Expected is that the negative relationship between foreign bank presence and private credit is conditional on a dimension of the liability of foreignness that foreign banks face: information asymmetry

(1)

Foreign Bank Presence and Financial Development: The Role of Asymmetric Information

University of Groningen, Faculty of Economics and Business

Master Thesis International Economics and Business Master Thesis Finance

Name Student: Dennis Veenstra Student ID: s2214555

Student Email: D.Veenstra.5@student.rug.nl Date Thesis: 13 June, 2016

Name Supervisor IE&B: Anna Samarina, PhD Name Supervisor Finance: prof. dr. Theo Dijkstra

(2)

2 Abstract

In this paper, I analyze the relationship between foreign bank presence and financial development, as measured by credit to the private sector. Expected is that the negative

relationship between foreign bank presence and private credit is conditional on a dimension of the liability of foreignness that foreign banks face: information asymmetry.

This paper analyzes empirically whether and how the relationship between foreign bank presence and financial development is influenced by this dimension of liability of foreignness.

Using a sample of 117 countries, a time period of 10 years (2004-2013) and a fixed effects panel setting, an overall positive impact is found for different measures of foreign bank presence and asymmetric information. This indicates that a higher degree of information asymmetry has a positive influence on the relationship between foreign bank presence and financial development, contrary to the liability of foreignness theory.

Field Keywords: Financial Development, Private Credit, Foreign Bank Presence, Liability of Foreignness, Asymmetric Information.

(3)

3 Content

I. Introduction ... 4

II. Literature Review ... 6

II.I. Financial Development ... 7

II.II. Foreign Bank Presence and Financial Development ... 7

II.III. Foreign Bank Presence, Financial Development and the Liability of Foreignness ... 9

II.IV. The Asymmetric Information Channel as a Dimension of Liability of Foreignness ... 10

II.V. Asymmetric Information and Private Credit ... 10

II.VI. Theoretical Model ... 11

III. Methodology ... 11

IV. Data ... 14

V. Analysis ... 17

V.I. Descriptive Statistics ... 17

V.II. Correlation Analysis ... 18

V.III. Trends ... 18

V.IV. Scatter Analysis ... 18

VI. Empirical Results ... 19

VI.I. Models 1 and 2, The Relationship between Foreign Bank Presence and Financial Development ... 19

VI.II. Models 3-6: Asymmetric Information as a Conditional Factor in the Relationship between Foreign Bank Presence and Financial Development ... 19

VII. Robustness ... 23

VII.I. Sample ... 23

VII.II. Time Period ... 29

VII.III. Control Variables ... 31

VII.IV. Estimation Technique ... 34

VIII. Discussion of Results ... 36

VIII.I. Results ... 36

VIII.II. Limitations and Further Research ... 38

IX. Conclusion ... 40

References ... 41

Appendices ... 45

(4)

4 I. Introduction

21^st century globalization made the world ‘flat’ (Friedman, 2005). According to Friedman’s top selling book worldwide “The World is Flat” (Friedman, 2005), this metaphor indicates that distances do not play a role anymore in todays globalized environment. Although this thesis is heavily criticized by various economists, they do all agree that technological and institutional changes reduced spatial transmission costs (Leamer, 2007; Gray & Friedman, 2005; McCann, 2008). With lower costs of moving goods and information, distance became less of an obstacle for conducting business. As a result, economic entities (firms, institutions etc.) seized the opportunity to expand abroad and increased their international trade flows by more than 116 percent between 1995 and 2005¹. All of a sudden, formerly remote areas were connected in multiple ways.

Especially the enhanced worldwide interconnectedness of financial institutions is of particular interest to economists and finance practitioners, as they concern not only local, but also global financial stability and fragility. The recent catastrophe that reminded us of the downside of increased globalization was the global financial crisis that began in 2008; what began as a subprime mortgage crisis in the US only, quickly became a crisis that resulted in severe liquidity problems affecting nearly the entire globe. Crises like this encourage discussions on the topic of international banking to answer question like: what are the consequences for financial sector development, stability and fragility? Or, should it be more regulated, controlled or restricted?

As could be concluded from the financial crisis, the increased interconnectedness of financial institutions makes it more likely that local financial shocks will be transmitted across borders.

Popov & Udell (2010) provide empirical prove for this thesis, by showing that distress at the parent bank led to less bank lending by foreign subsidiaries in the years before the financial crisis. In order to prevent local shocks for becoming regional or global ones, it is important that the international activities of banks are not detrimental to the development and stability of the financial system. This paper will investigate the international activities of banks, with a particular focus on its consequences for financial system development.

1 Based on own calculations from United Nations Comtrade Database, http://comtrade.un.org/data/.

(5)

5 Even though many studies exist that address the issues of international banking, it is hard to generalize the results obtained due to data availability issues². In an attempt to overcome these data availability issues, this research uses the relatively new Foreign Bank Ownership

Database constructed by Claessens & Horen (2015). By including nearly all banks in a

country and their ownership status over the years 1995-2009, they try to generalize the impact of foreign bank ownership on private credit. Claessens & Horen (2014) conclude that

heterogeneity of banks and countries is crucial for understanding the relationship between foreign bank presence and private credit. This indicates that bank- and host-country

characteristics influence this relationship significantly. Since private credit is closely related to financial development in existing literature (De Gregorio & Guidotti, 1995; King & Levine, 1993), private credit is used to investigate the relationship between foreign bank presence and financial development. Therefore, the terms ‘private credit’ and ‘credit’ refer to financial development throughout the rest of this paper. Using an updated version of the ‘Foreign Bank Ownership Database’ (Claessens & Horen, 2015) that covers the years 1995-2013, the role of a channel that might influence the relationship between foreign bank presence and private credit will be investigated as well, namely; the liability of foreignness.

The concept of liability of foreignness as a cost of doing business abroad was introduced by Hymer (1976), who claimed that foreign firms face disadvantages relative to domestic firms when entering a market. Whereas many research has been conducted on the cultural distance, institutional distance and efficiency dimensions of the liability of foreignness (Calhoun, 2002;

Djankov, McLiesh & Schleifer, 2007; Eden & Miller, 2004; Miller & Parkhe, 2002), the information asymmetry dimension seems to be missing in the foreign banking and private credit literature. This is why this research is focused on this subject in particular.

Asymmetric information between lender and borrower is one of the prime reasons for the existence of banks (Bhattacharya & Thakor, 1993). While individuals usually do not have the ability to evaluate and monitor customers, banks do. Therefore, they are better able to provide credit in imperfect market conditions such as information asymmetry. Usually, local banks have acquired longstanding relationships with local customers, facilitating the credit even further (Boot, 2000). Longstanding relationships result in the collection of soft information about these customers; information that is collected over time through many personal

2 In previous studies, conclusions are mostly drawn based on a small sample of countries and/or banks. In other words, there exists no database comprising the entire banking system. Due to this small sampling, it is not possible to account for heterogeneity between countries and/or banks.

(6)

6 meetings (Petersen, 2004). However, foreign banks have not established these relationships before entering the market. Coming from a different country, they have informational disadvantages compared to locally operating banks because they need to rely only on hard information; information that is codified (Petersen, 2004). In other words, foreign banks face a liability of foreignness since hard information is not always available. In order to address this dimension of liability of foreignness, this research will focus on the following question:

What is the impact of the degree of asymmetric information on the relationship between foreign bank presence and private credit in the domestic market?

According to the arguments above, it is hypothesized that the higher the degree of asymmetric information in a country, the harder it will be for foreign banks to provide for credit in the domestic market. In this research, the asymmetry of information will be measured by the loan spreads and the availability of creditor information. This paper will therefore contribute to the literature by explaining how foreign banks affect domestic banking markets in terms of private credit. Using a sample of 117 countries, a time period of 10 years (2004-2013) and a fixed effects panel setting, we find that overall the degree of asymmetric information

positively influences the relationship between foreign bank presence and private credit.

The rest of this paper is structured as follows: Section II reviews existing literature regarding the topic and provides the main hypotheses. Sections III, IV and V outline the methodology, data specifications and preliminary analyses. Section VI describes the results, followed by robustness checks in section VII and a thorough discussion in section VIII. Section IX concludes.

II. Literature Review

This research analyzes whether the effect of foreign bank ownership on domestic financial development, as measured by the supply of private credit, differs by the degree of information asymmetry. According to international economics theory, information asymmetry implies a liability of foreignness for foreign owned banks in the form of an unfamiliarity hazard (Zaheer, 2002; Eden & Miller, 2004). These liabilities should negatively influence their private credit supply compared to domestically owned banks. Where foreign banks have only hard information (measurable performance indicators) available about customers before entering the market, domestic banks benefit from soft information as well (Petersen, 2004).

Soft information is gathered by domestic banks over many years through intensive

(7)

7 relationships with their customers (Boot, 2000). Therefore, the presence of foreign banks is expected to hamper the overall financial system development, as measured by credit to the private sector within the host country, via the asymmetric information channel.

In order to structure the views of existing literature regarding this topic, the literature review starts with the concept of financial development and its importance. Subsequently, the effect of foreign bank presence on financial development will be discussed. Finally, this effect will be related to the liability of foreignness that foreign banks face, with the information

asymmetry channel acting as an important dimension of the liability of foreignness.

II.I. Financial Development

Financial development is actually the development of the financial system. The financial system comprises financial instruments, markets and institutions, and its purpose can be given by the following quote: “Financial systems serve one primary function: they facilitate the allocation of resources, across space and time, in an uncertain environment” (Merton & Bodie 1995, p. 12).

Financial development is commonly measured as domestic credit to the private sector;

therefore, (private) credit provision is used in the remainder of this paper to reflect financial development³. Its importance is demonstrated by the extensive literature, in particular with regard to its relationship with economic growth⁴, which illustrates the importance of private credit in other key economic variables.

II.II. Foreign Bank Presence and Financial Development

Many researchers have found a theoretical link between foreign bank presence and financial development. On the positive side of the spectrum, some argued that foreign banks improve domestic banking markets by increasing the availability and access to credit (Levine, 1996;

Clarke, Cull, Peria & Sanchez, 2003). This increased availability and access is believed to come from enhanced competition and increased lending to small customers.

3 The measurement of financial development is discussed more thoroughly in the data section of this paper.

4 Where some researchers find a positive relationship between the two (Bagehot, 1873; Hicks, 1969;

Schumpeter, Becker & Knudsen, 2002), others conclude that financial development is endogenous on economic growth (Robinson, 1952; Goldsmith, 1969). In other words, these results are contradictory with respect to the causality of the relationship. Other researchers concede that there is no relationship between finance and economic growth at all (Lucas, 1988; Stern, 1989). Levine (1997) argues instead that financial markets and institutions do play a crucial role in economic growth. By mitigating information and transaction costs, the financial system affects items that are crucial to economic growth. In fact, financial development is not only seen as a key player for current economic performance; it also turns out to be a serious indicator of future growth performance.

(8)

8 Enhanced competition by foreign banks forces incumbent banks to become more cost-

efficient, since foreign banks are usually larger and superior performing banks (Focarelli &

Pozzolo, 2000). Economic theory suggests that domestic banks that do not become more efficient are forced to exit the market because they cannot compete with these larger, better performing banks. This may result in more credit provided at lower rates in the host country.

Clarke, Cull & Peria (2001) find evidence for this by showing that borrowers have easier access to credit at lower rates due to the effect of foreign bank presence on domestic banking competition.

Foreign bank presence tends to improve lending to small customers because of improvements in hard information and the shift of focus of domestic banks to niche markets. Improvements in hard information such as credit scores enables the loans to small businesses to be part of an automated process (Oppenheim, 1997). Without credit scores, loans to small businesses were usually provided by small neighborhood lenders that knew the businesses well. Due to the increased availability of credit scores, large lenders become interested in smaller businesses as well, increasing the overall lending to the private sector (Mester, 1997). Access to credit is also believed to improve because foreign banks force domestic banks to act in niche markets, which improves small customer lending via a different channel. Bonin & Abel (2000) find evidence for this last argument in Hungary, whereas Jenkins (2000) finds evidence for this for a wider set of countries.

However, this theoretical positive link between foreign bank presence and private credit is under scrutiny. Other researchers describe the phenomenon of foreign bank lending as “cherry picking”, or “cream skimming” which is actually seen as a cost of foreign bank presence (Stiglitz, 1993; Detragiache, Tressel & Gupta, 2008; Clarke, Cull, Peria & Sanchez, 2003). In addition, foreign banks are expected to decrease their cross-border lending when the parent bank gets hit by a shock (Cetorelli & Goldberg, 2012).

Foreign banks focus primarily on the largest, most transparent and best customers, since they have informational disadvantages relative to domestic banks (Stiglitz, 1993; Detragiache, Tressel & Gupta, 2008). In other words, foreign banks tend to “cherry pick” the best customers out of the market. Leaving only the moderate or bad customers to be served by

(9)

9 domestically operating banks, this negatively influences overall private credit within the country when domestic banks exit the market (Clarke, Cull, Peria & Sanchez, 2003)⁵.

In times of crisis, credit may be haltered even further. Cetorelli & Goldberg (2012) expect that funds will flow from foreign affiliates to the parent when it gets hit by a shock. This indicates that foreign affiliates will provide less credit in the host-country when the parent bank is in financial distress. Cetorelli & Goldberg (2012) find empirical evidence for this for US banks in the financial crisis.

The evidence on foreign bank presence and private credit gives mixed results. Using a relatively new and complete foreign bank ownership database, Claessens & Horen (2014) have tried to generalize the empirical relationship between foreign bank presence and private credit, and find an overall negative relationship. In line with Claessens & Horen (2014), our first hypothesis is as follows:

H1: Foreign bank presence has a negative effect on private credit within the host-country II.III. Foreign Bank Presence, Financial Development and the Liability of Foreignness Claessens & Horen (2014) notice that the negative effect of foreign bank presence on private credit depends on home- and host-country characteristics that are likely to be attributed to the liabilities of foreignness that foreign banks face. For example, when the home country of the foreign bank is very distant to the host country, the negative effect of foreign bank presence on credit is more pronounced. In addition, this negative effect is more existent in developing countries, countries where contract enforcement is poor and the availability of creditor information is below standard.

Hymer (1976) mentions foreigner discrimination, exchange rate risk and differences in

language, law and politics as additional liabilities of foreignness that banks could face. Zaheer

& Mosakowski (1997) mention that foreign banks must have some sort of competitive advantage (compared to locally operating banks) to outweigh the costs arising from their liabilities of foreignness. The results of the research of Claessens & Horen (2014) indicate that even though foreign banks are likely to be more productive than local banks (Focarelli &

Pozzolo, 2000; Helpman , Melitz & Yeaple, 2003), their liabilities of foreignness makes them

5 The best customers are not only the best in terms of performance. In addition, best performers are seen as most transparent and able to monitor. These customers have hard information (measurable performance indicators, such as ratings or performance indicators) available (Petersen, 2004).

(10)

10 unable to translate this productivity advantage into an improvement of the domestic financial system as measured by private credit.

II.IV. The Asymmetric Information Channel as a Dimension of Liability of Foreignness This paper is going to enhance the knowledge of the benefits and risks of foreign bank presence by investigating whether and how the effect of foreign bank presence on financial development depends on a channel of the liability of foreignness; namely, asymmetric information.

Information asymmetry is seen as a barrier to perfect markets⁶. When markets are perfect there is no use for financial intermediaries, since borrowers and lenders only interact through the market to fulfill their funding needs. Hence, the reason for, and the idea of financial intermediation comes from the fact that markets are never perfect. In order to overcome this barrier, financial intermediaries (mostly banks) have resources and abilities that individuals do not have: they have many customers and therefore have the ability to share evaluation costs and monitoring effort (Gurley, Gurley & Shaw, 1960; Diamond, 1984). More specifically, Bhattacharya & Thakor (1993) mention that banks have an advantage over individuals when it comes to information gathering: they develop unique interpretation skills related to signals, and they can re-use information across customers.

Information asymmetry is seen as a dimension of liability of foreignness because foreigners do usually not possess that much information about the environment as natives. Due to the fact that native banks know where to look for information and usually have longstanding relationships with their customers, they have informational advantages compared to foreign banks (Caves, 1971; Boot, 2000). According to Boot (2000), relationship lending improves information production and sharing, and enables the use of tailored contracts. In other words, foreign banks that do not engage in relationship banking acquire less information from customers than local banks do.

II.V. Asymmetric Information and Private Credit

While many research has been conducted on the cultural distance, institutional distance and efficiency dimensions of liability of foreignness (Djankov, McLiesh & Schleifer, 2007;

Calhoun, 2002; Eden & Miller, 2004; Miller & Parkhe, 2002), little work has been done on

6 Asymmetric information is a market imperfection that a financial intermediary faces due to the fact that the intermediary (as a lender of funds) has less information (credibility, incentives) about the borrower than the borrower itself. This could lead to adverse selection and moral hazard problems (see Akerlof, 1995)

(11)

11 the asymmetric information dimension. Therefore, we now elaborate on how the information asymmetry market imperfection affects credit in the market and how it is considered a conditional factor in the relationship between foreign bank presence and private credit.

Where extensive literature has shown that asymmetric information affects credit provision negatively, there is some disagreement in terms of the channel trough which this occurs.

Some researchers conclude that asymmetric information results in higher risk premiums on loans (Ivashina, 2009; Mazumdar & Sengupta, 2005), while others find that asymmetric information leads to credit rationing, rather than increased interest rates (Jaffee & Russell, 1976; Stiglitz & Weiss, 1981; Carpenter & Petersen, 2002).

The risk premium on loans is actually an extra price charged on these loans. If, due to

information asymmetry, the prices of the loans increase, less loans will be demanded. In other words, an increased risk premium results in lower credit due to the channel of supply and demand. Credit rationing however, implies an excess demand for loans for a given optimal price. This means that when a bank is unable to distinguish the good from the bad borrowers, it is unable to match the supply and demand for loans (Stiglitz & Weiss, 1981). Even though the channels are different, asymmetric information appears to result in an equilibrium with lower credit in both cases.

II.VI. Theoretical Model

Following the reasoning above, it is expected that foreign banks have a more detrimental impact on domestic credit when the market is more imperfect in terms of information asymmetry. The underlying reason is that foreign banks are supposed to be negatively affected by liabilities of foreignness, with asymmetric information acting as a channel of this liability. As a result, these banks have less information about borrowers in foreign markets than locally operating banks do, leading to higher loan spreads and/or credit rationing. This, in turn leads to less credit provided to the private sector. Therefore, our second hypothesis states:

H2: A high degree of asymmetric information in a country will negatively influence the relationship between foreign bank presence and private credit within the host-country.

III. Methodology

In order to test the two hypotheses, the following models have been formulated. Model (1) tests for hypothesis 1 and model (2) tests for hypothesis 2, respectively:

(12)

12 𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝑜𝑟𝑒𝑖𝑔𝑛 𝐵𝑎𝑛𝑘 𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒_𝑖𝑡−1+ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙_𝑖𝑡−1+ µ_𝑖+ 𝜀_𝑖𝑡 (1) 𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝑜𝑟𝑒𝑖𝑔𝑛 𝐵𝑎𝑛𝑘 𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒_𝑖𝑡−1+

𝛽₂𝐴𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛_𝑖𝑡−1+ 𝛽₃𝐹𝑜𝑟𝑒𝑖𝑔𝑛 𝐵𝑎𝑛𝑘 𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒_𝑖𝑡−1∗

𝐴𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛_𝑖𝑡−1+ 𝐶𝑜𝑛𝑡𝑟𝑜𝑙_𝑖𝑡−1+ µ_𝑖+ 𝜀_𝑖𝑡 (2) Within these models, the dependent variable, FinDev, represents the financial development in country i and year t and is measured as domestic credit provided to the private sector as a percentage of GDP.

The first independent variable Foreign Bank Presence is the proxy for foreign bank presence in a country. The variable is measured is two ways: FBNUM represents the absolute presence of foreign banks, by dividing the number of foreign banks by the total amount of banks in country i and year t-1. FBSHR looks at the relative presence of foreign banks, by dividing foreign bank assets by total bank assets in country i and year t-1.

The second independent variable Asymmetric Information proxies the degree of information asymmetry in a country and is also measured in two ways: Premium stipulates the domestic risk premium on lending in country i and year t-1, and CreditInfo stands for the depth of creditor information score in country i and year t-1.

Control represents the various control variables included in the model, µ includes the unobserved country specific effects and ε is the error term. In the rest of this paper, we are referring to these terms combined when mentioning the error term.

In this analysis, Foreign Bank Presence and Asymmetric Information are used in interaction variables in order to determine whether the effect of foreign bank presence on financial

development differs by the degree of asymmetric information. In total, the following 4 models will be run:

𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝐵𝑁𝑈𝑀_𝑖𝑡−1+ 𝛽₂𝑃𝑟𝑒𝑚𝑖𝑢𝑚_𝑖𝑡−1+ 𝛽₃𝐹𝐵𝑁𝑈𝑀_𝑖𝑡−1∗ 𝑃𝑟𝑒𝑚𝑖𝑢𝑚_𝑖𝑡−1+

𝐶𝑜𝑛𝑡𝑟𝑜𝑙 + µ_𝑖+ 𝜀_𝑖𝑡 (3)

𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝐵𝑁𝑈𝑀_𝑖𝑡−1+ 𝛽₂𝐶𝑟𝑒𝑑𝑖𝑡𝑖𝑛𝑓𝑜_𝑖𝑡−1+ 𝛽₃𝐹𝐵𝑁𝑈𝑀_𝑖𝑡−1∗ 𝐶𝑟𝑒𝑑𝑖𝑡𝑖𝑛𝑓𝑜_𝑖𝑡−1+

𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝐵𝑆𝐻𝑅_𝑖𝑡−1+ 𝛽₂𝑃𝑟𝑒𝑚𝑖𝑢𝑚_𝑖𝑡−1+ 𝛽₃𝐹𝐵𝑆𝐻𝑅_𝑖𝑡−1∗ 𝑃𝑟𝑒𝑚𝑖𝑢𝑚_𝑖𝑡−1+

(13)

13 𝐹𝑖𝑛𝐷𝑒𝑣_𝑖𝑡 = 𝛽₀+ 𝛽₁𝐹𝐵𝑆𝐻𝑅_𝑖𝑡−1+ 𝛽₂𝐶𝑟𝑒𝑑𝑖𝑡𝐼𝑛𝑓𝑜_𝑖𝑡−1+ 𝛽₃𝐹𝐵𝑆𝐻𝑅_𝑖𝑡−1∗ 𝐶𝑟𝑒𝑑𝑖𝑡𝐼𝑛𝑓𝑜_𝑖𝑡−1+

This paper investigates how the relationship between foreign bank presence and financial development is influenced by asymmetric information within host countries over time.

Observing N individuals (countries) over T time periods, a panel setting is required.

Using a panel model requires the validity of various assumptions, in particular for the proper use and interpretation of inferential statistics like the t- and F-statistics. These assumptions are: constant error variance, uncorrelated errors through time and a complete set of regressors, uncorrelated with the residuals. Since heteroscedasticity and autocorrelation result in incorrect standard errors and test statistics, cluster-robust standard errors are used to correct for these potential problems. Cluster-robust standard errors are higher, increasing the confidence intervals in the estimations to allow for correlation of the errors within country-clusters and through time.

Uncorrelated errors with respect to the regressors relates to the type of panel estimation that should be performed: fixed or random effects. Both the fixed- and random effects models are believed to give consistent estimates when there is no correlation between the regressors and the residuals. However, when the unobserved country effects error component µ is correlated with the regressors, the parameters of the random effects model become inconsistent. To allow for correlation between the unobserved country effects µ and the regressors while remaining consistent, the fixed effects model is preferred because this model still provides consistent estimates. The Hausman test is used to formally check for correlation between µ and the regressors. For this analysis, the Hausman test recommends the use of the fixed effects model due to the existence of correlation between the regressors and the unobserved country effects error component µ⁷. The fixed effects model hereby controls for time- invariant characteristics of countries. Therefore, the fixed effects panel model is used in this analysis.

The adjusted R-squared, or the coefficient of determination, can also be used as an unbiased model selection criterion⁸. This means that by selecting the model with the highest adjusted

7 The Hausman test rejects the null hypothesis of equal estimates for the fixed and random effects model (p- value<0.001). This means that due to correlation between the unobserved country effects and the regressors, the random effects model gives inconsistent estimates, while the estimates of the fixed effects model are consistent.

8 The R-squared, or the coefficient of determination, is seen as a measure for the goodness of fit of a model. This measure indicates the amount of variance that is explained by the independent variables. However, due to the fact that the R-squared increases whenever adding additional independents, the measure overstates the goodness

(14)

14 R-squared, the correct model specification is chosen on average (Theil, 1957). However, the chance of choosing the wrong models remains quite high. As an extra check, the adjusted R- squared for the fixed and random effects models have been compared. Since for model 1 and 2, the measure is higher for the fixed effects model than for the random effects model, the fixed effects model is more suitable for this analysis.

However, investigating this relationship using a panel model has a disadvantage compared to OLS estimation in terms of data availability. As this was the case for our independent variable Premium, we introduced an alternative measure of asymmetric information to deal with this problem, as will be further discussed in the data section of this paper.

IV. Data

The final sample consists of 117 countries over 2004-2013 and is constructed as follows. The Foreign Bank Ownership database constructed by Claessens & Horen (2015) lists all banks, including their home- and host-country, that report to Bankscope. In total, 5498 banks are covered in 139 countries. However, in order to calculate reliable foreign bank shares and to align the Foreign Bank Ownership database with Bankscope and World Bank data, a few changes needed to be made.

First of all, Taiwan is excluded from the sample because there are different views on the independency of Taiwan with respect to China⁹. In other words, Taiwan and its banks have been left out entirely. Secondly, the foreign bank shares are based on commercial banks only in an attempt to rule out heterogeneous bank-type effects on financial development in a country¹⁰. Excluding non-commercial banks leaves us with more than 82 percent of the original sample of banks, indicating that the vast majority of banks are commercial. Thirdly, banks without ownership data are left out of the sample. These banks are mostly inactive.

of fit. To account for this, the adjusted R-squared is usually used to determine the goodness of fit for a model.

Whereas the R-squared increases whenever additional variables are added, the adjusted R-squared only increases when these variables improve the model more than would be the case by chance. In fact, the adjusted R-squared can even decrease when adding additional variables.

9 Even though Taiwan is listed as a separate country in the Bank Ownership Database, it is not according to the World Bank. To avoid any involvement in this issue, we excluded Taiwan from the sample. According to the World Bank, data for Taiwan is not included in China so it is safe to say that the analysis will not encounter measurement issues due to this exclusion.

10 It could be possible that the effect of foreign bank presence on domestic credit provision is driven by

differences in bank types. When we keep only commercial banks in the sample (which are the majority of banks) the effect on financial development is solely due to its foreign or domestic status.

(15)

15 By deleting countries for which there is no private credit information, we end up with a final sample of 117 countries over 2004-2013. The full list of countries included in this analysis is shown in table 9 in the appendix, that classifies the countries by income group. Income group classification distinguishes OECD countries, other high income (OHI) countries, emerging markets (EM) and developing countries (DEV) (Claessens & Horen, 2015). OECD only includes the core OECD countries, whereas OHI includes all countries classified as high- income by the World Bank in 2000 but not belonging to the OECD. For simplicity, we merge OHI and OECD countries into a single income group. EM includes all countries that are included in the Standard and Poor's Emerging Market and Frontier Market indexes and that were not high-income in 2000. All other countries are included in DEV. By making this distinction we are able to perform a more thorough analysis and see whether the overall results are driven by countries from specific income groups.

We now turn to the description, construction and sources of the variables that are being used in the models.

FinDev represents the financial development in a country and is measured as domestic credit provided to the private sector as a percentage of GDP. The measure is obtained from the World Bank¹¹. The reason that we use this measure comes from De Gregorio & Guidotti (1995), who described and compared three potential measures of financial development that are frequently used in this line of research: interest rates, monetary aggregates and credit provision. Their findings indicate that real interest rates appear to be good proxies for investment efficiency, but bad proxies for financial development overall¹². Monetary

aggregates, such as M1, M2 or M3, are good proxies for the provision of liquidity, but not for the provision of funds¹³. With these caveats in mind, they use domestic credit provision to the private sector as their measure of financial system development¹⁴. Note that this measure excludes funds channeled to the public sector. The importance of this exclusion is

underscored by Freedman & Click (2006), who claim that the private sector uses funds more productively than the public sector does. Therefore, an improvement in economic

11 World Development Indicators 2016, http://data.worldbank.org/products/wdi.

12 De Gregorio & Guidotti (1995) conclude this based on findings by Dornbusch (1990); since the positive relationship between real interest rates and economic growth is not driven by the effect on the volume of investment, it has to come due to its effect on the efficiency of investment. Because financial development is believed to stimulate the volume of investment, real interest rates are not a good proxy for this measure.

13 The importance of these two roles of the financial system is discussed in Fama (1980).

14 The use of domestic credit provision to the private sector as a fraction of GDP as a proxy of financial development has its disadvantages as well, as will be discussed in the limitations section of this paper.

(16)

16 development is mostly related to increased funding to the private sector, instead of increased funding to the public sector (Levine, Loayza & Beck, 2000). In line with De Gregorio &

Guidotti (1995), we will therefore use credit to the private sector as a fraction of GDP as our proxy for financial development¹⁵.

Foreign bank presence is measured in two ways, in line with Claessens, Demirgüç-Kunt &

Huizinga (2001) and Lensink & Hermes (2004): FBNUM represents the absolute presence of foreign banks and is calculated by dividing the number of foreign banks by the total amount of banks in country i and year t-1. FBSHR looks at the relative presence of foreign banks, by dividing foreign banks assets by total bank assets in country i and year t-1. Hereby, FBSHR incorporates the relative size of foreign banks compared to domestic banks¹⁶. The remainder of this paper uses absolute foreign bank presence to reflect FBNUM and relative foreign bank presence to reflect FBSHR. The variables are obtained by own calculations based on the Foreign Bank Ownership database constructed by Claessens & Horen (2015) and Bankscope data¹⁷. The authors argue that this database on bank ownership is superior to previous

databases frequently used for this purpose. For example, in comparison to a widely used database developed by Micco, Panizza & Yanez (2007) it lists almost all banks that are active in a country, it includes countries of different development levels, it lists bank ownership including changes in ownership over time and it lists the home-country of the foreign bank.

The degree of asymmetric information in a country is also measured in two ways: Premium stipulates the domestic risk premium (over the treasury bill rate) on lending. Its justification comes from Ivashina (2009) and Mazumdar & Sengupta (2005). Ivashina (2009) finds that banks charge higher premiums to compensate for higher asymmetric information and Mazumdar & Sengupta (2005) conclude that firms that provide high quality disclosures benefit from lower interest on their bank loans. The measure is obtained from the World Bank. However, the data on risk premiums is far from complete and lacks comparability: data is available for about 40 percent of the observations and comes with comparability issues because the treasury bill rate and lending terms may be different for countries¹⁸. Therefore, we use CreditInfo as our alternative second independent variable. This variable stands for the depth of creditor information score and is obtained from World Bank’s doing business

15 The measure is scaled by GDP to correct for country size.

16 The importance of including both measures of foreign bank presence in discussed in Claessens, Demirgüç- Kunt & Huizinga (2001) p. 895-897.

17 Bureau van Dijk, Bankscope Database, http://bankscope-bvdinfo-com.

18 The downside of using the risk premium on lending will be further discussed in the limitations section of this paper.

(17)

17 indicators. The index ranges from 0 to 6, and is based on the availability of credit information from credit bureaus or credit registries. A higher score in the index indicates better creditor information. Because asymmetric information is usually investigated on a micro level rather than a macro level, existing literature regarding private credit does not seem to proxy

asymmetric information this way¹⁹. Still, the measure is considered suitable because it reflects the quality, availability and accessibility of creditor information within host-countries.

Control represents the various control variables that are based on frequently used controls in existing literature regarding credit²⁰. GDPperCapita (PPP, constant 2011 international $) and GDPGrowth (annual %) are included based on La Porta et al. (1997). It is expected that more developed economies usually have a higher level of domestic private credit, and economies that witness large growth rates need more credit to fuel this growth. Inflation and

ContractEnforcement come from Djankov, McLiesh & Schleifer (2007): Inflation (GDP deflator, annual %) diminishes the value of debt. Because outstanding loans lose their value more quickly in high inflationary environments, less credit will be provided.

ContractEnforcement, measured as the days necessary to enforce a contract, is included as an institutional control variable. As a doing business indicator, it is assumed that lower values of this variable are related with a better institutional environment and more credit supply.

All independent and control variables that are included in the models are lagged by 1 year, in order to avoid endogeneity problems. For example, the private credit supplied in the host- country could influence foreign bank presence, instead of the other way around. As mentioned in Claessens & Horen (2014) the reasons may stem from a higher expected market growth or bad business circumstances in economies with low levels of private credit. The full

specification and sources of the variables can be found in table 10 in the appendix.

V. Analysis

V.I. Descriptive Statistics

The descriptive statistics for the variables used in the models are shown in table 11 in the appendix. As mentioned earlier, it can be clearly seen that the data on risk premiums covers

19 Asymmetric information is usually investigated in the corporate finance literature, who include the measure in theoretical and empirical models on a micro level. Since this analysis is performed on a macro level, such measures cannot be used. Hence, a new variable to measure asymmetric information on a macro level has been introduced.

20 Djankov, McLiesh & Schleifer (2007) is a frequently cited paper when it comes to variables affecting private credit.

(18)

18 only about 40 percent of the observations. For the depth of creditor information score

however, data is available on almost 80 percent of the observations. Furthermore, it is notable that the mean of absolute and relative foreign bank presence is roughly equal. However, the share of foreign bank presence in terms of assets has a larger standard deviation. This means that in terms of assets, there is much more variance in foreign bank presence compared to the amount of subsidiaries. Based on the descriptive statistics, it is unclear whether there are any outliers in the sample.

V.II. Correlation Analysis

The correlation table for all the independent variables is shown in table 12 in the appendix.

Since all correlations are well below the cutoff points, we conclude that there is no reason for concern in terms of multicollinearity in the models²¹. Even though multicollinearity is always mildly present in regression estimations, the independent variables are not influencing each other too much so that the individual predicted estimates remain valid. That is, the estimation parameters roughly reflect the individual stand-alone effects on the dependent variable, ceteris paribus.

V.III. Trends

Figures 7 and 8 in the appendix show data trends for our key variables of financial

development and foreign bank presence. Figure 7 illustrates this trend for private credit, based on host-country averages per year. Figure 7 suggests that private credit increased over the years, with a slight downturn after 2010. The trend for foreign bank presence, as shown in figure 8, shows an increasing pattern in absolute foreign bank presence with a temporary downturn in 2008 (figure 8A). Relative foreign bank presence is fluctuating over the time period: it first shows a sharp increase, followed by an enormous decline after 2008, after which it increased again (figure 8B). Apparently, the financial crisis that started in 2008 had a tremendous impact on the offshore activities of banks. I elaborate further on this trend in the robustness section of this paper.

V.IV. Scatter Analysis

The last part of the preliminary data analysis contains the scatterplot analysis, which is illustrated in figure 9 in the appendix. The scatterplots relate financial development, as measured by private credit, to foreign bank presence. Figure 9A includes foreign bank

21 In performing statistical analysis, correlations higher than 0.6 give potential reason for multicollinearity concerns. Even though the correlation between absolute and relative foreign bank presence is above this cutoff point, we do not have to be concerned: the variables are used in different regression models.

(19)

19 presence in absolute terms and figure 9B in relative terms. As can be noticed, both figures depict a slightly decreasing negative relationship between the two variables (as indicated by the fitted line). This is in line with the findings by Claessens & Horen (2014).

VI. Empirical Results

VI.I. Models 1 and 2, The Relationship between Foreign Bank Presence and Financial Development

In order to capture the initial effects of foreign bank presence on financial development as measured by private credit, table 1 provides some interesting insights. The table shows the regression estimates for model 1 and 2 respectively for the full sample. So first the

relationship between foreign bank presence and financial development will be described without involving asymmetric information.

In contrast to Claessens & Horen (2014), foreign bank presence does not seem to negatively influence private in this setting. Instead, absolute foreign bank presence implies a significant positive relationship (model 1), whereas the relative foreign bank presence has an

insignificant effect (model 2). The sign for GDP per capita is positive as expected: higher levels of economic development go hand in hand with higher credit supply in the host country. The rest of the controls turn out to be insignificant.

The fact that these results are not in line with Claessens & Horen (2014) gives an indication that this analysis of the relationship between foreign bank presence and private credit is sensitive to the model specifications. For example, Claessens & Horen (2014) use a different sample, time period and estimation technique. The robustness section of this paper looks further into this.

VI.II. Models 3-6: Asymmetric Information as a Conditional Factor in the Relationship between Foreign Bank Presence and Financial Development

Table 2 estimates models 3-6 including the potential conditionality of asymmetric information in the relationship between foreign bank presence and private credit for the full sample.

Models 3 and 4 estimate this relationship using absolute foreign bank presence, while models 5 and 6 use relative foreign bank presence. In addition, models 3 and 5 use the risk premium on lending as the proxy for asymmetric information, while models 4 and 6 use the creditor information score.

(20)

20 Table 1: Financial Development and Foreign Bank Presence

(1) (2)

VARIABLES Full Sample Full Sample

Foreign Bank Presence (absolute) 18.07**

(8.858)

Foreign Bank Presence (relative) -5.384 (3.539)

Ln(GDP per Capita) 26.28*** 31.30***

(6.058) (6.162)

GDP Growth -0.0803 -0.143

(0.120) (0.119)

Inflation -0.0341 -0.0277

(0.0527) (0.0538)

Contract Enforcement -0.0101 -0.00898

(0.0140) (0.0145)

Constant -189.8*** -226.1***

(57.27) (59.60)

Observations 963 946

R-squared 0.098 0.091

Adj. R-squared 0.093 0.086

Number of countries 105 104

Country FE YES YES

Cluster robust (country-level) standard errors in parentheses: *** p<0.01, ** p<0.05, * p<0.1.

Model 1 investigates the relationship between absolute foreign bank presence and financial development, whereas model 2 investigates the relationship between relative foreign bank

presence and financial development.

For all models (3-6) the foreign bank presence terms are insignificant. However, for models 3 and 5 the asymmetric information terms and the interaction terms between foreign bank presence and asymmetric information are significant at the 5 percent level. This indicates that for these models, foreign bank presence tends to affect credit through its interaction with the degree of asymmetric in the host-country. In these models, Premium has a negative

coefficient. This is in line with existing literature, since higher risk premiums are expected to negatively influence credit supply. Higher levels of GDP per capita again correspond to a higher amount of private credit, whereas GDP growth seems to negatively influence it.

ContractEnforcement also has a negative coefficient, indicating that the amount of days necessary to enforce a settlement dispute negatively influences credit supply in the host country.

(21)

21 Table 2: Financial Development and Foreign Bank Presence: Asymmetric Information

(3) (4) (5) (6)

VARIABLES Full

Sample

Full Sample

Full Sample Foreign Bank Presence (absolute) 2.420 12.68

(7.254) (10.51)

Foreign Bank Presence (relative) -5.252 4.980

(3.558) (4.632)

Premium -0.354** -0.780**

(0.170) (0.297)

Creditor Information 0.192 1.061*

(0.707) (0.549)

Foreign Bank Presence (absolute)#Premium 0.674***

(0.207) Foreign Bank Presence (absolute)#Creditor

Information

-0.378 (1.285)

Foreign Bank Presence (relative)#Premium 1.024***

(0.351) Foreign Bank Presence (relative)#Creditor

Information

-2.189**

(0.941)

Ln(GDP per Capita) 25.48*** 23.35*** 28.35*** 25.29***

(7.658) (6.387) (7.103) (6.870)

GDP Growth -0.161 -0.0326 -0.220** -0.0479

(0.109) (0.125) (0.108) (0.125)

Inflation 0.00846 -0.0128 -0.0134 -0.0156

(0.0446) (0.0400) (0.0436) (0.0409)

Contract Enforcement -0.0194** -0.0205 -0.0214*** -0.0202

(0.00819) (0.0185) (0.00777) (0.0186)

Constant -178.1** -154.1** -198.0*** -168.7**

(68.51) (59.88) (64.93) (64.37)

Observations 435 767 425 758

R-squared 0.246 0.052 0.255 0.056

Adj. R-squared 0.234 0.043 0.243 0.047

Number of countryid 55 102 54 102

Country FE YES YES YES YES

Models 3 and 4 investigate the relationship between absolute foreign bank presence and financial development, whereas models 5 and 6 investigate the relationship between relative

foreign bank presence and financial development. In addition, models 3 and 5 use the risk premium on lending as the proxy for asymmetric information, while models 4 and 6 use

creditor information.

(22)

22 To be able to determine the actual effect of asymmetric information in the relationship

between foreign bank presence and private credit, the models with significant interaction terms will be interpreted using marginal effects. According to Brambor et al. (2006), estimating the marginal effect of X (Foreign Bank Presence) on Y (FinDev) is extremely important when using a conditioning variable Z (Asymmetric Information). The intuition behind this is because the coefficient of X represents the effect on Y only if the conditioning variable Z is zero. However, since the conditioning variable is almost never zero, it is unclear what is the actual effect of X on Y. In line with Brambor et al. (2006), the marginal effect of Foreign Bank Presence on FinDev in this analysis is calculated as follows:

𝜕𝐹𝑖𝑛𝐷𝑒𝑣

𝜕𝐹𝑜𝑟𝑒𝑖𝑔𝑛 𝐵𝑎𝑛𝑘 𝑃𝑟𝑒𝑠𝑒𝑛𝑐𝑒= 𝛽₁+ 𝛽₃ ∗ 𝐴𝑠𝑦𝑚𝑚𝑒𝑡𝑟𝑖𝑐 𝐼𝑛𝑓𝑜𝑟𝑚𝑎𝑡𝑖𝑜𝑛 (7) Figure 1 shows the marginal effects of Foreign Bank Presence on FinDev, conditional on Asymmetric Information for models 3 and 5 respectively. According to these graphs, Premium enhances the marginal effect of foreign bank presence on private credit (figures 1A and 1B).

To be more specific, the effect of foreign bank presence on the amount of credit supplied in the host-country becomes significant when both the lower and upper bounds of the 95 percent confidence interval are below or above the zero line. This means that absolute foreign bank presence significantly influences private credit for risk premiums higher than 19.5 percent (figure 1A). However, since only 4 percent of the observations have a risk premium higher than this threshold level, it is unlikely that absolute foreign bank presence has a significant effect for a randomly chosen country-year combination. Relative foreign bank presence has a significant influence for risk premiums lower than -2 and higher than 14 percent (figure 1B).

Because 8 percent of the observations fall in this range, it is twice as likely that relative foreign bank presence affects credit compared to absolute foreign bank presence in this setting. The sign of the impact is somewhat puzzling. Where theory suggests that higher risk premiums should negatively influence the impact of foreign bank presence on private credit, these results suggest the opposite.

(23)

23 Figure 1: The Marginal Effect of Foreign Bank Presence on Financial Development

1A: Model (3) 1B: Model (5)

VII. Robustness

As noted earlier, the initial findings on the relationship between foreign bank presence and financial development, as measured by private credit, are not exactly in line with existing literature. Instead of an expected negative relationship, the results in this analysis point to a positive relationship for absolute foreign bank presence, and an insignificant relationship for relative foreign bank presence. To be able to shed light on the sensitivity of this analysis, we run models 3-6 using different subsamples, time periods, control variables and estimation techniques.

VII.I. Sample

Detragiache, Tressel & Gupta (2008) find a negative relationship between relative foreign bank presence and credit for 89 low-income countries. In addition, Claessens & Horen (2014) find that the negative relationship between relative foreign bank presence and private credit is mainly driven by developing countries. To account for this host-country heterogeneity, models 3-6 will be run using developing economies, emerging markets and high income countries separately. Income classifications are in line with table 9 in the appendix.

-50 050100150

Marginal Effect of Foreign Bank Presence (absolute)

-20.5 -.5 19.5 39.5 59.5 79.5 99.5 119.5 139.5 Premium

-100 0

100200300

Marginal Effect of Foreign Bank Presence (relative)

-20.5 -.5 19.5 39.5 59.5 79.5 99.5 119.5 139.5 Premium

(24)

(3) (4) (5) (6)

VARIABLES Developing

Economies

Developing Economies

Developing Economies Foreign Bank Presence (absolute) 5.007 25.12**

(8.255) (11.18)

Foreign Bank Presence (relative) -5.139 0.870

(3.556) (5.295)

Premium -0.306** -0.570*

(0.116) (0.296)

Creditor Information 0.591 -0.0881

(1.442) (0.962)

Foreign Bank Presence (absolute)#Premium 0.655***

(0.151) Foreign Bank Presence (absolute)#Creditor

Information

-2.516 (2.339)

Foreign Bank Presence (relative)#Premium 0.825**

Information

-1.358 (1.476)

Ln(GDP per Capita) 17.33** 14.07 21.75*** 21.25**

(6.984) (9.891) (7.687) (9.873)

GDP Growth -0.131 -0.155 -0.204* -0.235*

(0.105) (0.121) (0.108) (0.123)

Inflation 0.0546 -0.0101 0.0238 -0.0206

(0.0597) (0.0610) (0.0580) (0.0621)

Contract Enforcement -0.0139* 0.00378 -0.0177** 0.0123

(0.00752) (0.0118) (0.00702) (0.0122)

Constant -98.45* -82.91 -125.4** -134.5

(53.69) (80.75) (60.34) (82.29)

R-squared 0.291 0.079 0.279 0.057

Adj. R-squared 0.263 0.059 0.250 0.037

(25)

25 Table 3 shows the empirical results for models 3-6 for developing countries. Only model 4 shows a positive significant foreign bank presence term. However, since the asymmetric information term and the interaction term are insignificant there is no reason to assume that this relationship is influenced by asymmetric information in this model. For models 3 and 5 the asymmetric information term and the interaction term are significant at the 5 percent level, indicating that foreign bank presence influences private credit through its interaction with asymmetric information. In these models, the risk premium on lending has a negative coefficient as expected. The coefficient for GDP per capita is positive, and contract

enforcement has a negative coefficient. In model 6, all independent variables are insignificant.

Figure 2 shows the marginal effect of foreign bank presence on the supply of private credit for models 3 (2A) and 5 (2B) respectively. Notice that the risk premium on lending has an

enhancing effect again on the marginal effect of both absolute and relative foreign bank presence. Absolute foreign bank presence becomes significant at risk premiums higher than 17 percent, while relative foreign bank presence is significant for premiums lower than 0 percent or higher than 19 percent. This translates to 7,5 and 8 percent of the observations.

Again, the risk premium on lending seems to positively influence the relationship between foreign bank presence and private credit, in contrast to findings by Detragiache, Tressel &

Gupta (2008) and Claessens & Horen (2014).

Figure 2: The Marginal Effect of Foreign Bank Presence on Financial Development

2A: Model (3) 2B: Model (5)

-50 050100150

Marginal Effect of Foreign Bank Presence (absolute)

-20.5 -.5 19.5 39.5 59.5 79.5 99.5 119.5 139.5 Premium

-50 050100150200

-20.5 -.5 19.5 39.5 59.5 79.5 99.5 119.5 139.5 Premium

(26)

(3) (4) (5) (6)

VARIABLES Emerging

Markets

Emerging Markets

Foreign Bank Presence (absolute) -0.860 -15.79

(23.11) (22.86)

Foreign Bank Presence (relative) 6.453 9.860

(11.89) (8.511)

Premium -2.562 -0.853

(2.253) (1.104)

Creditor Information -0.0905 2.494**

(0.998) (0.971)

Foreign Bank Presence (absolute)#Premium 3.461 (3.102) Foreign Bank Presence (absolute)#Creditor

Information

0.894 (1.703)

Foreign Bank Presence (relative)#Premium 1.012

Information

-3.945**

(1.664)

Ln(GDP per Capita) 34.20** 29.14** 35.01** 26.24**

(15.86) (10.99) (16.08) (11.96)

GDP Growth 0.0348 -0.0577 0.0553 -0.0796

(0.206) (0.137) (0.200) (0.136)

Inflation -0.149 0.000919 -0.166 0.0268

(0.111) (0.0485) (0.106) (0.0473)

Contract Enforcement -0.0432 -0.0211** -0.0355 -0.0230**

(0.0252) (0.00922) (0.0250) (0.0101)

Constant -240.4 -198.7* -259.9 -182.3

(148.5) (107.9) (156.0) (118.2)

R-squared 0.337 0.114 0.337 0.153

Adj. R-squared 0.302 0.087 0.301 0.127

(27)

27 The results for the emerging countries are displayed in table 4. According to these results, only relative foreign bank presence seems to influence private credit significantly through the interaction with creditor information (model 6). Where creditor information itself has a positive sign as expected, the sign of the interaction term is negative. This indicates that the creditor information score has a reductive impact on the marginal effect of relative foreign bank presence, again in contrast to what theory suggests.

This marginal effect, as shown in figure 3, becomes significant at creditor information scores of 5 and 6. Since 45 percent of the observations have a creditor information score like this, it is considered likely that relative foreign bank presence influences private credit negatively for a randomly chosen emerging country.

Figure 3: The Marginal Effect of Foreign Bank Presence on Financial Development 3A: Model (6)

The final subsample contains the high income countries, which consist of OECD and OHI countries. The results are shown in table 5. For the high income countries, only relative foreign bank presence significantly influences private credit when using the risk premium on lending as conditioning variable (model 5). However, since the interaction term is

insignificant, it can be concluded that relative foreign bank presence influences the supply of credit, but not through its interaction with asymmetric information. The marginal effect is therefore equal to the coefficient estimate. The coefficient on relative foreign bank presence is negative, in line with Claessens & Horen (2014). The independent variables of models 3, 4 and 6 are all insignificant, which means that for these models foreign bank presence does not influence private credit in high income countries.

-40-20 02040

0 1 2 3 4 5 6

Creditor Information

(28)

(3) (4) (5) (6)

VARIABLES OECD

+ OHI

OECD + OHI

Foreign Bank Presence (absolute) -12.76 62.21

(33.21) (50.81)

Foreign Bank Presence (relative) -17.24** 20.91

(6.455) (31.69)

Premium -0.562 -0.0115

(0.573) (0.873)

Creditor Information 3.201 2.759

(2.404) (2.222)

Foreign Bank Presence (absolute)#Premium 3.559 (2.379) Foreign Bank Presence (absolute)#Creditor

Information

-1.678 (3.588)

Foreign Bank Presence (relative)#Premium 2.900

Information

-5.092 (5.409)

Ln(GDP per Capita) 38.57* 62.12** 37.18* 55.85**

(17.99) (28.16) (19.86) (23.75)

GDP Growth -0.495* -0.0358 -0.509* -0.0808

(0.271) (0.339) (0.257) (0.316)

Inflation -0.0432 -0.0275 -0.0390 -0.00857

(0.140) (0.220) (0.136) (0.229)

Contract Enforcement -0.0147 -0.0556 -0.00966 -0.0674

(0.0194) (0.0824) (0.0176) (0.0910)

Constant -348.9* -599.9* -337.1 -504.2**

(183.9) (294.1) (207.9) (226.5)

R-squared 0.272 0.070 0.326 0.067

Adj. R-squared 0.220 0.034 0.276 0.031