided by the total amount of banks in each country in each year. In other words, for each host country and year, I calculate the ratio of foreign banks in each group to the total amount of banks.
The Doing Business Report depth of creditor information (henceforth creditor information)8 will be used to proxy for the use of hard and soft information (Claessens and van Horen, 2012). This indicator measures the rules and practices affecting the availability and scope of creditor information in a country. The value is an integer between 0 (lowest) and 6 (highest). Data is available for the years 2004-20169. A higher value implies a greater possibility to use hard information in a country, whereas a lower value indicates that the use of soft information is more likely. The foreign bank shares split on creditor information will be FBNUM depth AM (Above Mean) and FBNUM depth BM (Below Mean) for the banks with an above- or below average home country creditor information, respectively.
The measure for regulatory quality is taken from the Worldwide Governance Indicators (Kaufmann, Kraay, Mastruzzi, 2011) . The regulatory quality indicator captures
7
The countries which are not host countries but have banks active in developing and emerging countries are: Andorra, Bermuda, Cayman Islands, Gabon, Gambia, Iraq, Liechtenstein, Malta, Netherlands Antilles, San Marino, Syria and the British Virgin Islands.
8
World Bank, Doing Business Report. http://www.doingbusiness.org/data.
9
13
the “perceptions of the ability of the government to formulate and implement sound policies and regulations that permit and promote private sector development” (Kaufmann et al., 2011). The measure can take any value between -2.5 and 2.5, where a higher figure indicates a better perceived ability of the government to formulate and implement sound policies. Data is available for the years 1996 to 2014, where the data is recorded biyearly between 1996 and 2002.10 For the regulatory quality, the foreign bank shares are named FBNUM RQ AM and FBNUM RQ BM for above- and below average home country regulatory quality, respectively. This measure is used by Claessens and van Horen (2012) as well.
The effect of foreign bank entry will be measured on the efficiency of domestic banks. The efficiency will be measured along the lines of income, profits and costs of incumbent domestic banks. Income is measured by the net interest income to total assets (NMARGIN) and the non-interest income to total assets (NINTINC). Profits are measured by the before tax profits to total assets (PROF) and the costs are measured by the total overhead costs to total assets (OVERHEAD) and the loan loss provision to total assets (LLPROV). All the data to construct these five variables are obtained from Bankscope and are also used by Claessens et al. (2001) and Lensink & Hermes (2004).
The following variables will be included to control for bank specific characteristics and are all obtained from Bankscope. First, equity to total assets (EQUITY) is included to control for the cost of risk. As markets associate low levels of capital with a higher probability of default, banks with low degrees of equity incur higher costs to obtain funds (De Young & Nolle, 1996). Second, banks can hold cash for various reasons. Among other things, it depends on the type of bank, services provided, and regulation. The amount of cash held is different for every bank in a country and affects its income, profits or costs. Therefore, the cash and due from banks (NINTASS) is included as well. Third, the overhead costs (OVERHEAD) is a measure to control for differences in costs efficiencies among domestic banks. For instance, banks with large overhead costs might have lower profits before taxes compared to banks with lower overhead costs. The same goes for the net interest income and non-interest income. However, introducing the overhead costs in the model might cause some problems. For instance, foreign bank entry might affect the dependent variables in question via a spillover effect on the overhead costs. Even though this is a genuine concern, it turns out
10
14
that foreign bank entry does not have an effect on the overhead costs (see Table 9, column 2 in the next section). The overhead costs measure thus does not capture part of the foreign bank entry effect. Including it does not bias the estimations.
Controls at the country level are also included in the model. Growth domestic product per capita (GDPPerCapita)11, GDP growth (GDPGrowth) and inflation (Inflation) are included to control for country-year specific effects and are obtained from the World Development Indicators12. Both the bank and country controls are, generally, in line with Claessens et al. (2001) and Lensink & Hermes (2004).
To summarize, this paper analyzes the effect of foreign bank entry on the incumbent domestic banks by taking into account both home- and host country characteristics. Data on 1150 banks in developing countries and 2127 banks in emerging countries from 1995 to 2013 is used to analyze the effects of foreign bank entry. The impact is measured along the lines of income, profits and costs of incumbent domestic banks. Domestic banks are split based on their efficiency and foreign banks are split according to their home country characteristics. This allows me to test the three hypotheses.
3.2 Methodology
The empirical models, based on Claessens et al. (2001) and Lensink & Hermes (2004), are as follows:
∆𝐼𝑖𝑗𝑡 = ∆𝛼0+ 𝛽∆𝐹𝐵𝑃𝑗𝑡+ 𝛿𝑖∆𝐵𝑖𝑗𝑡+ 𝜃𝑗∆𝑋𝑗𝑡+ 𝜀 (1)
∆𝐼𝑖𝑗𝑡 = ∆𝛼0+ 𝛽∆𝐹𝐵𝑃(> 𝑚𝑒𝑎𝑛̃ )𝑗𝑡+ 𝛾∆𝐹𝐵𝑃(< 𝑚𝑒𝑎𝑛̃ )𝑗𝑡+ 𝛿𝑖∆𝐵𝑖𝑗𝑡+ 𝜃𝑗∆𝑋𝑗𝑡+ 𝜀 (2) Where 𝐼𝑖𝑗𝑡 is the dependent variable measuring the income, profitability or costs of domestic bank i in country j at time t. 𝐹𝐵𝑗𝑡 is the measure of Foreign Bank Presence in country j at time t (i.e. FBNUM or FBSHR). 𝐵𝑖𝑗𝑡 and 𝑋𝑗𝑡 are vectors of control variables at the bank- and country level, respectively. Eq. (1) analyzes the overall impact of foreign bank entry on the domestic banks, while Eq. (2) allows me to take into account home country characteristics.
11 Ideally, real GDP per capita would be used. However, this reduces the available observations for developing
and emerging countries significantly. Using GDP per capita in 2005 US$ instead of real GDP does not alter the significantly alter the results.
12
15
Controlling for every country-year combination would be ideal. However, because the foreign bank share is constant for all banks in each country and year, it is perfectly collinear with a country-year control dummy variable. Therefore, controlling for each country and year separately is the second best option. To limit the drawback of this second best approach, country level control variables are included, as mentioned in the previous section.
The models are estimated in first differences, as indicated by the ∆ signs. The main interest is the effect of foreign bank entry on the domestic market. In a first differences model, the change in the independent variables is regressed on the change in the dependent variable. As the data described above includes a foreign share variable for each country in each year, the change in the share of foreign banks is regressed on the change in any of the five dependent variables. This change in foreign bank share represents, in essence, the entry of foreign banks in the market13. For instance, if the coefficient of the FBNUM is positive, foreign bank entry is associated with an increase in the dependent variable in question.
This interpretation is slightly different compared to a fixed effect model. If Eq. (1) would be estimated in a fixed effect model, the interpretation of a positive coefficient for FBS would indicate that countries with a higher share of foreign banks are associated with a higher value of the dependent variable. Intuitively, it seems to be similar. However, on the one hand the first differences model explicitly measures the direct effects of a change in the ratio of foreign banks (i.e. foreign bank entry). On the other hand, a fixed effect model indicates whether countries, for instance, with a high ratio of foreign banks are associated with higher or lower profits, relative to countries with low ratios of foreign banks. From the fixed effect model, it is therefore not possible to directly measure the effect of foreign bank entry. Hence, because our main interest lies in the effect of foreign bank entry on the host market, a first difference model is preferred. This is in line with both Claessens et al. (2001) and Lensink & Hermes (2004).
All equations include country- and time-specific fixed effects dummies (not reported). No constant term is included because it is mathematically cancelled out with first differencing. Furthermore, the estimations use clustering of the robust standard errors at the country-year level to control for possible heteroscedasticity. Countries with less than 3 banks in a certain year are dropped from the data to ensure proper coverage of the banking sectors.
13
16
Moreover, only domestic bank observations are used. In other words, besides the calculation of foreign bank shares, foreign banks are not used in the analysis. This is, for the most part, in line with to Claessens et al. (2001) and Lensink & Hermes (2004). Finally, the regressions only include developing- and emerging host countries. This analysis, in effect, measures the short-term effects of foreign bank entry because the change of FBS in year t is regressed on the dependent variable in year t.
The regressions will be run with developing- and emerging host countries, both separately and combined to see if there are any significant differences. It is tested whether or not the coefficients for the foreign entry measures are significantly different between different samples and between the two foreign bank presence variables when split in two. First, the coefficients are tested between different income groups. Second, it is tested if foreign banks with a below average home country characteristic have a significantly different impact on the domestic market compared to banks with an above average regulatory quality. If the coefficients are significantly different at the 10% level, both coefficients are in bold.14 The footnotes of the tables state which variables are significantly different from one another and which are not, meaning across different samples or within.
4 Empirical results and discussion
Section 4.1 the results will be presented and discussed for profits before taxes as a measure of efficiency, while section 4.2 will discuss the remain measures of income and costs as measures for efficiency. The results will be compared with previous studies in section 4.3 and robustness checks will be performed in section 4.4.
4.1 Profitability as a measure of efficiency
Profits are the bottom line for all banks, thus indicating the overall effect of foreign bank entry. After all, the income and costs measures all affect the profits before taxes of a bank. Table 4 shows the results for ten variations on the variations of Eq. (1) and (2). All variables are as explained above. At the bottom of all tables the income groups used in the regressions are indicated. Due to data availability constraints, the period in question throughout the paper is 2004-2013 when splitting foreign banks based on home country
14
17
creditor information, 1996-2013 when foreign banks are split according to the home country regulatory quality and 1995-2013 all other regressions.
Turning to the results in Table 4, column 1 shows the impact of foreign bank entry, in terms of numbers, on the profits before taxes to total assets of domestic bank. Both developing and emerging countries pooled together. Foreign bank entry has a significant positive impact on the profitability of domestic banks. The results can be interpreted as follows.An 1% increase in the foreign bank share increases profitability of domestic banks by 0.04%. The result is significant at the 5% level. When turning to the foreign bank presence in terms of assets in column 2, it turns out that the foreign entry has an insignificant impact on the profitability of the domestic banks.
As was already considered in the previous section, the two measures of foreign bank entry give different results, both in terms of significance and in the sign of the effect. Claessens et al. (2001) concluded that the impact is transmitted immediately after foreign entry instead of after having gained more market share in terms of assets. Because the FBSHR has an insignificant effect on the domestic market and Claessens et al. (2001) concluded that the measure in numbers is better, the following regressions will all include FBNUM only.
Looking at different income groups, the sample is split based on the host country level of development. Regression 3 measures the effect of entry on domestic banks in developing countries and the fourth model estimates it for emerging countries. Interestingly, foreign bank entry does not have any significant impact on the banking markets in developing countries. The opposite is true for emerging markets, where foreign entry has a positive impact on the profitability of the domestic banks. Moreover, the coefficients in columns 3 and 4 are significantly different. It thus shows that foreign bank entry has a significant effect on the profits before taxes of domestic banks in emerging countries. The effect on developing countries is insignificant. The remaining models in Table 4 will use only domestic banks in emerging countries.
18
However, the effect of banks below average have a significant impact as well and the effects are not significantly different. This concludes that the H2 is accepted, while H3 is rejected.
Turning to column 6, banks with an above average home country regulatory quality have a significant positive impact on the profitability of incumbent domestic banks. Banks with below average regulatory quality, however, do not have any significant impact on the domestic banks in emerging countries. The difference, though, is not significant. Nevertheless, in terms of regulatory quality and profitability of domestic banks, I can accept both H1 and H2.
In columns 7 – 10, domestic banks are split according to their profitability. Looking at columns 7 and 8 first, it is evident that the least profitable banks are affected by foreign entry while the most profitable banks are not significantly affected. Moreover, the coefficients are statistically different. This indicates that the least profitable banks are benefiting most from foreign entry, for instance via spillovers. The most profitable banks are not affected. I interpret these results as that the least profitable banks are able to improve more, relative to the most profitable banks. This is simply because they have more ground to gain. This is also in line with H1, which I accept when looking at the profits before taxes as a measure of efficiency. Columns 1 and 2 in Table A.4 (all tables with the A. prefix are in the appendix) show that the results are insignificant, when looking at developing countries.
The last two regressions analyze the effect of foreign entry on both the least- and most efficient domestic banks if the foreign banks are split according to their home country regulatory quality. The results indicate that only the least profitable banks are affected by foreign banks with an above average regulatory quality. All other coefficients are insignificant. This suggests that the least efficient banks learn most from bank originating from a country with an above average regulatory quality.
Columns 1 and 2 in Table A.6 show the effect of foreign bank entry on the least- and most profitable domestic banks if the foreign banks are grouped based on the creditor information. Neither the least- or most profitable banks are affected by foreign entry. Foreign entry thus does not seem to affect the profits of domestic banks, when split on the home country creditor information.
19
Table 4: The effect of foreign bank entry on profits before taxes by taking into account home- and host-country characteristics and domestic bank level characteristics
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF top
50% banks ∆PROF bottom 50% banks ∆PROF top 50% banks ∆PROF bottom 50% banks ∆FBNUM 0.0483** 0.00474 0.0788** 0.0278 0.136** (0.0227) (0.0292) (0.0344) (0.0270) (0.0589) ∆FBSHR -0.000674 (0.0117) ∆FBNUM depth AM 0.0700* (0.0357) ∆FBNUMdepth BM 0.0616** (0.0313) ∆FBNUM RQ AM 0.0906** 0.0478 0.139** (0.0376) (0.0339) (0.0587) ∆FBNUM RQ BM 0.0536 0.00307 0.104 (0.0416) (0.0332) (0.0717) ∆OVERHEAD -0.0680** -0.0680** -0.0562 -0.0677* -0.0587 -0.0695* -0.0130 -0.136* -0.0112 -0.153* (0.0311) (0.0311) (0.0427) (0.0388) (0.0478) (0.0410) (0.0235) (0.0707) (0.0230) (0.0793) ∆CSTFUN 0.00347 0.00337 0.0236 -0.00857 -0.0231 -0.00711 -0.0417 0.0434 -0.0364 0.0434 (0.0255) (0.0255) (0.0566) (0.0288) (0.0259) (0.0291) (0.0354) (0.0298) (0.0356) (0.0297) ∆EQUITY 0.248*** 0.248*** 0.178** 0.311*** 0.174*** 0.301*** 0.205** 0.424*** 0.189** 0.419*** (0.0582) (0.0582) (0.0870) (0.0715) (0.0608) (0.0727) (0.0827) (0.0798) (0.0811) (0.0823) ∆NINTASS -0.0265 -0.0265 -0.0206 -0.0267 -0.0427* -0.0330 0.0378 -0.0433 0.0313 -0.0461 (0.0187) (0.0188) (0.0343) (0.0208) (0.0245) (0.0204) (0.0236) (0.0310) (0.0209) (0.0325) ∆GDPGrowth 0.000444 0.000447 1.86e-05 0.00126** -0.000518 0.00119* 0.000727 0.00151 0.000525 0.00150 (0.000317) (0.000317) (9.44e-05) (0.000615) (0.000392) (0.000643) (0.000455) (0.00103) (0.000459) (0.00107)
∆GDPPerCapita 1.85e-06 2.00e-06 4.03e-07 4.08e-06 5.60e-06 4.78e-06 -1.36e-06 8.56e-06 -3.65e-08 8.81e-06
(4.99e-06) (5.00e-06) (3.11e-06) (6.84e-06) (4.47e-06) (6.98e-06) (5.27e-06) (1.29e-05) (5.14e-06) (1.35e-05)
∆Inflation -8.63e-06 -9.67e-06 6.62e-06 -4.98e-06 0.000722 -2.12e-05 3.69e-05 -6.07e-05 2.13e-05 -8.34e-05
(1.96e-05) (1.97e-05) (1.06e-05) (4.96e-05) (0.000508) (4.87e-05) (6.03e-05) (5.49e-05) (5.24e-05) (6.04e-05)
Observations 12,951 12,951 4,032 8,919 5,480 8,521 4,342 4,576 4,158 4,362
R-squared 0.163 0.163 0.137 0.196 0.088 0.190 0.167 0.259 0.150 0.261
Country FE YES YES YES YES YES YES YES YES YES YES
Year FE YES YES YES YES YES YES YES YES YES YES
DEV Countries YES YES YES NO NO NO NO NO NO NO
EM Countries YES YES NO YES YES YES YES YES YES YES
Note: The regressions are estimated in first differences, as indicated by the “∆”. Only the effect on domestic banks is being estimated. Country and year fixed effects are included. Robust standard errors clustered at the
20
entry seems to positively influence the non-interest income activities of the least profitable banks, which in turn will increase the profits before taxes. An increase in non-interest income is not evident for the most profitable banks. This thus explains why the most profitable banks see no increase in the profits before taxes.
In summary, it appears that taking into account the heterogeneity of host and home countries as well as the heterogeneity of domestic banks are both important. The profits in the developing countries are not significantly affected by foreign bank entry, while the banking markets in emerging economies are positively and significantly affected. Moreover, the sheer presence of foreign banks seems to matter more than the size of the entry. Overall, if there is any significant effect on profits it is always positive. The results indicate that the most profitable banks are not affected while the least profitable banks are positively affected, confirming the first hypothesis. On the one hand, banks with either an above or below average home country creditor information both have a significant impact on the profits of domestic banks, while the effects are not significantly different from each other. The home country regulatory quality does matter, confirming both the second and third hypothesis when looking at the profits of domestic banks and regulatory quality of the foreign banks’ home country.
4.2 Income and costs as measures of efficiencies
The previous section discussed the effect of foreign entry on profits, based on different samples and specifications. This section will deal with the remaining measures of efficiency, namely the non-interest income, net interest income, overhead costs, and the loan loss provision. As mentioned above, it is suggested that the cause of the increase in profits is due to an increase in the non-interest income. The columns labelled NINTINC represent the results for the non-interest income. Starting with the Table 5, it appears that the non-interest income is positively affected by foreign bank entry, at least in emerging countries. Foreign entry as measured in assets (Table A.1) or when looking at developing countries (Table A.2) shows an insignificant effect on the non-interest income.
21
Table 5: The effect of foreign bank entry on the performance of domestic banks in emerging countries
(1) (2) (3) (4) (5)
∆PROF ∆NMARGIN ∆NINTINC ∆LLPROV ∆OVERHEAD
∆FBNUM 0.0788** -0.00349 0.0518** -0.0287 0.0598 (0.0344) (0.0164) (0.0228) (0.0202) (0.0608) ∆OVERHEAD -0.0677* 0.0262*** -0.0840* 0.0653 (0.0388) (0.00726) (0.0447) (0.0666) ∆CSTFUN -0.00857 -0.0338*** 0.00646 0.00624 -0.187* (0.0288) (0.0113) (0.0130) (0.0194) (0.103) ∆EQUITY 0.311*** 0.0906*** 0.0302* -0.148*** -0.161 (0.0715) (0.0181) (0.0156) (0.0484) (0.115) ∆NINTASS -0.0267 -0.00160 0.0126 -0.00102 0.273** (0.0208) (0.0112) (0.0141) (0.0157) (0.106) ∆GDPGrowth 0.00126** -0.000515** 0.000335 -0.00162*** -0.00276* (0.000615) (0.000230) (0.000385) (0.000552) (0.00146)
∆GDPPerCapita 4.08e-06 8.63e-06*** -4.59e-06 3.42e-07 4.92e-06
(6.84e-06) (2.87e-06) (3.86e-06) (4.44e-06) (1.81e-05)
∆Inflation -4.98e-06 4.93e-05** 0.000204*** 0.000144** 0.000376
(4.96e-05) (2.12e-05) (4.73e-05) (5.76e-05) (0.000240)
Observations 8,919 8,908 8,875 7,846 8,919
R-squared 0.196 0.124 0.072 0.146 0.090
Country FE YES YES YES YES YES
Year FE YES YES YES YES YES
DEV Countries NO NO NO NO NO
EM Countries YES YES YES YES YES
Note: See footnote of Table 4. NMARGIN (net interest income), NINTINC (net non-interest income), LLPROV (loan loss provisions) and
OVERHEAD (overhead) are all relative to assets.
Next, the foreign banks are split into group based on their home country characteristics. Table 6 shows the results based on the creditor information. Both groups of banks have positive and significant influence on the non-interest income. Moreover, the difference between the two effects are not significantly different from each other. This suggests that the home country creditor information does not determine the effect of foreign bank entry. Therefore I can accept H2, which states that banks with an above average home country creditor information positively impact efficiency of the domestic banks. H3 cannot be accepted in this case, because both coefficients are significant and not different from each other.
22
not push domestic banks towards more fee-based activities. H2 and H3 are accepted when looking at the regulatory quality and non-interest income as a measure of efficiency.
Table 6: The effect of foreign bank entry on the performance of domestic banks, controlling for the home country creditor information
(1) (2) (3) (4) (5)
∆PROF ∆NMARGIN ∆NINTINC ∆LLPROV ∆OVERHEAD
∆FBNUMdepth AM 0.0700* -0.00369 0.0628** -0.00867 0.140 (0.0357) (0.0147) (0.0316) (0.0181) (0.106) ∆FBNUM depth BM 0.0616** -0.00422 0.0453* -0.00831 0.0545 (0.0313) (0.0140) (0.0272) (0.0176) (0.134) ∆OVERHEAD -0.0587 0.0202*** -0.108** -0.0251 (0.0478) (0.00588) (0.0536) (0.0440) ∆CSTFUN -0.0231 -0.0167* -0.0103 0.0168 -0.282** (0.0259) (0.00963) (0.0156) (0.0146) (0.143) ∆EQUITY 0.174*** 0.0428** -0.00183 -0.101** -0.158 (0.0608) (0.0170) (0.0284) (0.0482) (0.227) ∆NINTASS -0.0427* 0.00797 -0.00237 -0.00122 0.262* (0.0245) (0.00744) (0.0137) (0.0150) (0.138) ∆GDPGrowth -0.000518 -0.000128 -0.000646 -0.000242* -0.00484** (0.000392) (0.000160) (0.000396) (0.000143) (0.00239)
∆GDPPerCapita 5.60e-06 2.04e-06 -2.43e-06 -8.35e-06*** 1.46e-05
(4.47e-06) (2.35e-06) (3.84e-06) (3.14e-06) (2.63e-05)
∆Inflation 0.000722 4.81e-05 3.71e-05 -0.000496 -0.000941*
(0.000508) (6.71e-05) (0.000164) (0.000317) (0.000506)
Observations 5,480 5,477 5,470 4,727 5,480
R-squared 0.088 0.056 0.123 0.055 0.110
Country FE YES YES YES YES YES
Year FE YES YES YES YES YES
DEV Countries NO NO NO NO NO
EM Countries YES YES YES YES YES
Note: See footnote of Table 4 & 5. In neither of the 5 equations is the FBNUM depth AM coefficients significantly different from the
FBNUM depth BM coefficient.
Table 7: The effect of foreign bank entry on the performance of domestic banks, splitting foreign banks based on home country regulatory quality
(1) (2) (3) (4) (5)
∆PROF ∆NMARGIN ∆NINTINC ∆LLPROV ∆OVERHEAD
∆FBNUMRQ AM 0.0906** -0.0103 0.0617** -0.0315 0.0571 (0.0376) (0.0189) (0.0279) (0.0219) (0.0680) ∆FBNUMRQ BM 0.0536 0.0144 0.0334 -0.0209 0.0174 (0.0416) (0.0195) (0.0271) (0.0270) (0.0697) ∆OVERHEAD -0.0695* 0.0264*** -0.0873* 0.0583 (0.0410) (0.00776) (0.0476) (0.0643) ∆CSTFUN -0.00711 -0.0278*** 0.00353 0.00875 -0.195* (0.0291) (0.00920) (0.0137) (0.0192) (0.106) ∆EQUITY 0.301*** 0.0867*** 0.0258 -0.142*** -0.168 (0.0727) (0.0181) (0.0161) (0.0486) (0.119) ∆NINTASS -0.0330 -0.000981 0.00898 0.00251 0.264** (0.0204) (0.0110) (0.0142) (0.0156) (0.107) ∆GDPGrowth 0.00119* -0.000513** 0.000303 -0.00160*** -0.00282* (0.000643) (0.000236) (0.000401) (0.000575) (0.00151)
∆GDPPerCapita 4.78e-06 7.96e-06*** -3.55e-06 3.75e-07 7.05e-06
(6.98e-06) (2.93e-06) (3.96e-06) (4.64e-06) (1.88e-05)
∆Inflation -2.12e-05 3.19e-05** 0.000211*** 0.000149** 0.000372
(4.87e-05) (1.61e-05) (4.47e-05) (6.30e-05) (0.000229)
Observations 8,521 8,510 8,480 7,536 8,521
R-squared 0.190 0.118 0.075 0.145 0.096
Country FE YES YES YES YES YES
Year FE YES YES YES YES YES
DEV Countries NO NO NO NO NO
EM Countries YES YES YES YES YES
Note: See footnote of Table 4 & 5. In neither of the 5 equations is the FBNUM RQ AM coefficients significantly different from the FBNUM
23
Turning to Table A.5, foreign banks are split based on regulatory quality and domestic banks on their profitability. The non-interest income of the most profitable domestic banks are still unaffected, while the least profitable banks are affected by both types of foreign banks. This suggests that the least profitable banks can learn from foreign banks with both a below- and above average home country regulatory quality. The effects are not significant within and across the samples. The only difference between Table A.5 and A.6, lies in the fact that foreign banks are now grouped based on the home country creditor information in the latter. Again, the coefficients for the non-interest income are all insignificant, suggesting that the home country depth of creditor information does not matter.
Summarizing the effects of foreign bank entry on the non-interest income, it turns out that only emerging markets are affected. Moreover, only the least profitable banks are affected. When looking at the home country characteristics of banks, only regulatory quality seems to play a significant role. All of this is in line with the effects on profits before taxes. Therefore, I interpret these results as follows. Foreign bank entry positively impacts the non-interest income, mainly due to spillovers. Because the net income non-interest is not negatively affected, it is unlikely that foreign banks push the domestic banks out of the loan markets. This, in turn, increases the overall results of the least profitable domestic banks: the profits before taxes.
Looking at the table for the net interest income, the results are only significant when looking at Table A.6. Both banks with an above- and below average home country creditor information have affect the net interest income of the least profitable domestic banks positively. This suggests that the depth of creditor information play a role when considering the impact on the net interest income. However, all other variables are still insignificant. There is thus no conclusive evidence that the net interest margin is affected by foreign bank entry. Foreign banks do not seem to gain borrowers at the expense of the incumbent domestic banks. This suggests that the domestic banks hold knowledge superior to the foreign banks, allowing them to remain competitive.
24
with the fact that the home country creditor information does not matter. This suggests that there are no spillovers in terms of screening technologies.
The last measure of efficiency, the overhead costs, remains insignificant throughout all regressions as well. Domestic banks do not increase their costs to implement new practices and techniques, brought into the market by foreign banks. Lensink and Hermes (2004) found that banks actually increase their costs to adopt the new methods. To cover the increased costs, they raised their interest margins and non-interest income. While the non-interest income also increase in this analysis, the net margins remain unchanged in general. This shows that there is no sign that banks increase their overhead costs in order to adapt new practices and techniques.
In summary, the results indicate that foreign bank entry affects the profits of domestic banks positively via an increased non-interest income. The other variables, namely net margin, loan loss provision and overhead costs are not affected by foreign bank entry. While the regulatory quality of the home country matters, the effect of foreign bank entry does not depend on the home country creditor information.
4.3 Comparison with previous studies
I will now compare the results of this research to the results of Claessens et al. (2001) and Lensink & Hermes (2004), who both used the same baseline empirical model. Claessens et al. (2001) found a significant negative effect of foreign bank entry on the profits before taxes to total assets. The results for the net margin and loan loss provision are insignificant and the coefficients for the non-interest income and overhead costs are significant only at the 10% level. The effects seem to be negative, while the results presented here all are generally positive. One explanation for this might be that the authors do not distinguish between the host country level of development. Moreover, the time period (1988-1995) is different from the time period used here.
25
to my results. Firstly, a threshold method is used. This requires balanced panel data reducing the data available significantly. To apply this method, they keep the three most recent observations of all banks, leaving only 990 banks with 3 observations for each bank across 48 countries. Secondly, the time period 1990-1996 is, again, different from the time period used in this research.
The reason that I stipulate the time period used by the two studies mentioned above is the following. During the recent years, more data has become available. Even in the dataset used in this study, the availability of data is higher in the later years. This suggests that for the time periods used by the other studies, less data for banks was available. Moreover, it is not unlikely that the earlier periods data was available for banks with specific characteristics only. For instance, in the past, Bankscope might have gathered data only of the largest banks in each country. This might result in some form of bias in the sample, which is also present in the early years of my dataset. As a result, the outcomes between the two studies and my study are different from one another. Moreover, as indicated by Table 1 and 3, the landscape has changed as well. Nowadays, banks are becoming more international and originate more often from developing countries.
4.4 Robustness checks
Up until this point, the analysis assumed that the foreign entry in year t and country j affected the domestic bank performance of bank i in year t and country j. However, it is possible that spillovers will materialize only after a certain period of time. For the profitability measure, this is analyzed in Table 8. In column 1 the basic regression results are shown, which correspond to column 4 in Table 4. Column 2 introduces a lagged term of the foreign bank entry measure and in column 4 the regression results are shown when incorporate both terms. It seems to be the case that the previous year foreign bank entry does not explain the current year variation in profits before taxes to total assets.
26
Table 8: The effect of foreign bank entry on the performance of domestic banks in emerging countries, introducing a lag and lead term
(1) (2) (3) (4) (5) (6) (7) (8) (9)
∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆NMARGIN ∆NMARGIN ∆NMARGIN
∆FBNUM 0.0788** 0.0703** 0.0565* 0.0481 -0.00349 -0.00279 (0.0344) (0.0348) (0.0332) (0.0337) (0.0164) (0.0173) ∆FBNUM t-1 0.0216 0.0105 -0.00481 (0.0407) (0.0410) (0.0418) ∆FBNUMt+1 -0.0577* -0.0659* -0.0670** -0.0264* -0.0260 (0.0328) (0.0343) (0.0333) (0.0154) (0.0158) ∆OVERHEAD -0.0677* -0.0680 -0.0311 -0.0682 -0.0313 -0.0295 0.0262*** 0.0286*** 0.0286*** (0.0388) (0.0417) (0.0204) (0.0417) (0.0204) (0.0228) (0.00726) (0.00819) (0.00819) ∆CSTFUN -0.00857 -0.00127 -0.0127 -0.00111 -0.0126 -0.00465 -0.0338*** -0.0393*** -0.0393*** (0.0288) (0.0305) (0.0277) (0.0305) (0.0277) (0.0292) (0.0113) (0.0123) (0.0123) ∆EQUITY 0.311*** 0.360*** 0.311*** 0.360*** 0.311*** 0.367*** 0.0906*** 0.0915*** 0.0915*** (0.0715) (0.0652) (0.0772) (0.0652) (0.0772) (0.0703) (0.0181) (0.0194) (0.0194) ∆NINTASS -0.0267 -0.0278 -0.0220 -0.0273 -0.0214 -0.0231 -0.00160 -0.00244 -0.00247 (0.0208) (0.0220) (0.0202) (0.0219) (0.0202) (0.0212) (0.0112) (0.0120) (0.0120) ∆GDPGrowth 0.00126** 0.00121* 0.00163*** 0.00117* 0.00160*** 0.00150*** -0.000515** -0.000516** -0.000515** (0.000615) (0.000625) (0.000585) (0.000621) (0.000580) (0.000579) (0.000230) (0.000236) (0.000235) ∆GDPPerCapita 4.08e-06 6.60e-06 1.92e-06 6.48e-06 1.84e-06 4.24e-06 8.63e-06*** 8.83e-06*** 8.84e-06***
(6.84e-06) (7.07e-06) (7.14e-06) (6.96e-06) (7.04e-06) (7.14e-06) (2.87e-06) (3.02e-06) (3.02e-06) ∆Inflation -4.98e-06 -1.77e-05 -3.94e-05 -1.56e-05 -3.83e-05 -4.63e-05 4.93e-05** 4.70e-05** 4.69e-05** (4.96e-05) (4.90e-05) (3.60e-05) (4.95e-05) (3.59e-05) (3.72e-05) (2.12e-05) (2.12e-05) (2.12e-05)
Observations 8,919 8,471 8,016 8,471 8,016 7,592 8,908 8,005 8,005
R-squared 0.196 0.215 0.204 0.216 0.204 0.228 0.124 0.130 0.130
Country FE YES YES YES YES YES YES YES YES YES
Year FE YES YES YES YES YES YES YES YES YES
DEV Countries NO NO NO NO NO NO NO NO NO
EM Countries YES YES YES YES YES YES YES YES YES
Note: See footnote of Table 4 & 5.
Table 9: The effect of foreign bank entry on the performance of domestic banks in emerging countries, introducing the control variables
(1) (2) (3) (4) (5) (6) (7) (8) (9)
∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF ∆PROF
∆FBNUM 0.132*** 0.131*** 0.117*** 0.113*** 0.134*** 0.0943*** 0.125*** 0.117*** 0.0788** (0.0428) (0.0422) (0.0425) (0.0425) (0.0416) (0.0362) (0.0403) (0.0403) (0.0344) ∆OVERHEAD -0.0877** -0.0677* (0.0372) (0.0388) ∆CSTFUN -0.183*** -0.00857 (0.0402) (0.0288) ∆EQUITY 0.301*** 0.311*** (0.0596) (0.0715) ∆NINTASS -0.0496** -0.0267 (0.0225) (0.0208) ∆GDPGrowth 0.00160** 0.00126** (0.000661) (0.000615)
∆GDPPerCapita 1.52e-05** 4.08e-06
(6.59e-06) (6.84e-06)
∆Inflation -0.000111** -4.98e-06
(5.27e-05) (4.96e-05)
Observations 9,482 9,436 9,344 9,284 9,148 9,444 9,482 9,444 8,919
R-squared 0.015 0.034 0.074 0.159 0.017 0.023 0.017 0.018 0.196
Country-Year FE YES YES YES YES YES YES YES YES YES
DEV Countries NO NO NO NO NO NO NO NO NO
EM Countries YES YES YES YES YES YES YES YES YES
27
As discussed in the previous section, the control variables in the model are used to control for (i) the country-year specific effects not captured by the fixed effects and (ii) to control for changes in the domestic bank variables which cannot be explained by foreign bank entry. Table 9 introduces all the control variables in a step-wise process. The dependent variable used is the profitability in emerging countries.
The results show that the foreign presence variable remains significant through all regressions, although its effect is slightly smaller when all variables are added to the regression at the same time. Moreover, all control variables are significant when they are included on their own, indicating that all variables control for bank- and country specific characteristics, reducing the possible bias when estimating the effect of foreign bank entry.
Another note on the control variables is that they may change sign along different dependent variables and different sub-samples. First, the signs may differ among the different dependent variables because its impact on each variable might be different. For instance, higher overhead costs are associated with reduced profits and increased loan loss provisions (see Table 5). Higher overhead costs come at the expense of profits, whereas they may also be associated with better screening technologies and thus lower loan loss provisions. Second, different banks might react differently to changes in the control variables. For instance, the least profitable banks might be adversely affected by an increase in inflation while the most profitable banks might see a positive effect (see Table 4, columns 7 and 8). This may be caused by the fact that the least profitable banks are by definition different and thus affected differently by inflation, relative to the most profitable banks. Third, there are no signs of multicollinearity among the control variables.
5 Conclusion
28
Consistent with Lensink & Hermes (2004), the results indicate that the host country does affect the impact of foreign bank entry. However, the evidence presented in the previous sections indicates that the impact is not significant on the banking sector of developing countries. Nevertheless, the impact on emerging countries is significant. This indicates that foreign banks do not have any significant influence in developing markets.
Moreover, this study only finds a significant impact on the profits and non-interest income of the least profitable domestic banks in emerging countries. No evidence is found for a change in the net margins, loan loss provisions or overhead costs, as opposed to what previous studies found. However, this might be explained by the fact that a different time period is analyzed alongside a larger sample of domestic banks. These results indicate that foreign bank entry stimulates non-interest income activities among the least profitable banks and consequently improve their profits before taxes.
Contrary to what was expected, the home country creditor information does not determine the impact of foreign banks. Moreover, it does not explain any change in the loan loss provision of domestic banks. However, the home country regulatory quality does determine the impact of foreign bank entry. Especially banks coming from countries with an above average regulatory quality seem to have a positive impact on the non-interest income, resulting in an increase in the profits before taxes as well.
In line with previous analysis, the ratio of foreign banks to total banks is superior in explaining any change in domestic banking markets, compared the ratio of foreign assets to total assets. This is most likely due to the fact that an increase in foreign assets does not mean that a foreign entry occurred per se, as it could be an expansion of an already active foreign bank. Therefore, differentiating between foreign entry and foreign expansion, as measured in assets, is a topic for future research.
The analysis in this paper could be further extended in the following ways. First, by distinguishing between foreign entry between a takeover of an existing bank or a greenfield investment. In general, a takeover will increase the amount of active banks in the market and thus the competitive pressure. It is not unlikely that this increase in competition will have different effects compared to a takeover of an existing bank.
