Faculty of Economics and Business
Master Thesis Economics Monetary Policy, Banking and Regulation April-August 2016
Prof. Dr. Aerdt Houben Dr. Ward Romp
What is the impact of foreign bank presence on financial stability in West-European countries?
Annemijn van Rheden 10204512
Abstract
Using bank-level data, a panel fixed effects regression will be done to estimate the effect of foreign bank presence on financial stability over the timespan of 1995-2013 in 15 West-European countries. In contrast to earlier research, two different proxies are used to measure financial stability: the Z-score and credit growth. This research does not show that foreign bank are significantly more financially stable than domestic banks, nor that financial stability of foreign presence is significantly lower during the financial crisis. However, the estimated coefficients do suggest the particular direction. In addition, no statistical evidence is found that the home country of the foreign player matters, although evidence suggests that other than high-income (oecd) countries improve financial stability. Furthermore, foreign owned cooperative banks are found more financially stable than their domestic counterparts and shows the importance of correcting for heterogeneity among business models.
2
Table of Contents
1. INTRODUCTION 3-4
2. LITERATURE REVIEW 5-12
The transmission channels of foreign bank presence on financial stability The spillover effects form home country to host country
Empirical results to date Contribution of this paper
3. MODEL AND DATA 13-23
Data sources Measures
Descriptive statistics Empirical Method
4. RESULTS & DISCUSSION 23-36
Baseline measure: Z-score Second measure: credit growth Robustness on bank characteristics The role of the crisis
5. CONCLUSION 37
REFERENCES 38-41
3
1. Introduction
The Dutch banking sector is highly concentrated and dominated by a small number of large national banks. This can be a threat to financial stability if the services provided are not stable and efficient due to high market power. The general opinion of both the supervisory authorities and the Ministry of Finance is that the stability and efficiency of banking services are best guaranteed in a sector characterized by less concentration and more diversity. Consequently, entrance of new players could increase the financial stability in the Dutch banking sector (DNB, 2015). New players can take the form of new, innovative financial institutions or allowing multinational banks entering the Dutch financial market. This paper will focus on the latter, and investigates if foreign players enhance financial stability in the host country.
On the one hand, a large part of the literature supports the hypothesis that foreign bank presence increases competition, improves the robustness of the financial system and brings greater financial stability.1 Especially in developing countries, the influence of foreign bank entry seems to improve financial services and increases the efficiency of domestic banks and hence improves financial stability (Hermes and Lensink, 2004). For high-developed countries this effect is less clear-cut. This stems from the fact that high-developed countries generally have a more developed financial system, are more globally integrated and are therefore less sensitive to foreign entrants (Claessens and Van Horen, 2013).
On the other hand, foreign bank presence can also result in increased financial volatility in times of financial distress (Claessens and Van Horen, 2013). In the crisis of 2008, the banking interdependence has threatened financial stability. Especially in European countries, cross-border lending has fallen from 2008 onwards, leading to liquidity constraints in the financial markets. In turn, this instability was transmitted from the banking sector to the real economy by influencing private agent’s decisions. This lead to a decrease in both consumer confidence and spending with negative consequences for economic growth.
The interconnectedness of the financial system and its impact on the real economy has spread fear among policymakers whether or not financial stability is at jeopardy. Therefore, more regulation and policies might be quintessential to achieve stability in the banking sector
1
Claessens et al, 2001; Clarke et al., 2003; Claessens 2006; Claessens, & Van Horen, 2008; Cull, & Martinez Peria, 2013.
4 (Berger, Klapper, & Turk-Ariss, 2008).
The purpose of this paper is to research the impact of foreign players on financial stability in West-European countries. In order to do this, two key questions will be answered. First, what is the effect of foreign bank presence on financial stability? Second, what is the effect of foreign bank presence during the financial crisis? To answer these questions, a panel fixed effects regression will be done over the timespan of 1995-2013, analyzing 833 banks from 15 countries. In contrast to earlier research, two different measures are used to measure financial stability. Using the Z-score and credit growth as proxies for financial stability, this research does not show that foreign bank are significantly more financially stable than domestic banks, nor that financial stability of foreign presence is significantly lower during the financial crisis. However, the estimated coefficients do suggest the particular direction. In addition, no statistical evidence is found that the home country of the foreign player matters, although evidence suggests that other than high-income (oecd) countries improve financial stability. Furthermore, foreign owned cooperative banks are found more financially stable than their domestic counterparts and shows the importance of correcting for heterogeneity among business models.
This paper is structured as follows. It commences with Section 2 in which the theoretical background, as well as an overview of previously conducted studies are discussed. Additionally, it provides a reflection of these studies along with the contributions of this study. Section 3 describes the data, followed by an explanation on the research method used. Section 4 presents the results of the empirical study and discusses the findings. Finally, section 5 concludes.
2. Literature Review
In the literature, several issues on the effects of foreign bank presence on financial stability are discussed. This paragraph provides an overview. Firstly, the theoretical mechanisms of foreign presence on financial stability are explained in section 1 and 2. Subsequently, the empirical findings of earlier research is presented. Section 4 concludes with the contribution of this paper.
2.1 The transmission channels of foreign bank presence on financial stability
On a macro-level, foreign bank presence may affect financial stability in two ways. First of all, foreign bank presence can increase competition. In order to sustain their market share, domestic banks can be forced to improve the quality of their services and cut costs, thereby increasing efficiency (Lensink and Hermes, 2004). However, does increased competition lead to more financial stability? Two competing views arise in the literature. Proponents of the competition-stability view argue that less market power of domestic banks can induce lower bank risk. If a few banks with large market power charge higher prices for their services, they might exacerbate moral hazard incentives, shifting customers into riskier projects (Berger et al., 2009). Banks with less market power cannot conduct this monopolistic behaviour. Nonetheless, opponents of the competition-stability view emphasize that more bank competition puts margins under pressure. This, in turn, could lead to increased bank risk taking. Less bank competition should lead to higher profits and thus higher capital buffers that protect them against financial-wide shocks, therefore reducing financial fragility (Boyd et al., 2004).
Secondly, domestic banks are often more likely to benefit from a government safety net. Following the too big to fail doctrine, this may give financial institutions an incentive for excessive risk-taking, making financial instability more likely (Mishkin, 1999). Mishkin therefore argues that improvement of supervisory, regulatory and institutional factors can be a consequence of foreign entry (2009). For instance, broadening the mandate of an independent financial supervisor can lead to stricter rules in the domestic sector and thereby enhance financial stability for the system as a whole. However, Allen and Gale (2000) point out that
6
monitoring few larger domestic banks may be more effective in terms of supervision and implementation of new regulation.
2.2 Spillover effects from home country to host country
Financial stability can also arise via spillover effects of individual institutions across countries. Most subsidiaries do not operate autonomously, but are part of the bank holding with an international risk-diversifying strategy. The decisions made by the subsidiary in the host country are then, to a certain extent, restricted by the vision of the parent bank in the home country. Consequently, this creates positive and negative spillover effects.
On the one hand, on a positive note, financial stability in the host country can be enhanced due to risk diversification of cross-border banks. De Haas and Van Lelyveld (2006) argue that deterioration of the home country economic conditions might induce the parent bank to increase activities and investments abroad to seek for new, profitable opportunities. Subsequently, foreign subsidiaries might be less sensitive to country-specific shocks as they act in line with the risk-diversifying strategy of the parent bank that is active in various countries. Therefore, the likelihood of a bank failure declines. Secondly, the parent bank can ultimately step in during times of financial distress of the foreign subsidiaries to allocate capital and liquidity through the internal capital market. The parent bank’s support will make the subsidiaries less prone to shocks in the host country (De Haas and Van Lelyveld, 2006). If this strategy is translated into stable credit supply, foreign subsidiaries can act as a stabilising force in the host country. Especially when a parent bank considers its subsidiary strategic, it will not pull back those foreign operations in times of financial distress. Where domestic banks are forced to decrease credit supply, foreign banks can then substitute for the domestic players in the loan market. Thirdly, foreign banks can be more efficient and introduce new and more developed services in the host country. If domestic banks are forced to improve their banking operations, this can benefit the stability of the financial system as a whole. However, this is primarily the case for developing markets, as foreign banks are more likely to introduce more advanced services and technologies than in developed regions (Allen, 2006; Hermes and Lensink, 2004).
7
On the other hand, foreign bank presence can also have negative spillover effects. First, parent banks have more flexibility to reallocate their capital than domestic banks. In line with their risk-diversifying strategy foreign banks can reallocate capital to their subsidiaries in different regions to avoid country-specific shocks. This flexibility in capital flows can lead to a less stable credit supply in the host country, decreasing financial stability. In contrast, purely domestic banks lack this specific internal capital market and are therefore less likely to be as sensitive to home country shocks. Consequently, foreign banks tend to have a higher lending volatility and can enhance financial instability (Goldberg, Dages, & Kinney, 2000; De Haas and Van Lelyveld, 2006). A second destabilising force considered is contagion (Allen, 2006). If the parent bank prioritizes the activities in the home country and experiences a domestic shock, it is likely to reduce activities of foreign subsidiary first to protect primary activities back home. In this scenario, the foreign bank presence results in a negative force for financial stability.
Theoretically, this means that risk diversification across countries can increase financial stability in individual countries. However, when a crisis hits, the diversification benefits might be overshadowed by the risk of contagion from foreign players to domestic banks.
2.3 Empirical results to date
Before the empirical results are discussed, it is important to understand how financial stability is defined and what proxies are used (Allen and Gale, 2005; Andrianova et al., 2015). Financial stability explains the likelihood of financial institutions suddenly collapsing and causing damage to the real economy (Allen and Gale, 2005). Put differently, it captures the risk that the financial volatility of institutions spills over to the real economy. To date, however, no general indicator for financial stability is found due to the complexity of capturing its full definition in one measure. The three main proxies for financial stability used in the literature on foreign bank presence are: credit supply, the Z-score and lastly profit and margin.
Most authors that study the spillover effects of foreign subsidiaries on the host country’s financial stability exploit (changes in) credit supply. In that sense, a decrease in credit supply by a foreign bank reflects the financial distress of the banking sector. In the literature, foreign subsidiaries are found to be sensitive to home country shocks and less so to host
8
country shocks. Empirical evidence on the effect of financial distress in the home country on the credit supply abroad starts with research by Peek and Rosengren (2000). They find that the credit supply of Japanese subsidiaries in the U.S. decreased when their parent bank was hit by the Japanese banking crisis in the period 1988-1995. More recently, Cetorelli and Goldberg (2012) and Claessens and Van Horen (2013) analyse the internal capital market and find that, when hit by a home country shock, capital allocation to foreign subsidiaries is reduced. In turn, foreign subsidiaries were forced to cut lending abroad, thereby increasing financial instability in the host country. In view of this, the literature points to the importance of parent bank characteristics.2 For example, if the parent bank experiences a decrease in financial health it is reluctant to cut credit supply of the foreign subsidiary. In contrast, when a shock takes place in the country where the foreign bank is active, it seems that the subsidiary plays a stabilizing role in terms of credit growth. De Haas and Van Lelyveld (2010) show that domestic banks cut their lending more than foreign banks when the domestic economy is hit by a shock.
However, results change when the financial crisis is taken into account. Studies suggest that, at the height of the recent crisis in 2008, multinational banks created negative spillover effects through their foreign subsidiaries by reducing credit supply (De Haas et al., 2011; De Haas and Van Lelyveld, 2014). Especially in Europe, findings suggest that European banks cut their cross-border lending during the financial crisis. Dekle and Lee (2015) show that this result is partly driven by the levels of sovereign debt, as banks in European countries with high sovereign debt levels cut their lending by more than the parent banks that are situated in other countries. This decline in international bank loans had dramatic consequences for output growth. On a positive note, foreign banks are found to enhance financial stability when the internal capital market is strong and the parent bank is strategically committed to the foreign market in terms of real estate (Navaretti et al., 2010; Claessens and Van Horen, 2013).
The second measure considered isthe Z-score. This bank-specific measure describes the probability that the value of a banks’ assets become lower than the value of its debt. In other words, it indicates the likelihood of a bank becoming insolvent and hence conveys financial
2
E.g. Peek and Rosengren, 2000; Hermes and Lensink, 2004; De Haas and Van Lelyveld, 2006; Buch and Goldberg, 2014; Claessens and Van Horen, 2014
9
instability. To my knowledge, the Z-score has not yet been used to describe the effect of foreign presence on financial stability. However, in order to shed light over the usefulness of this proxy, related literature will be presented.
The Z-score is widely used to explain the effect of concentration ratio on financial stability.3 In this specific strand of the literature, the role of foreign players increasing competition is widely recognized in theory. However, no specific measure is taken into account to capture the interaction of foreign players with concentration ratio and the effect on financial stability. Furthermore, Cihak and Hesse (2010) introduce the Z-score to estimate the differences in the nature of financial institutions (i.e., Islamic banks vs. commercial banks) on financial stability. Their results suggest that the size of the institutions matter for financial stability. Put differently, small Islamic banks are more financially stable than small commercial banks, whereas large commercial banks tend to be financially more stable than large Islamic banks.
Lastly, profit and margins are used to specifically describe the effect of foreign bank presence on domestic behaviour. When the entry of foreign banks results in a downward pressure on the margins and profitability of their domestic counterparts, it can increase the likelihood of bank failure. This effect differs across the levels of economic development (Claessens et al., 2001; Lensink and Hermes, 2004). In the short run, evidence shows that foreign bank presence leads to higher costs, due to investment expenses for innovation of systems and financial services, and lower margins for domestic banks at lower levels of economic development. In the long run, the increased efficiency can lead to a more developed financial system and hence greater financial stability. For higher levels of development, foreign presence seems to have positive effects, with falling costs, or does not show any statistically significant effect on domestic banks. This suggests that countries with a less developed economy are more sensitive to foreign bank entry, influencing the financial sector as a whole. Instead, in a both economic and financial well-developed country, foreign entry does not seem to have a large impact on the banking sector. Moreover, the results of Claessens and Van Horen
3
e.g. Boyd and Runkle, 1993; De Nicolo et al., 2004; Maechler et al., 2005; Uhde and Heimeshoff, 2009; Beck et al, 2010; Beck et al., 2013
10
(2013) suggest that when multinational banks enter a country with relative similarities in income level, institutional development, and geographical location, foreign banks perform better than domestic banks in terms of profitability. Ultimately, they find that the equity ratio and capital adequacy is higher and the provision for non-performing loans is lower for foreign banks than for domestic banks (Claessens and Van Horen, 2012;2013). This result suggests that foreign banks are more conservative than their domestic counterparts with respect to their risk-taking behaviour and can act as a stabilizing force in the host countries’ banking sector. In table 1 the most important conclusions of the studies discussed above are summarized.
In conclusion, the estimated effect of foreign presence on financial stability is dependent on methods and measures used. By using credit supply as an indicator for financial stability, transmission mechanisms are explained. Findings suggest that foreign banks increase financial stability if the host country is hit by a shock, but create negative spillover effects if the parent bank experiences a home country financial crisis. This is, however, dependent on the health of the parent bank and to what extent it is committed to the specific host market. Moreover, literature that adopts profitability and margin as an indicator for domestic behaviour finds that domestic banks in high-developed countries are not impacted by foreign bank entry. In addition, foreign banks that have their home country close by act more conservative than their domestic counterparts.
Table 1 Authors No. of banks Country Measure of financial stability Effect foreign presence on
financial stability Driving force for financial stability
P&R (2000) 137 U.S. Credit Supply - Home country shock
C&vH (2012) 3615 Cross-Country Credit Supply + Commitment to host country
DH&VL (2014) 401 Cross-Country Credit Supply - Financial Crisis; Reduce in credit supply
D&L (2015) 21000 Cross-Country Credit Supply - Financial Crisis; Decrease in European cross-border lending due to high sovereign debt levels
C&G (2012) 44 U.S. Internal Capital Market + Host country is strategically important
DH&VL (2010) 45 Cross-Country Internal Capital Market + Absorb host country shock if host country is priority L&H (2004) 990 Cross-Country Profits and Margins + No/low effect on domestic behaviour in developed
countries
C&vH (2013) 1318 Cross-Country Profits and Margins + Foreign banks act conservative if host country is geographically closeby
Empirical literature on the effects of foreign presence on financial stability
Abbreviations authors: P&R (Peek and Rosengren); C&vH (Claessens and Van Horen); DH&VL (De Haas and Van Lelyveld); D&L (Dekle and Lee); C&G (Cetorelli and Goldberg); L&H (Lensink and Hermes)
2.4 Contribution of this paper
The literature shows that foreign presence can increase financial stability, depending on the nature of shocks to the financial system, the level of development and bank-specific characteristics. This paper will contribute to the literature by analysing data of 914 banks from 1995-2013, correcting for heterogeneity in both markets and banks. Firstly, the differences in regulations and level of economic development are minimized by analysing West-European countries only. Secondly, the proxies for financial stability are defined on bank-specific level. The advantage of this approach is that many banks across countries can be compared, using the same criteria for measuring bank’s soundness. Therefore, financial stability in this paper takes the form of financial soundness of an individual institution, rather than for the country as a whole. Furthermore, the financial crisis of 2008 is considered to capture a system-wide financial shock. The main objective of incorporating this event in the study is to describe the behaviour of foreign players in times of financial distress.
The hypothesis is as follows: during normal times in developed countries, the foreign players are expected to be more financially stable than domestic players. When the geographical distance is small and the level of development is relatively the same, foreign banks are assumed to be committed and aim for a risk-diversifying strategy across borders. However, when foreign banks experience both a host and a home country shock and will, due to home bias, decrease activities of their foreign subsidiaries. Hence, foreign banks are assumed to be less stable during times of financial stress.
3. Methodology & data
First, the sources of the data will be presented. Secondly, the measures used are explained and thirdly , the model used will be presented.
3.1 Data Sources
Data was collected using different databases. The starting point is the dataset on (foreign/domestic) bank ownership provided by Claessens and Van Horen (2013), available for the 1995-2013 period. Consequently, additional information on bank-level is extracted from
13
income sheets and accounting information using Bureau van Dijk BankScope database, including bank specialization. As BankScope does not provide data for all banks for the full sample period, the panel data set is unbalanced. In addition, the balance sheet information reports different measures and different currencies and therefore all financial variables from BankScope are converted into thousands of U.S. dollars, on yearly basis. BankScope provides information in both consolidated and unconsolidated status. As this paper focusses on the behaviour and effects of individual entities, only the unconsolidated balance sheets are used. One drawback of this setup is that no characteristics of the parent back can be taken into consideration. However, it ensures that an individual bank’s riskiness is captured and avoids double counting. Of 44 institutions only consolidated balance sheet information is available and they are therefore deleted from the sample. Lastly, macro-economic information on country-level is obtained from several databases. GDP growth and inflation is taken from the International Financial Statistics database from IMF. Concentration ratio is taken from the World Development Indicators of the World Bank, which is available from the period 1997-2013. This results in 15333 observations for 833 individual banks, over the time-span of 19 years (1995-2013). The majority of banks in the sample are commercial (N = 424, 51%), followed by saving banks (N = 273, 33%) and cooperative banks (N = 134, 16%).
3.2. Measures
A sample was constructed to explore the effects of foreign banks on financial stability based on the available data. Fifteen countries4 that are considered to diminish heterogeneity in economic development levels are included. To select foreign owned banks in these countries, a dummy variable is computed (1= foreign owned, 0 = not foreign owned). If at least 50% of the bank shares are foreign owned, evaluated in terms of the bank’s assets relative to the total assets of the country, the bank is defined as foreign owned (Claessens et al., 2001; Lensink and Hermes, 2004a; Cihak and Hesse, 2010). Following the hypothesis, foreign banks are expected to be more financially stable than domestic banks in normal times, but are expected to reduce stability in times of financial distress.
4
The countries considered in this paper are Austria, Belgium, Denmark, France, Finland, Germany, Greece, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden and the United Kingdom.
14
As a second step, two variables are constructed reflecting financial stability of the bank: individual bank soundness and credit growth. The measure for individual bank soundness is the Z-score. This proxy explains the probability of a bank becoming insolvent, by comparing the buffer of a bank, indicated by capitalization and returns, with the volatility of those returns. In other words, the Z-score measures how volatile returns can turn before the bank depleted its equity. Over the last two decades, the Z-score has gained popularity as a proxy of bank soundness. This is mainly due to the simplicity of the measure in terms of calculation, using accounting information only. The Z-score is defined as 𝑍𝑖,𝑗,𝑡 ≡
(𝜅𝑖,𝑗,𝑡+µ𝑖,𝑗)
𝜎𝑖,𝑗 , where 𝜅𝑖,𝑗,𝑡 is equity capital divided over assets for bank i in country j at time t; µ is average return over the full period as percentage of assets for bank i in country j, and σ is the standard deviation of return on assets as a proxy for return volatility for bank i in country j.5 As the proxy combines capital, average returns and return volatility in one indicator, it increases with a bank’s capital ratio and decreases in volatility. A higher (lower) Z-score corresponds to low (high) probability insolvency risk (Cihak and Hesse, 2010; Uhde and Heimeshoff, 2009).
The advantage of the Z-score is that it is applicable to any level of risk-adjustment. For example, a bank that uses a high risk-high return strategy can lead to the same risk-adjusted returns as a bank conducting a low risk-low return strategy. In case of lower risk-adjusted returns, the bank can still compensate the Z-score by a high level of capitalization (Cihak and Hesse, 2010).
As a second proxy for financial stability, credit supply is added. This measure primarily explains the effect on the economy via the banking sector: an increase in credit supply can increase physical investment and GDP growth. Especially foreign banks can play a significant role in adding to output growth in the host country if they provide additional capital. In contrast, a decrease in the supply of credit can introduce financial instability. For instance, if a bank experiences a shock to bank liabilities, due to a bank run or a tightening internal capital market, this shock can force the bank into credit contraction in order to meet their reserve
5 Literature using the Z-score are for example:. Boyd and Runkle, 1993; De Nicolo et al., 2004; Maechler et al., 2005; Uhde and Heimeshoff, 2009; Beck et al, 2010; Cihak and Hesse, 2010; Beck et al., 2012; Andrianova et al., 2015 and many others.
15
requirements (Dekle and Lee, 2015). Following De Haas and Van Lelyveld (2006), credit growth was defined as 𝐶𝑟𝑒𝑑𝑖𝑡 𝐺𝑟𝑜𝑤𝑡ℎ𝑖,𝑗,𝑡 ≡
∆(𝑇𝑜𝑡𝑎𝑙 𝑙𝑜𝑎𝑛𝑠)𝑖,𝑗,𝑡
𝑇𝑜𝑡𝑎𝑙 𝐿𝑜𝑎𝑛𝑠𝑖,𝑗,𝑡−1, where total loans is defined as the sum of loans and gross loans, divided by total loans of bank i, lending in country j, at time t-1.
In order to correct for bank-specific and country-specific characteristics, control variables were added.6 Three bank-specific control variables were added in line with the research of Claessens et al. (2001): the cost-income ratio (total operating expenses/total operating income) to measure the bank’s efficiency, the liquidity ratio (liquid assets/total assets) and the profitability (net income/total equity). A bank is expected to be less financially stable if it is less efficient or, put differently, has higher costs of operations. Therefore, the expected sign is negative. For both liquidity and profitability, a bank is expected to be more stable if the ratios increase as there is room to improve the capital ratio (De Haas and Van Lelyveld, 2006).
Furthermore, macro-economic controls were used to control for cross-country differences. In particular, GDP growth and inflation rate were added as controls to capture macroeconomic developments that possibly affect the quality of bank assets. To correct for business cycles, one-period lagged values of GDP growth and inflation rate were included (Laeven and Majoni, 2003). A positive sign is expected for GDP growth, as lending opportunities increase and revenue streams are expected to go up as a consequence. In addition, private agents’ solvability might increase due to economic prosperity. The impact of inflation is also expected to be positive. With an increasing inflation rate the real interest rate is expected to rise, enabling banks to charge higher premiums for their services. In addition, real GDP per capita is added to correct for the level of development in a country. As the sample only considers high-developed countries, no specific sign is expected. To account for cross-country variation in financial stability caused by differences in market concentration, a proxy for concentration ratio is added to the model (Cihak and Hesse, 2010). Concentration ratio is defined as the share of assets of the 5 largest banks in comparison to the total assets of the country. For concentration ratio, the expected sign is ambiguous following the discussion on
6
16
the concentration stability vs. the concentration fragility view in the literature review.
3.3 Descriptive statistics
Firstly, a preliminary look at the Z-scores in figure 1a suggests high variability across the sample. Moreover, the distribution of the Z-score shows large differences in the Z-scores between both the size and the ownership of banks. First, banks with a small market share (measured by total bank asset ratio) take values that are higher and more diversified than for banks with a high market share. Second, the variability for domestic banks is higher than for foreign banks.
Table 2 presents the summary statistics of the sample and categorizes by ownership. The average Z-score for domestic banks (39.03) significantly higher than for foreign banks (20.29). This is explained by the components of the Z-score (equity ratio, return on assets and the standard deviation of the return on assets), that are all significantly higher for foreign banks than for their domestic counterparts. Therefore, it seems that due to the higher volatility of the returns, the foreign players in general have a lower Z-score.
Figure 1b presents the scatterplot of credit growth linked to market size of a bank, and categorized by ownership. The distribution is very similar to that of the Z-score, though the difference between foreign and domestic players seems less clear. A closer look at the outliers shows that a few observations are not representative due to miscalculations by, for instance, missing observations. Therefore, I remove 6 observations manually. In turn, the average credit growth for domestic banks (0.70) does not significantly differ from foreign banks (0.51).
17
Figure 1a: Scatterplot Z-score and Total Bank Asset Ratio, categorized by ownership
Figure 1b: Scatterplot Credit Growth and Total Bank Asset Ratio, categorized by ownership
Fig. 1. The upper figure (1a) shows the relationship between the Z-score and market share, as measured by the total bank asset
ratio. The lower figure (1b) shows the relationship between credit growth and market share, as measured by the total bank asset ratio. The information is categorized by ownership, where a bank is foreign owned if more than 50% of its assets are in foreign hands. The Z-score is defined as the sum of equity over total assets and return on assets divided by the standard deviation of the return on assets. Credit growth is defined as the sum of loans and gross loans, divided by total loans of bank i, lending in country j, at time t-1. The total bank asset ratio is calculated by the total assets of the bank divided by the total assets in the country. The indicators are calculated at individual bank-level and averaged over the time span (1995-2013). On horizontal axis, the total bank asset ratio is shown, whereas the Z-score and credit growth, resp., are shown on the vertical axis. The left panel displays domestic banks and the right panel presents foreign owned banks.
18
The bank-specific controls are presented in the third panel of table 2, where only liquidity shows a significant difference between domestic and foreign ownership. For macroeconomic control variables, displayed in the fourth panel, it seems that foreign subsidiaries enter countries with high levels of economic development and low concentration ratios.
Subsequently, a distinction is made between low-development and high-development (i.e. OECD countries). In addition, I correct for the geographical distance of high-developed countries, extending the research of Claessens and Van Horen (2013) who find that closer distance to the home country increases foreign bank performance and hence financial stability.
Table 2 Summary Statistics Mean N Mean N Dependent Variables Z-score 39.03 7620 20.29*** 2138 Equity Ratio 0.09 7629 0.10*** 2141 Return on Assets 0.005 7610 0.007** 2108 S.D. Return on Assets 0.007 11,243 0.01*** 2,899 Credit Growth 0.70 6888 0.51 1925 Variable of Interest
Bank Asset Ratio 0.02 7629 0.03** 2141
Bank-specific Controls Liquidity 0.20 7622 0.44*** 2132 Efficiency 0.64 7306 0.64 1968 Profitability 0.07 7610 0.06 2108 Macroeconomic Controls GDP per capita 39545.61 11438 50206.83*** 2994 GDP growth 0.02 10872 0.023*** 2877 Inflation Rate 85.95 10872 82.03*** 2877 Concentration Ratio (5) 0.67 9651 0.60*** 2624 Foreign Ownership Domestic Ownership
This table shows the total sample summary statistics for the variables used throughout the paper. The total sample consists of 15333 observations. A bank is defined foreign owned if more than 50% of the assets are in foreign hands. The Z-score is defined as z≡(κ+µ)/σ, where κ is the capital equity ratio, µ the return on assets and σ is the standard deviation of the return on assets. Credit Growth is defined as the first difference of total loans to the first lag of total loans. The difference between value of domestic and foreign owned banks at 95% confidence level is significant at 10%(*); at 5%(**); at 1%(***).
19
The sample is split and the dummy of ownership has now been replaced by 3 group of countries: if the home country of the foreign player is from a neighbour country, ownership is categorized as “Neighbour” (e.g. Germany for the Netherlands). All other home countries that are part of OECD, but are not direct neighbour countries, are grouped under “OECD” (e.g. U.S. or Italy for the Netherlands). All other countries are grouped under “Other”.7 In figure 2, the distribution is presented of the market share owned by foreign subsidiaries in each category. In addition, the Z-score per country is shown above the columns.
Figure 2
Figure 2 depicts the home country of the parent bank of the subsidiary, according to its total market share in the host country. Data is averaged over the sample 1995-2013. The bank can be domestically owned (Domestic) or foreign owned. The foreign owned banks are grouped in three categories: the home country of the parent bank is a direct neighbour of the country the foreign subsidiary is active in (Neighbour), the home country is not a direct neighbour but part of the OECD countries (OECD) and all other countries are grouped under “Other”. The average Z-score of the countries over the time period is shown above the columns.
7
Although the categorization is somewhat arbitrary, it primarily filters the main background characteristics of the home country of foreign entrants. In addition, it captures the advantage of cross-border banking close by versus overcoming cultural differences from far away. This is in line with Claessens and Van Horen (2013). OECD countries are defined following the OECD (2016). Other countries are e.g. Andorra, Brazil, Egypt, India, Indonesia and many more. 0% 20% 40% 60% 80% 100%
Market share owned by domestic and foreign banks per
country
OECD Other Neighbour Domestic
20
Besides a large variation of Z-scores between countries, the distributions of market share differ largely between the countries considered. For instance, Greece has only domestic banks and a Z-score of 14.53, whereas the bank assets of Norway are almost solely foreign owned and has a Z-score of 42.64. All other countries have a more diversified distribution, but no clear link between home country and financial stability can be seen directly from figure 2. Most interestingly, the market share of ‘other’ countries is only present in few European countries. A closer look at the data reveals that this specific form of foreign bank presence is mainly driven by historical reasons, such as foreign colonies. This pattern is not surprising since research has shown that foreign banks tend to follow their customers and therefore tend to enter countries with strong trade linkages (Goldberg and Grosse, 1991).
From figure 2, it seems that financial stability is not primarily driven by the origin of the parent banks of the foreign subsidiaries. Therefore, I also consider the business models of banks separately, defined as Savings Bank, Commercial Bank and Cooperative Bank. For savings banks, only 1 bank in the sample is foreign owned and hence this specialization is not considered separately in further analysis.
51% of the sample consists of commercial banks, with, on average, a total foreign share of 3%. The average Z-score for foreign banks is 21.07 and average credit growth is 21%. In contrast, 16% of the sample consists of cooperative banks, where 1% of the market share is in foreign hands. The average Z-score is 60.24 and the credit growth is 10%. These statistics suggest a significant impact on financial stability and are taken into account for further analysis.
3.4 Empirical method Baseline model
To measure the impact of foreign players on financial stability, the model used will distinguish the financial soundness of domestic and foreign banks. In other words, the model will estimate how risky a foreign player is in comparison to their domestic counterparts. The impact will be estimated using a time-fixed effects regression. Specifically, the model is defined as follows:
21
(1) 𝑆𝑖,𝑗,𝑡 = 𝛽1𝐹𝑖,𝑗,𝑡+ 𝛽2log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 + 𝛽3𝐹𝑖,𝑗,𝑡log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 + 𝛽4𝐵𝑖,𝑗,𝑡+ 𝛽5𝑀𝑖,𝑗,𝑡 + 𝜆𝑡+ 𝜀𝑖,𝑗,𝑡
𝑆𝑖,𝑗,𝑡 is the measure for financial stability for bank i in country j at time t. 𝐹𝑖,𝑗,𝑡 is the variable created to measure foreign ownership for bank i in country j at time t. The dummy turns 1 if a bank is foreign and is zero otherwise. The sign of the coefficient is expected to be positive, as foreign players are hypothesized to enhance financial stability. Secondly, log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 is the log-transformed measure of the bank’s market share, defined by the individual bank assets over the total bank assets in a country for bank i in country j at time t. No specific sign is expected for the coefficient of market share. In addition, the interaction between ownership and market share is added in the model by 𝐹𝑖,𝑗,𝑡log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 to capture the different impact between domestic and foreign banks in terms of market shares. This measure is estimated for each individual bank i in country j at time t. More specifically, if a foreign player plays a larger role in the market (i.e. owns a higher percentage of total bank assets in the country) it is expected to have a higher level of bank soundness and, hence, the expected sign is positive.
𝐵𝑖,𝑗,𝑡 is a set of variables for bank i in country j at time t, and 𝑀𝑖,𝑗,𝑡 are macroeconomic control variables for bank i in country j at time t. The bank-specific controls used in the estimations are liquidity (+), efficiency (+) and profitability (+) and are defined as ratios, where the expected signs of the coefficients are in parentheses. The country-specific variables included in the estimations are annual growth of GDP (+), the indexed inflation rate based on consumer prices (+), real GDP per capita in US dollars (?) and concentration ratio (?). Both the inflation rate and the GDP growth are added with the first lag to correct for business cycles. Furthermore, 𝜆𝑡 are the time-fixed effects and are added to reduce the threat of omitted variable bias. More specifically, these effects are included to control for common shocks across countries, such as business cycles and regulatory changes considering the financial system. Lastly, 𝜀𝑖,𝑗,𝑡 represents the residual.
22
financial stability. Following Lepetit and Strobel (2015) a yearly time-varying Z-score is used to overcome distributional problems.8 As a second measure, credit growth is used.
In attempt to account for the limitations in the dataset, two steps are taken. Firstly, after testing the data for heteroscedasticity and autocorrelation, clustered standard errors are added. Secondly, one of the most important issues in doing this estimation is that profitability has a high correlation with one of the components of the Z-score (return on assets).9 Therefore, the control variable will be added sequentially to obtain more robust results.
Financial Crisis
As hypothesized, it is expected that the effect of ownership during a financial crisis will differ from normal times. More specifically, in normal times (i.e. non-crisis periods) foreign players are expected to be more stable than domestic players. In contrast, during a financial shock in both their host and home country, they are expected to be less financially stable. Therefore, a second model will be estimated to capture the differences in time periods. More specifically, the following model is estimated:
(2) 𝑆𝑖,𝑗,𝑡 = 𝛽1𝐹𝑖,𝑗,𝑡+ 𝛽2log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 + 𝛽3𝑡𝐹𝑖,𝑗,𝑡log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 + 𝛽4𝐵𝑖,𝑗,𝑡+ 𝛽5𝑀𝑖,𝑗,𝑡 + 𝜆𝑡+ 𝜀𝑖,𝑗,𝑡
The model uses the same variables as in the baseline model. Again, 𝑆𝑖,𝑗,𝑡 is measured by both the Z-score and credit growth; 𝐹𝑖,𝑗,𝑡 represents the absolute number of foreign owners; log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 measures the market share of an individual bank; 𝐹𝑖,𝑗,𝑡log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡 is the interaction term of ownership and market size, and it’s coefficient can take different values over the different time periods, to correct for the crisis. The control variables take the form of 𝐵𝑖,𝑗,𝑡, where only liquidity is added to the model, and 𝑀𝑖,𝑗,𝑡, where GDP per capita, GDP growth and inflation rate are used for estimation.
8
See Lepetit and Strobel (2013) for a discussion on the different existing approaches to the construction of time-varying Z-score measures. By comparison of the Z-scores, their research shows that the Z-score with the best fit for G20 data between 1992-2009 is by the use of mean and standard deviation estimates of the return on assets calculated over the full sample combined with current capital-asset ratio.
9
23
In turn, the years of the crisis are set for 2008 and 2009. For this period, the expected sign of the interaction term of ownership (𝐹𝑖,𝑗,𝑡log(𝐴𝑠𝑠𝑒𝑡 𝑅𝑎𝑡𝑖𝑜)𝑖,𝑗,𝑡) is negative. In addition, the non-crisis periods are divided by a pre-non-crisis period (1995-2007) and a post-non-crisis period(2010-2013). The signs for the interaction terms are in these cases assumed to be positive.
Considering only one system-wide shock has two main drawbacks. Firstly, the European financial system has experienced more shocks in the sample period reviewed. Nevertheless, the dataset consists of yearly observations. By accounting for short-lived crises, the noise would be large as the exact dates cannot be captured in the data. Hence, only one crisis is considered that has a large impact on the financial landscape. In this case, the crisis started in the third quarter of 2008 and had a large impact on the financial system in 2009.10 Therefore, the period considered as the crisis are both years. Secondly, the period 2010-2013 is considered as a non-crisis period in the analysis, although in Europe the sovereign debt non-crisis took place. However, this event did not primarily influence the financial sector to the extent as the financial crisis did. Therefore, the analysis will only focus on the financial crisis of 2008-2009.
4. Results & Discussion
The results section consists of three subsections. First, the baseline model will be estimated for the Z-score as dependent variable. Second, the model will be estimated for credit growth. In the third section, I will analyze the specific bank-characteristics, such as home country of the parent bank and the bank’s specialization. In the last section, the effect of foreign presence during the crisis is analyzed.
4.1 Baseline measure: Z-score
Table 3 presents the results of the baseline model for the Z-score, as described in paragraph 3. The interaction coefficient of foreign presence enters regression (1) significantly positive at a 10%-level, suggesting that foreign banks are more financially stable than their domestic counterparts . However, although the coefficient remains positive, the effect turns insignificant
10
24
when regressions are run for robustness in column (2)-(5). Based on these results I do not conclude that foreign presence are significantly more stable than domestic banks in West-European countries.11 However, evidence does suggest a weak positive relation between foreign players and its financial soundness. This outcome is supported by the coefficient of ownership, that measures the amount of foreign banks in absolute numbers and also enters all specifications positively, but statistically insignificant. These results are in line with the conclusions of earlier research that foreign presence does not have a significant positive impact on domestic bank behavior in high-developed countries (Lensink & Hermes, 2004; Claessens and Van Horen, 2014). Apparently, correcting for the level of development in the sample selection has indeed shown that the foreign banks in West-Europe are not more financially sound than domestic players. This result could stem from the fact the financial sectors of these countries are largely integrated an cross-border banking does not have a significant influence on financial stability.
The scatterplot of the Z-score in figure 1a showed that the distribution of the Z-score seemed less dependent on ownership, but more on market share. Veritably, the coefficient of total bank assets enters all regressions negatively and significantly. This means that the Z-score decreases if a bank increases its market share, or put differently, large banks are more risky than small banks.12 This is also found by Laeven et al. (2014), who show that due to changes in the financial system over the last two decades, large banks engage in more market-based activities, hold less capital, have less stable funding and are organizationally more complex than small banks.
Concerning the bank-specific variables, only liquidity enters the model significantly and is negatively associated with financial stability. This contradicts the hypothesis. In turn, it seems that holding liquid assets does not increase financial stability, but increases the riskiness of banks. Intuitively, this can be explained by liquidity not being an indicator for solvability of a
11
Evaluation of the first model suggests that the significant result of foreign presence is mostly driven by outliers. I rerun the model with exclusion of the 1st and 99th percentile of the distribution of the Z-score and the significant impact of foreign presence disappears. I therefore conclude foreign bank presence generally does not have a significant positive impact on financial stability.
12 Following Cihak and Hesse(2010) and Laeven et al. (2014) an attempt has been made to account for small and large banks. Unfortunately, only 5 banks in the dataset can be defined as “large” (total assets of more than 1 billion U.S. Dollars) and the sample is therefore too small for empirical estimations.
25
bank. Put differently, if a bank holds more liquid assets, this does not mean a bank is more solvable and hence able to absorb financial shocks. Therefore, the negative relation shows that a higher liquidity ratio lowers the Z-score, as these assets are not used to increase capital ratios. Profitability and efficiency do not enter the model significantly. To control for possible endogeneity effects, both variables are omitted from the analysis in column (2)-(5) and seem to improve the model. In terms of the macro-economic control variables, GDP growth is positively associated with higher financial stability and inflation with lower financial stability, as also found by De Haas and Van Lelyveld (2006). Interestingly, GDP per capita has a small, but negative impact on financial stability. Hermes and Lensink (2004) show that foreign bank presence enhances financial stability in low-developed countries, but not per se in developed countries. The small, negative effect of GDP per capita on financial stability suggests that a higher level of financial development does not per se lead to more bank stability.
Lastly, the concentration ratio is added in column (5) and is positively associated with financial stability. Although the coefficient of the market share of banks is negative, it seems that if the five largest banks have a large share of the market, it enhances financial stability. This finding seems to contradict the finding of Laeven et al. (2014), stating that large banks are more risky than small banks. However, the fact that concentration ratio has a positive impact on financial stability is more likely to be related to the financial landscape of individual countries. As suggested by Berger et al. (2009) and Mishkin (2009), few institutions with much market power in industrial countries are easier to control for supervisory authorities. When more banks become large and concentration ratio decreases, the results found suggest that financial stability decreases as it is more difficult to control them. Apparently, only a few large banks is better for financial stability. This supports the concentration-stability view.
26
4.2 Second measure: Credit Growth
As a second measure, credit growth is used. I rerun the baseline model and replace the Z-score for credit growth. Results can be found in table 4, with the full baseline model in column (1) and additional regressions for robustness in columns (2)-(5). None of the models show any significant signs for the estimates and can be explained by the small variability of credit growth, as seen in the scatterplot in figure 1b.13 Therefore, using the second proxy for financial stability,
13
To improve the predictive ability of the model, both the dependent and independent variables are transformed, Table 3
Z-score and Foreign Presence
(1) Z-score (2) Z-score (3) Z-score (4) Z-score (5) Z-score
Ownership 10.91 6.257 5.609 5.374 4.310
(8.837) (8.140) (8.245) (8.190) (9.694) Log (Bank Asset Ratio) -6.798*** -6.269*** -6.252*** -6.253*** -6.902** (2.141) (1.838) (1.858) (1.862) (2.184) Ownership X Log (Bank Asset Ratio) 3.275* 2.392 2.199 2.149 1.991 (1.906) (1.757) (1.784) (1.778) (2.072) Inflation Rate(t-1) -1.244*** -1.124*** -1.161*** -1.408*** -1.392*** (0.323) (0.301) (0.302) (0.302) (0.342) GDP Growth(t-1) 66.22*** 68.95*** 71.35*** 64.69*** 72.88*** (19.20) (18.67) (18.63) (18.33) (20.78) GDP per Capita -0.000116* -0.000142** -0.000136** -0.0000844 (0.0000541) (0.0000556) (0.0000555) (0.0000590) Efficiency 0.409 (1.223) Liquidity -7.214*** -6.396*** -6.689*** -6.735** (2.167) (2.027) (1.974) (2.125) Profitability 0.658 (0.434) Concentration Ratio (5) 21.61*** (3.812) Constant 89.14*** 82.20*** 86.46*** 100.4*** 83.15** (26.85) (23.96) (24.24) (23.49) (26.90)
Time Fixed Effects Yes Yes Yes Yes Yes
Observations 8975 9463 9447 9447 8827
Adjusted R-squared 0.130 0.117 0.120 0.117 0.122
Clustered standard errors in parentheses, significance levels indicated with *p<.1, **p<.05, ***<.01.
Z-score is defined as the Equity and the mean of the Return on Assets, divided by the standard deviation of the Return on Assets. A bank is defined foreign owned if more than 50% for the assets are in foreign hands, and is domestic otherwise. Concentration ratio available from 1997 onwards.
27
it cannot be concluded that foreign players are significantly more financially stable than domestic banks. However, the coefficients found leave room for further analysis.
The coefficient found of the interaction between ownership and market size suggests a positive relationship between foreign owners and their credit growth. In addition, the coefficient for ownership in absolute numbers supports this conclusion and also enters the regression positively. Although the statistical evidence is not convincing, these results are in line with the estimations of the Z-score and suggest that foreign players are more likely to increase credit supply in general.
In contrast, the impact of market share is less clear. As seen in the distribution of credit growth (figure 1b), the size of a bank does not explain large part of the credit growth. In column (1) and (2), the market share (Log(Total Bank Asset Ratio)) enters the regressions negatively, indicating that larger banks are less likely to increase their credit growth. However, in column (3) and (4), efficiency and profitability are removed as bank-specific control variables and the coefficient turns positive. Apparently, bank characteristics are of essence to take into account for explaining the differences in bank size. As found by Laeven et al. (2014), large banks do behave differently than domestic banks and depending on characteristics this influences credit growth. This intuition is supported by adding concentration ratio in column (5), that enters the regression negatively and turns the interaction coefficient between ownership and market share also negative. The latter indicates that if a market is highly concentrated, banks in general, but large banks specifically, are likely to reduce credit growth. This result supports the concentration-fragility view and contradicts the results found for the Z-score. However, one should note that although the Z-score and credit supply both measure financial stability, they explain different mechanisms. More specifically, the Z-score explains individual bank soundness. Apparently, a higher concentration ratio improves the balance sheets of all financial institutions. In contrast, credit growth focusses on the transmission of bank stability to the real economy through loan supply. In this line of reasoning, large banks and a concentration ratio but results do not change. In turn, I reformulate the credit growth following different papers (e.g. Cihak and Hesse, 2010; De Haas and Van Lelyveld, 2010), using nonperforming loans to assets or the loan growth to assets. No improvement in the predictive ability is found.
28
have a negative impact on credit growth. This mainly stems from the loan supply of large banks being less stable (Laeven et al., 2014). In addition, with a highly concentrated banking sector and few large banks, their unstable credit supply has a higher impact on credit growth than with many small banks that are willing to provide funding to private agents. In conclusion, this supports the concentration-fragility view.
Table 4
Credit Growth and Foreign Presence (1) Credit Growth (2) Credit Growth (3) Credit Growth (4) Credit Growth (5) Credit Growth Ownership 0.986 1.093 1.020 1.020 0.0582 (0.949) (0.916) (0.894) (0.893) (0.557)
Log (Total Bank Asset Ratio) -0.0194 -0.00231 0.00635 0.00633 -0.0306
(0.0876) (0.0779) (0.0808) (0.0808) (0.0868)
Ownership X Log (Total Bank Asset Ratio) 0.278 0.305 0.285 0.285 0.116
(0.189) (0.183) (0.176) (0.176) (0.0805)
Inflation Rate(t-1) 0.00579 -0.0179 -0.0203 -0.0205 -0.0288
(0.0104) (0.0302) (0.0301) (0.0277) (0.0316)
GDP Growth(t-1) 0.771 -2.224 -2.108 -2.113 -2.520
(0.964) (2.917) (2.916) (2.875) (3.163)
GDP per Capita -0.00000109 -0.000000410 -0.000000120 1.94e-08
(0.00000331) (0.00000369) (0.00000363) (0.00000276) Efficiency 0.106 (0.0949) Liquidity -0.374 -0.397 -0.397 -0.121 (0.266) (0.255) (0.259) (0.132) Profitability -0.0842 (0.0617) Concentration Ratio (5) -0.137 (0.188) Constant -0.00741 1.831 2.136 2.150 2.553 (0.782) (2.327) (2.325) (2.239) (2.763)
Time Fixed Effects Yes Yes Yes Yes Yes
Observations 8389 8800 8791 8791 8204
Adjusted R-squared 0.018 0.005 0.005 0.005 0.004
Clustered standard errors in parentheses, significance levels indicated with *p<.1, **p<.05, ***<.01.
Credit Growth is defnied as the first difference of loans and gross loans, divided by its first lag. A bank is defined foreign owned if more than 50% for the assets are in foreign hands, and is doemstic otherwise. Concentration ratio available from 1997 onwards.
29
4.3 Robustness on bank characteristics
Next, I am interested whether background characteristics of the subsidiaries affect the relationship between foreign presence and financial stability. Two features will be taken into account: firstly, level of development of the home country will be considered and secondly the business model of banks are examined. The baseline model is run for all categories and the results are presented in the Appendix, for both the Z-score and credit growth (table 3 and 4). The setup of the paragraphs is as follows: first the Z-score will be analyzed, followed by credit growth.
Home country characteristics
Specifically, I test whether the level of development of the home country positively affects the relationship between foreign bank presence and financial stability. A distinction is made between low-development and high-development (i.e. OECD countries). In addition, I correct for the geographical distance of high-developed countries, by the categories ‘neighbour’, ‘OECD’ and ‘other’. Specifically, it is expected that foreign subsidiaries from neighbour countries perform better in terms of financial stability than domestic banks (Claessens and Van Horen, 2012;2013). For foreign banks with their origin in other OECD countries the expected sign is ambiguous. For other (non-OECD) countries, I assume a positive linkage between foreign presence and financial stability, as these foreign subsidiaries are likely to be committed to the host market, for historical reasons, and will therefore act more conservative serving their emigrants.
For the Z-score, no statistical significant effects are found for the ownership variables, in contrast to suggestions from earlier research (e.g. Navaretti et al., 2010). However, the ownership coefficient for neighbour countries is found to be negative and is positive for oecd countries. This result contradicts the findings of Claessens and Van Horen (2012;2013) and suggests that subsidiaries from countries that are geographically close by are less financially stable than domestic banks. On interpretation could be that this stems from home bias, where parent banks decide to prioritize home markets rather than stimulate their subsidiaries via the internal capital market. As local shocks are more likely to influence neighbour countries than
30
countries that are not geographically close by, this can explain the suggestive negative impact of foreign players in neighbour countries. For instance, a political or economic event could have a negative impact on the economy in the host country, leading to lower trade with the neighbour country where the parent bank is established, hence creates contagion. Subsequently, the parent bank could decide to decrease activities of the foreign subsidiaries, due to the turmoil and uncertainty in the real economy. In contrast, oecd countries that are not geographically close by will experience less negative effects from a host country shock, and might therefore still support the foreign subsidiary (Claessens and Van Horen, 2012).
For other than high-income countries, no statistically significant, but positive sign is found for foreign bank presence. A closer look at the data indicates that these subsidiaries are active for a longer period of time in the host country. This suggests that these banks act in line with domestic bank behaviour and have little or no direct influence from the home country and its possible risks. Put differently, although these banks come from less developed countries, historical relations with the host country to serve their migrants in the particular market do not seem to influence their style of business anymore. Due to their commitment, they might have a suggestive, positive effect on financial stability.
Estimations for credit growth show that foreign players from both neighbour and oecd countries have a small positive impact in absolute terms and a small negative impact when the market share is taken into account. However, the estimated coefficients are not significant. The small negative effect from foreign market share supports the earlier conclusions, where larger banks are assumed to be more risky. Apparently, larger foreign banks that come from high-developed countries have lower credit growth than their domestic counterparts. This result suggests geographical distance does not have a different impact on credit supply. In contrast, the ownership coefficient for other than high income countries enters the specification positively, but not statistically significant. It seems that if these banks play a more important role in the host country, they are likely to have a higher credit supply than domestic banks. This could stem from the intuition that they are committed to the host market for historical reasons and less sensitive to host country shocks.
Table 5
Home Country Grouped (1) Z-score (2) Credit Growth (3) Z-score (4) Credit Growth (5) Z-score (6) Credit Growth Neighbour -0.181 0.130 (0.954) (0.0828)
Neighbour X Log (Total Bank Asset Ratio) -4.923 -0.00918
(4.942) (0.653)
OECD 3.178 0.120
(10.55) (0.273)
OECD X Log (Total Bank Asset Ratio) 1.080 -0.0253
(1.978) (0.0562)
Other 13.41 0.646
(13.69) (0.512)
Other X Log (Total Bank Asset Ratio) 1.945 0.0942
(2.322) (0.0776)
Log (Total Bank Asset Ratio) -5.021*** -0.0275 -5.392*** -0.00408 -5.168*** -0.0233
(1.280) (0.0386) (1.549) (0.0523) (1.313) (0.0410) Inflation Rate(t-1) -1.035*** 0.00208 -1.028*** 0.00157 -1.036*** 0.00150 (0.303) (0.00776) (0.305) (0.00786) (0.303) (0.00775) GDP Growth(t-1) 78.43*** 0.668 78.82*** 0.611 78.80*** 0.654 (18.56) (0.922) (18.40) (0.914) (18.53) (0.908) GDP per Capita -0.000152** 0.000000609 -0.000155** 0.000000357 -0.000153** 0.000000165 (0.0000540) (0.00000273) (0.0000564) (0.00000266) (0.0000555) (0.00000264) Liquidity -6.873*** -0.105 -6.822*** -0.137 -7.074*** -0.141 (1.985) (0.120) (2.032) (0.119) (2.000) (0.118) Constant 83.30*** -0.205 80.95*** -0.0811 82.45*** -0.145 (23.56) (0.659) (24.35) (0.707) (23.43) (0.647)
Time Fixed Effects Yes Yes Yes Yes Yes Yes
Observations 9429 8778 9429 8778 9429 8778
Adjusted R-squared 0.118 0.028 0.118 0.025 0.118 0.025
Z-score is defined as the Equity and the mean of the Return on Assets, divided by the standard deviation of the Return on Assets. Credit Growth is defined as the first difference of loans and gross loans, divided by its first lag. A bank is defined foreign owned if more than 50% for the assets are in foreign hands, and is domestic otherwise. Foreign ownership specified by the home country of the parent bank. Countries grouped by neighbour countries, OECD contries that are not neighbour countries and other countries (non-OECD).
Clustered standard errors in parentheses, significance levels indicated with *p<.1, **p<.05, ***<.01.
32
Bank business models
According to Claessens and Van Horen (2014), correcting for heterogeneity in business models is of essence to capture differences in risks for financial stability. Therefore, the model is run for the two types of banks (commercial bank and cooperative bank), for both the Z-score and credit growth. Results are reported in table 6.
The estimated coefficients of ownership are negative and do not enter the model significantly for commercial banks, using either the Z-score of credit supply. This indicates that domestic and foreign owned commercial banks do not significantly differ in terms of financial stability. However, the negative coefficient suggests that foreign owned commercial banks are more likely to introduce financial instability than domestic banks. Intuitively, this can be explained by the fact that commercial banks will actively participate in the host market and also be most vulnerable to host country shocks. If the parent bank strategically diversifies risk in the host market and does not support the subsidiary by the internal capital market, the commercial bank is not likely to enhance financial stability in the host market.
In contrast, foreign owned cooperative banks are found to have a positive effect on financial stability and is in line with conclusions of Cihak and Hesse (2007). They show that cooperative banks increase financial stability due to lower volatility of cooperative banks’ returns, which more than offsets the lower profits they make in comparison to commercial banks. This stems from the ability to absorb shocks with the use of customer surplus and conducting a more conservative business model. However, the authors do not correct for foreign ownership. Appearing from the results for the Z-score found in this paper, cooperative banks that engage in cross-border banking seem to act even more conservative than their domestic counterparts. This can stem from the commitment a foreign subsidiary of a cooperative bank could have towards home country customers and do not enter a new market to increase profits in a hazardous manner. The statistically significant sign found for the Z-score can be explained by the fact that the conservative behavior takes mainly form of a higher buffer, which increases the equity ratio and hence the Z-score. Therefore, the credit growth of foreign cooperative banks is not per se higher than for domestic banks.
