T HE I MPACT OF F OREIGN B ANKS ON F INANCIAL
S TABILITY
BY
A SHLEY K LAPWIJK
University of Groningen Faculty of Economics and Business MSc International Economics & Business
June 2017
Parkstraat 35 BS 3581 PD Utrecht +31 (0) 6 48 58 74 70 ashleyklapwijk@gmail.com
Student number: s2022273
ABSTRACT
The global financial crisis reinvigorated the debate on the effect of foreign banks on financial stability. Whereas previous research mainly focused on the effect of foreign banks on financial stability at the bank level, this paper examines the effect of foreign banks on financial stability at the country level using financial soundness indicators proposed by the International Monetary Fund. Financial stability is measured by the regulatory capital to risk- weighted assets (or the capital adequacy ratio), nonperforming loans to total gross loans and return on equity. The sample consists of 38 countries between 2008 and 2013. To take heterogeneity into account, the sample is divided into two subsamples: (1) Latin American and developed countries and (2) Central and Eastern European and developing countries. The results indicate that foreign banks have a different impact on each subsample. This especially holds for the profitability of the banking system. More importantly, policymakers should not focus on the individual financial soundness indicators but look at multiple financial soundness indicators, since this provides a more comprehensive picture.
Keywords: Foreign banks, financial stability, financial soundness indicators
Supervisor: prof. dr. J. (Jakob) de Haan
Co-assessor: dr. M.J. (Michiel) Gerritse
TABLE OF CONTENTS
1 Introduction... p.1 2 Literature review... p.3
2.1 The impact of foreign banks 2.2 Financial stability
3 Data and methodology... p.9 3.1 Data
3.2 Methodology
3.2.1 Regulatory capital to risk-weighted assets 3.2.2 Nonperforming loans to total gross loans 3.2.3 Return on equity
4 Results... p.15 4.1 Regulatory capital to risk-weighted assets
4.2 Nonperforming loans to total gross loans 4.3 Return on equity
5 Sensitivity tests... p.22
6 Conclusion... p.24
7 References... p.27
8 Appendices... p.34
1 1. INTRODUCTION
Financial deregulation and further economic integration have resulted in the globalization of the financial sector. As a result, banks in saturated home countries expanded abroad attracted by higher margins. From 1995 to 2009, the total number of banks remained virtually the same. However, during this period the amount of domestic banks decreased by 18% while at the same time the amount of foreign banks increased by 69%. Hence, between 1995 and 2009 the relative importance of foreign banks increased from a share of 20% to 34% (Claessens &
Van Horen, 2014). Before the global financial crisis (GFC) the consensus was that the benefits of financial integration outweighed the costs. However, the GFC has shown that financial integration can have serious adverse effects. It therefore “reinvigorated the debate on the cost and benefits of financial integration in general and foreign bank ownership in particular” (Claessens & Van Horen, 2013: 1).
On the one hand, foreign banks can have a positive impact on the financial stability of the host country. Claessens, Demirguc-Kunt & Huizinga (2001) find that entry of foreign banks improves the efficiency of the banks in the host markets. Lensink & Hermes (2004) extend this study and find that the effect of foreign banks on domestic banks is conditional on the level of economic development of the host country. Furthermore, when domestic banks are hit by a shock, foreign banks can substitute for domestic banks and hence can be a stabilizing force (Allen, Beck, Carletti, Lane, Schoenmaker & Wagner, 2011). On the other hand, foreign banks may have a destabilizing effect and lead to financial instability. Since foreign banks can expose the host country to shocks from abroad, contagion can increase the likelihood of a systemic crisis. Moreover, foreign capital is likely to be more mobile than domestic capital (Allen et al., 2011). As a result, foreign banks may reduce credit provision in the host country more easily. This retrenchment of foreign capital can lead to financial instability.
Financial stability concerns the resilience of the financial system. Because of the
complex nature of the financial system and its interlinkages, it is very difficult if not
impossible, to construct a single aggregate measure of financial stability (Geršl & Heřmánek,
2006). The International Monetary Fund (IMF) published the Compilation Guide on Financial
Soundness Indicators (IMF, 2006) that consists of two sets of financial soundness indicators
(FSIs), namely the core set and the encouraged set. Although previous literature has studied
the effect of foreign bank on financial stability at the bank level, less research has been done
on the effect at the country level. Önder & Özyildirim (2016) examined the effect of foreign
2 banks in Emerging European countries on the real sector, measured by real GDP growth rate, which is not an FSI in accordance with the indicators proposed by the IMF (2006). Barth, Dopico, Nolle & Wilcox (2002) examined the effect of foreign banks on nonperforming loans (NPLs) and return on equity (ROE) in 1999 whereas Boudriga, Taktak & Jellouli (2009) examined the effect of foreign banks on NPLs over the period 2002-2006. To the author’s knowledge, this paper will be the first to examine the effect of foreign banks on financial stability, measured by various FSIs proposed by the IMF (2006), namely: regulatory capital to risk-weighted assets (or the capital adequacy ratio - CAR), NPLs to total gross loans and ROE (banking sector) at the country level. Each FSI measures a different area of the banking system: CAR measures the ability to absorb losses (how resilient is the banking system to shocks) and hence promotes financial stability. A decrease in CAR may therefore be a signal of an increase in risk exposure and possible capital adequacy problems (Sundararajan et al., 2002). NPLs to total gross loans measure the quality of the bank assets whereas ROE is commonly used to measure the profitability of the banking sector. More importantly, this study is the first to look at the effect of foreign banks on financial stability during the period 2008-2013. Hence, this paper looks at the impact of foreign banks on financial stability during the GFC (2008-2010) and after the GFC (2011-2013). Overall, this paper makes an attempt to combine two streams of literature, namely those on financial stability and foreign banks.
This paper uses a sample of in total 38 countries for which FSIs and foreign bank ownership are available. Previous research found that the impact of foreign banks is not similar across countries (Lensink & Hermes, 2004; Detragiache, Tressel & Gupta, 2008;
Kamil & Rai, 2010; Cull & Martinez Peria, 2013). Therefore, to take heterogeneity into account, the sample is divided into two subsamples: (1) Latin American and developed countries and (2) Central and Eastern European (CEE) and developing countries. The findings suggest that during the GFC, foreign banks reduced the CAR for the pooled dataset and for CEE and developing countries. On the other hand, foreign banks decreased NPLs to total gross loans in CEE and developing countries between 2008 and 2013, implying that foreign banks increased asset quality. Interestingly, foreign banks increased profitability in Latin American and developed countries between 2008 and 2013, whereas foreign banks decreased profitability in CEE and developing countries between 2008 and 2013, and after the GFC.
The rest of the paper is structured as follows. In section 2 the relevant literature is
reviewed. The data and methodology will be described in section 3. The results of the analysis
will be reported in section 4. Section 5 offers a sensitivity analysis, while the conclusions are
provided in section 6.
3 2. LITERATURE REVIEW
The literature review is structured as follows. In section 2.1 the literature on the impact of foreign banks is discussed. Section 2.2 will focus on the literature on financial stability.
2.1 The impact of foreign banks
Deregulation of the financial sector led to reforms and liberalization resulted in the reduction of entry barriers for foreign banks.
1In combination with technological changes and rapid globalization, the global banking sector underwent important transformations. Between 1995 and 2009, the total number of domestic and foreign banks virtually remained the same (Claessens & Van Horen, 2014). However, according to Claessens & Van Horen (2014), these aggregate numbers mask two counteracting trends: from 1995 to 2009 the number of domestic banks decreased, driven by technological changes and deregulation, while the number of foreign banks increased. Hence, the relative importance of foreign banks increased substantially, from a share of 20% in 1995 to 34% in 2009. (Claessens & Van Horen, 2014).
There are however important differences across countries, where the number of foreign banks grew faster in emerging and developing countries in comparison with OECD and high-income countries (Claessens & Van Horen, 2014). Especially for transition economies the expansion of foreign banks has been important (Buch, 1997; Naaborg et al., 2004; Bonin & Schnabel, 2011). The financial turmoil from the GFC however changed the global banking landscape and accelerated structural transformations. Especially American and European banks were affected by the GFC, which had adverse effects on the financial system worldwide. In the wake of the crisis, banks were required to inter alia meet stricter capital requirements and restore their balance sheets, which gave banks an incentive to reduce their international operations (Claessens & Van Horen, 2015). Furthermore, liquidity dried up in the interbank market, which caused funding problems. As a result, there is currently a large variety of banks active on the global market with banks from emerging and developing markets gaining importance (Van Horen, 2011; Beck, Fuchs, Singer & Witte, 2014; Claessens & Van Horen, 2014; Claessens & Van Horen, 2015). According to Claessens & Van Horen (2015), bank ownership by OECD home countries decreased from 89 to 83 percent after the crisis while banks from non-OECD countries more than doubled their presence.
1 In this paper, a bank is considered to be a foreign bank when 50% (or more) of the shares are hold by foreigners (Claessens & Van Horen, 2015).
4 Several studies have looked at the effect of foreign banks on the domestic banking market. Foreign bank presence can have a positive impact in several ways. To begin with, foreign bank presence is found to improve the efficiency of the domestic market by decreasing the costs of financial intermediation (Claessens, Demirguc-Kunt & Huizinga, 2001; Lensink & Hermes, 2004; Berger, Clarke, Cull, Klapper, & Udell, 2005). Furthermore, foreign banks can increase the quality of financial intermediation by lowering loan-loss provisions (Martinez Peria & Mody, 2004). Foreign banks can also be a driver for domestic reforms by stressing governments to increase transparency and strengthen supervision and regulation (Levine, 1996; Mishkin, 2006). Increased competition and technological spillovers are the determinants of these effects, where, according to Lensink & Hermes (2004), the effect is dependent on the level of economic development. Lensink & Hermes (2004) find that at lower levels of economic development, foreign bank entry raises margins of the domestic banks in the short run. They argue that the difference in the development between foreign and domestic banks leads to strong spillover effects. However, catching up comes with costs and since domestic banks still have relatively strong domestic market power, they are able to increase their margins while maintaining market share. Hence, the effect of an increase in competition is canceled out. Spillover effects are however less relevant in markets that are more developed since the development gap between foreign and domestic banks is smaller.
Subsequently, margins decrease since competition forces domestic banks to reduce costs, which increases efficiency. Not only a limited economic development but also tighter regulations on bank entry can hamper the effectiveness of foreign banks (Demirguc-Kunt, Laeven, & Levine, 2004, Garcia Herrero & Martinez Peria, 2007). Claessens & Lee (2003) look at the effect of foreign banks in low-income countries on the competitiveness and performance of the domestic banking system. They found that the positive effect of foreign banks in low-income countries is stronger in countries where foreign bank presence is already high. Hence, there may be a threshold that determines the effectiveness of foreign banks.
Furthermore, they suggest “it is the new entry of foreign banks that leads domestic banks to reduce their costs and lower their margins, while a greater presence by foreign banks lowers overall profitability in the domestic banking system, provided that the system is contestable”
(Claessens & Lee, 2003: 133). All in all, research has shown that the positive effect of foreign banks on the domestic market tends to be dependent on certain conditions.
Before the GFC, the benefits were assumed to outweigh the costs of foreign bank
presence. However, the GFC reinvigorated this debate and assessing the risks of foreign bank
presence is therefore an important topic for research and policymakers. Starting point for this
5 discussion is the premise that foreign banks introduce financial instability due to i) the exposure to contagion and ii) the withdrawal of liquidity from foreign subsidiaries by the parent bank. Claessens & Van Horen (2012) found that during the GFC foreign banks indeed reduced credit more than domestic banks, except when foreign banks had a dominant role.
2Furthermore, De Haas & Lelyveld (2014) found that parent banks during the GFC were not a significant source of strength to their subsidiaries. In comparison with domestic banks, De Haas & Lelyveld (2014) found that multinational subsidiaries had to slow down credit growth almost three times faster, controlling for other bank characteristics. Notwithstanding, heterogeneity should be taken into account since the impact of foreign banks is not similar across countries. Before the GFC greenfield foreign banks provided credit stability in Central and Eastern Europe particularly (De Haas & Lelyveld, 2006). During the GFC however foreign banks reduced credit earlier and faster than domestic banks in Eastern Europe (De Haas et al., 2012). On the contrary, banks that participated in the Vienna Initiative - an agreement to provide macroeconomic stability in Central and Eastern Europe by ensuring that foreign subsidiaries were sufficiently capitalized - were stable lenders (De Haas, Korniyenko, Loukoianova, & Pivovarsky, 2012). In Central Europe, multinational subsidiaries also reduced credit faster than domestic banks (Popov & Udell, 2012; De Haas & Van Horen, 2013). More importantly, emerging Europe was hit severely by the GFC due to cross-regional contagion effects, since the banking sector relied heavily on cross-border lending (Canales- Kriljenko, Coulibaly & Kamil, 2010; Árvai, Driessen & Ötker-Robe, 2010). Using a sample of 20 Emerging European countries (1998-2013), Önder & Üzylilirim (2016) found that the impact of foreign banks on macroeconomic volatility only temporarily had an adverse effect during the GFC. Furthermore, they found that after the crisis and during normal times foreign banks seem to help Emerging European countries stabilize macroeconomic volatility. Hence, in Central and East Europe foreign banks led to financial instability during the GFC (De Haas et al., 2012). However, according to Önder & Üzylilirim (2016), the adverse effect was only temporarily and foreign banks ensured financial stability after the GFC.
Latin American countries, on the other hand, were not hit that severely by the GFC and did considerably better than CEE countries (Kamil & Rai, 2010; Cull & Martinez Peria, 2013). Kamil & Rai (2010) showed that during the GFC, Latin America did better than e.g.
Central and Eastern Europe because lending by foreign banks was mostly denominated in domestic currency and funded from a domestic deposit base. More importantly, “Latin
2 In the study done by Claessens & Van Horen (2012), foreign banks dominate when they hold over 50 percent of all bank assets in the country.
6 America's lower reliance on foreign banks‘ cross-border credit made it less exposed to the risk of a homeward flow of foreign banks‘ assets” (Kamil & Rai, 2010: 14). Hence, in contrast with CEE countries, Latin American countries were mostly funded by local lending activities. According to Garcia Herrero & Martinez Peria (2007), local lending activities tend to me more stable and less responsive to negative shocks than cross-border claims, since local lending activities requires paying higher fixed and irreversible costs. Moreover, they argue that banks which try to shrink the size or close down their foreign affiliates, will have to pay the reputational costs of doing so and, therefore, may be less likely “to run” when economic conditions deteriorate. Surprisingly, almost no research has been done on the impact of foreign banks on developed countries. Detragiache, Tressel & Gupta (2008) found adverse effects of foreign bank presence in poor countries but found no such effect in advanced countries. Lensink & Hermes (2004) argue that foreign bank presence leads to efficiency gains in countries that are more developed since competition decreases costs. Schaeck, Čihák
& Wolfe (2009) found that more competitive banking systems are less likely to experience a systemic crisis. However, they look at the period 1980-2003 and hence do not look at the GFC. Since developed countries were significantly exposed to contagion during the GFC, because both domestic banks and foreign banks were highly integrated in the global banking market, these results likely do not hold for the period 2008-2010.
To summarize, literature has shown that foreign banks can have a positive effect on the domestic banking market but this effect is dependent on certain conditions. The GFC fundamentally changed the global banking market and reinvigorated the debate on the effect of foreign banks on financial stability. Authors have shown that certain countries were hit more vigorously than others and that foreign banks do not ensure financial stability per se.
Thus, heterogeneity should be taken into account. This leads to the following hypotheses:
Hypothesis 1: Foreign banks have a positive impact on financial stability after the GFC.
Hypothesis 2: During the GFC the impact of foreign banks on financial stability is positive for (i) Latin American countries and (ii) developed countries.
Hypothesis 3: During the GFC the impact of foreign banks on financial stability is negative
for (i) CEE countries and (ii) developing countries.
7 2.2 Financial stability
Financial stability concerns the resilience of the financial system and has become an important topic for policy makers. However, “unlike price stability, financial stability is not easy to define or measure given the interdependence and the complex interactions of different elements of the financial system among themselves and with the real economy” (Gadanecz &
Jayaram, 2009: 365). As such, Sundararajan et al. (2002) proposed two sets of FSIs, namely the core set (including indicators related to the banking sector) and the encouraged set (including additional indicators related to the banking sector as well as indicators related to other relevant markets, like the real estate and corporate sector). Subsequently, these were laid down in the Compilation Guide on Financial Soundness Indicators (IMF, 2006). “The primary purpose of the Guide is to provide guidance on the concepts and definitions, and sources and techniques, for the compilation and dissemination of the FSIs. The Guide is intended to encourage compilation of FSIs and promote cross-country comparability of these data, as well as assist compilers and users of FSI data, for the purpose of supporting national and international surveillance of financial systems” (IMF, 2006: 1). By definition, the IMF puts substantial effort in the comparability of the FSIs so that they can be used for cross-country comparisons. According to the IMF (2009), FSIs are mostly backward looking. For this research, however, this is not considered a drawback since this paper does not focus on FSIs as early warning indicators but focuses on cross-country analyses.
Since there are numerous FSIs that are designed to measure a specific factor regarding a specific sector, not all FSIs can be used to measure the impact of foreign banks on financial stability. For example, FSIs that focus on the household sector are less relevant for measuring bank stability. Önder & Üzylilirim (2016) look at the effect of foreign bank presence in Central and Eastern Europe at the country level, but measure financial stability in terms of macroeconomic volatility (measured in real GDP growth and its components consumption and investments), which is not a FSI as indicated by the IMF (2006) but a macroprudential indicator used by the European Central Bank (Mörttinen, Poloni, Sandras & Vesala, 2005).
To measure the impact of foreign bank presence on financial stability, it is required that the
FSIs used measure the health of the banking system correctly. This sounds rather trivial, but
literature on early warning indicators has shown that not all FSIs are as effective since they
are not all able to detect banking crises. Čihák & Schaeck (2007) examined how well
aggregate bank ratios identify banking crises for a dataset of 100 countries for the period
1994-2004. They use a multivariate logit model in which they test if (lagged) FSIs can
correctly predict a banking crisis. The FSIs included in their study are CAR, NPLs to total
8 loans, ROE (banking sector), ROE (corporate sector) and the debt to equity ratio (corporate sector). As acknowledged by the authors, the FSIs used deviate from the FSIs proposed by the IMF. Therefore, the FSIs may not be fully comparable across countries due to inter alia differences in definitions. The results show that not all FSIs are able to detect turmoil in the banking sector. ROE of banks (irrespectively of the lagged structure) and corporate leverage are robust indicators of banking vulnerabilities, where the first also serves as an indicator of the timing of the crisis (Čihák & Schaeck, 2007). The NPLs to total loans and the CAR are significant at the 10 percent significance level. On the other hand, ROE in the corporate sector is not a good indicator for predicting banking crises. Following the work of Čihák & Schaeck (2007), Costa Navajas & Thegeya (2013) test the effectiveness of FSIs as indicators of potential banking crises for a dataset of 80 countries for the period 2005-2012 by running reduced-form models. Hence, they look at the correlations between FSIs and banking crises and do not determine a causal relationship. They are the first to used FSIs collected by the IMF following the Compilation Guide on Financial Soundness Indicators (IMF, 2006).
Hence, by using these FSIs the authors are able to make cross-country comparisons. The FSIs included in their study are CAR, NPLs net of provisions to capital, NPLs to total loans, ROE (banking sector), interest margin to gross income, and noninterest expenses to gross income.
Similar to Čihák & Schaeck (2007) they find that especially CAR and ROE are useful early warning indicators for banking crises, where the lagged ROE for banks is a leading crises indicator. Furthermore, they found mixed evidence for NPLs to total loans and noninterest expenses to gross income. The authors however find no support for NPLs net provisions to capital and interest margin to gross income.
To the author’s knowledge, only two studies examined the effect of foreign banks on NPLs and ROE (banking sector). To begin with, Barth et al. (2002) studied the connections between the structure of bank supervision and the safety and soundness of banking systems for 70 countries in 1999. To control for the market structure of the banking system, they included foreign banks, measured by the percentage of foreign bank assets among total bank assets, as an explanatory variable. Barth et al. (2002) found that foreign banks raise asset quality by decreasing NPLs. Moreover, they found that foreign banks have a positive impact on ROE. On the contrary, they find no significant relationship between foreign banks and management quality, measured by the ratio of overhead costs to assets. Boudriga, Taktak &
Jellouli (2009) look at the cross-country determinants of NPLs. They are specifically
interested in the effect of the regulatory framework and supervision on NPLs. Boudriga,
Taktak & Jellouli (2009) used aggregate data on NPLs on FSIs from the IMF for 59 countries
9 over the period 2002-2006. Following Barth et al. (2002), they included the percentage of foreign bank assets among total bank assets as an explanatory variable. Their results confirm the findings of Barth et al. (2002), namely that foreign banks reduce NPLs and hence improve financial stability. However, they acknowledge that the GFC calls into question the positive impact of impact of foreign banks on financial stability.
To summarize, previous studies have not yet examined the impact of foreign banks on financial stability at the country level, measured by various FSIs proposed by the IMF (2006).
More importantly, in contrast with previous studies (Barth et al., 2002; Boudriga, Taktak &
Jellouli, 2009), this paper studies the period between 2008-2013, thereby examining the impact of foreign banks on financial stability during the GFC (2008-2010) and after the GFC (2011-2013). To measure the effect, the FSIs used should reflect the health of the banking system. Following the results of Barth et al. (2002), Čihák & Schaeck (2007), Boudriga Taktak & Jellouli (2009) and Costa Navajas & Thegeya (2013), CAR, ROE (banking sector) and NPLs to total gross loans are used as measures of financial stability.
3. DATA AND METHODOLOGY
3.1 Data
This paper uses the annual FSIs compiled by the IMF. Although the Compilation Guide on Financial Soundness Indicators (2006) provides guidance for the compilation of the FSIs, reporting countries do not always use the same methodology and hence different methodologies are used, which may impede cross-country analyses. This means that countries may use different definitions of, for example, regulatory capital to risk-weighted assets, which can have an impact on the results. Despite this potential drawback, the data is of high quality.
Moreover importantly, the database is currently the best available dataset for cross-country comparisons on the health of the financial system. Most countries started reporting FSIs from 2008 onwards. However, there are some countries that started earlier. As of April 2017, 114 countries report FSIs on a regular basis.
3The bank ownership database compiled by Claessens & Van Horen (2015) is used for measuring the percentage of foreign bank assets among total bank assets. The database contains information on bank ownership for 5,498 banks active in 139 countries over the period 1995-2013.
4As indicated in appendices 1 and 2, there are countries that are in the FSI
3 See appendix 1 for an overview of the reporting countries.
4 See appendix 2 for an overview of the included countries.
10 dataset but not in the bank ownership database and vice versa. For the analysis, the FSIs as well as the foreign bank assets among total bank assets are necessary. Consequently, countries that are in the FSI dataset but missing in the bank ownership database or vice verse are excluded from this study.
5In total, there are 94 countries that are in both databases. However, as shown in appendix 1, several countries started reporting FSIs later than 2008. Furthermore, there are countries that only report on a quarterly and/or monthly basis.
6Subsequently, these countries are also excluded. After deleting countries with missing values for more than two years, 38 countries remain over the period 2008-2013 (see table 1).
7This results in a dataset of 228 country-year observations.
TABLE 1 Countries in Sample
Latin American and developed countries CEE and developing countries
Argentina Armenia
Australia Bosnia and Herzegovina
Brazil Bulgaria
Chile Indonesia
Colombia Kenya
Costa Rica Latvia
Honduras Macedonia
Hong Kong Malaysia
Mexico Mexico
Panama Nigeria
Paraguay Panama
Seychelles Paraguay
Singapore Philippines
South Africa Romania
South Korea Russian Federation
Switzerland Rwanda
Trinidad and Tobago Thailand
Uruguay Uganda
Ukraine Vietnam
Total: 18 countries Total: 20 countries
5 Countries that are in the bank ownership database (Claessens & Van Horen, 2013) but missing in the FSI dataset are: Albania, Antigua and Barbuda, Azerbaijan, Bahrain, Barbados, Benin, Botswana, Burkina Faso, Cote D'Ivoire, Cuba, Dominican Republic, Egypt, Ethiopia, Haiti, Iceland, Iran, Jamaica, Jordan, Kuwait, Libya, Madagascar, Malawi, Mali, Mauritania, Mongolia, Montenegro, Morocco, Mozambique, Nepal, New Zealand, Nicaragua, Niger, Oman, Qatar, Senegal, Serbia, Sudan, Taiwan, Togo, Tunisia, Venezuela, Yemen and Zimbabwe.
Countries that are in the FSI dataset but missing in the bank ownership database (Claessens & Van Horen, 2013) are: Afghanistan, Bhutan, Brunei Darussalam, Comoros, Gambia, Guinea, Kosovo, Maldives, Malta, Micronesia, Papua New Guinea, Samoa, San Marino, Solomon Islands, Tajikistan, Tonga, Vanuatu and West Bank and Gaza.
6 Countries that report quarterly and/or monthly all started reporting later than 2008. Hence, the FSIs cannot simply be aggregated.
7 If data are missing between years, the average is calculated (e.g. if 2009 is missing, the average of 2008 and 2010 is used). If data are missing for 2008, data from 2009 is used and lastly if data are missing for 2013, data from 2012 is used.
11 Previous research found that the impact of foreign banks is not similar across countries (Lensink & Hermes, 2004; Detragiache, Tressel & Gupta, 2008; Kamil & Rai, 2010; Cull &
Martinez Peria, 2013). Therefore, to take heterogeneity into account, the sample is divided into two subsamples, namely (1) Latin American and developed countries and (2) CEE and developing countries.
The descriptive statistics of the dependent and independent variables can be found in
appendices 4 to 9. Notably, the CAR is above average in CEE countries, suggesting that the
banking system, on average, in CEE countries is more resilient, which is surprising since CEE
countries were severely hit by the GFC. On the contrary, the CAR in Hong Kong is rather
low. NPLs to total gross loans are especially high in Rwanda, Seychelles and Trinidad and
Tobago, implying that asset quality is low in these countries. The ROE in Seychelles,
Argentina and Paraguay is above average; hence the banking systems in these countries are,
on average, more profitable. On the other hand, the ROE in Ukraine and Latvia is negative,
indicating that in these countries the overall banking system is making a loss. Furthermore,
appendices 6 to 8 show that the CAR, NPLs to total gross loans and ROE vary tremendously
in Nigeria between 2008 and 2013. On average, the percentage of foreign bank assets among
total bank assets is 45%. However, for Estonia and Hong Kong have a percentage the
percentage of foreign bank assets among total bank assets is more than 90%, while for
Vietnam, Switzerland and Australia the percentage of foreign bank assets among total bank
assets is below 10%. Indicated by appendix 9, foreign banks increased their presence
significantly in Uruguay while foreign banks decreased their presence significantly in
Ukraine. In Seychelles, the largest three banks dominate the banking system since bank
concentration is 100%. Furthermore, the banking system is highly concentrated in Trinidad
and Tobago, Chile and Estonia while Russia has the least concentrated banking system. On
average, GDP growth between 2008 and 2013 in Latvia, Estonia, Hungary and Ukraine was
negative. Lastly, the unemployment rate in especially South Africa, Bosnia and Herzegovina
and Macedonia is high.
12 3.2 Methodology
3.2.1 Regulatory capital to risk-weighted assets
Building on the work of Babihuga (2007) and Kasselaki & Tagkalakis (2013), the following fixed effects model with panel robust standard errors is estimated:
8!"#
$%= '
$+ '
)*+,-./0
$%+ '
1!+02-03,43.+0
$%+ '
5!+67-3.3.+0
$%+ '
8!,-9.3
$%+
'
:;
$%+ =
%+ >
$%!"#
$%is the capital adequacy ratio (regulatory capital to risk-weighted assets) for country i at year t. '
$is the unobserved country effect and *+,-./0
$%is the percentage of foreign bank assets among total bank assets for country i at year t. To control for the structure of the banking system the percentage of assets held by the three largest commercial banks as a share of total commercial banking assets for country i at year t, !+02-03,43.+0
$%, is included.
According to Boyd & de Nicoló (2005) bank concentration is a proxy for (the lack of) competition. Beck, Demirgüc-Kunt & Levine (2006) found that crises are less likely in economies with more concentrated banking systems. On the other hand, de Nicoló, Bartholomew, Zaman & Zephirin (2005) found the opposite, namely that crises are more likely in economies with more concentrated banking systems. Claessens & Laeven (2004) argue that competition and bank concentration are two distinctive variables, which both measure different aspects of the banking system. Schaeck, Cihak & Wolfe (2009) find that this is indeed the case, implying that bank concentration cannot be used as a proxy for competition. To be specific, Schaeck, Cihak & Wolfe (2009) find that “more competitive banking systems are less prone to experience a systemic crisis and exhibit increased time to crisis while a more concentrated banking system is associated with higher probability of a crisis and shorter time to crisis” (p.1). Diallo (2015) also finds that banking systems which are more competitive, measured by the Boone indicator, are less stable. Hence, to control for the effect of banking competition, the Boone indicator for country i at year t, !+67-3.3.+0
$%, is included as explanatory variable. Following Demirgüç-Kunt & Detragiache (1998a), Demirgüç-Kunt & Detragiache (1998b) and Kasselaki & Tagkalakis (2013), !,-9.3
$%is added as an explanatory variable, which measures the financial development of country i at year at year t, measured by the domestic credit to private sector as a percentage of GDP.
8 According to the Hausman test, a fixed effects model is preferred. Robust standard errors are used to correct for heteroscedasticity.
13 According to Phill & Pradhan (1997), the ratio of private sector credit to GDP best captures the extent to which financial deepening has progressed.
;
$%are the macroeconomic variables included, namely the annual GDP growth, the GDP deflator based inflation rate, the change in real effective exchange rate (REER), the real interest rate and the deposit rate.
9The annual GDP growth measures the economy’s growth.
Research has found that CAR increases, when economic conditions worsen (Wong, Choi &
Fong, 2005; Babihuga, 2007; Kasselaki & Tagkalakis, 2013). In other words, during an economic downturn the quality and the value of the assets deteriorates and as a responds banks either increase capital or decrease their assets in order to fulfill the minimum CAR requirement (Babihuga, 2007; Kasselaki & Tagkalakis, 2013).
10The effect of the inflation rate and the REER on CAR is ambiguous. Babihuga (2007) found a negative and statistically significant relationship, possibly due to the erosion of bank capital, while Kasselaki &
Tagkalakis (2013) found no significant relationship. Furthermore, the impact of the REER depends on the share of banking systems assets held abroad (Babihuga, 2007; Kasselaki &
Tagkalakis, 2013). Babihuga (2007) found a negative relationship between CAR and the REER whereas Kasselaki & Tagkalakis (2013) did not find a significant relationship. The real interest rate is expected to have a negative impact on CAR, since an increase in real interest rate increases the fragility of the banking system (Hardy & Pazarbaşioğlu, 1998; Demirgüç- Kunt & Detragiache, 1998a; Demirgüç-Kunt & Detragiache, 1999). More importantly, the asset side of the bank’s balance sheet is likely to deteriorate when the real interest rate is high, since default risk of the borrowers increases and hence the borrower’s ability to repay their loan. According to Čihák (2004) “the impact of capital adequacy is different for lending rates and spreads. Banks with higher capital adequacy have lower lending rates, but they have even lower deposit rates, so that their spreads are higher than in banks with lower capital adequacy” (p.20). Hence, the relationship between in deposit rate and CAR is expected to be negative. Lastly, =
%corrects for the unobserved year effect
11and >
$%is the idiosyncratic error term.
9 See appendix 3 for the variable definitions.
10 As pointed out by Babihuga (2007), without access to the metadata from which FSIs are calculated, it is impossible to test if GDP has an effect on capital (numerator of CAR) or on assets (denominator of CAR).
11 Time-fixed effects are only included if the coefficients for all year dummies are not equal to zero.
14 3.2.2 Nonperforming loans to total gross loans
Following the work of Babihuga (2007), Boudriga, Taktak & Jellouli (2009), Kasselaki &
Tagkalakis (2013), the following fixed effects model with panel robust standard errors is estimated:
12?@A
$%= '
$+ '
)*+,-./0
$%+ '
1!+02-03,43.+0
$%+ '
5!+67-3.3.+0
$%+ '
8!"#
$%+ '
:#B"
$%+ '
C!,-9.3
$%+ '
D;
$%+ =
%+ >
$%?@A
$%is the nonperforming loans to total gross loans of country i at year t. '
$is the unobserved country fixed effect and *+,-./0
$%is the percentage of foreign bank assets among total bank assets for country i at year t. To control for the banking system’s structure,
!+02-03,43.+0
$%and !+67-3.3.+0
$%are included. !"#
$%is the capital adequacy rate of country i at year t. Regulatory and supervisory bodies stress the importance of capital stringency since capital requirements absorb losses and hence behave as a cushion to ensure solvency. Therefore, the relationship between CAR and NPLs to total gross loans is expected to be negative. Moreover, #B"
$%is the return on assets (aggregated on the country level) of country i at year t. Profitability, measured by ROA, is expected to have a negative effect on NPLs to total gross loans meaning that an increase in bank performances reduces NPLs to total gross loans. !,-9.3
$%captures the financial development and ;
$%are the macroeconomic variables included, namely the annual GDP growth, the GDP deflator based inflation rate, the change in real effective exchange rate, the real interest rate and the unemployment rate.
4When economic conditions improve, the ratio of NPLs to total gross loans decreases.
Therefore, GDP has a negative impact on the dependent variable. More importantly, when real interest rate increases and hence borrower default risk increases, NPLs to total gross loans increases since asset quality deteriorates. On the other hand, the effect of REER and the inflation rate on the NPLs to total gross loans is ambiguous (Babihuga, 2007; Kasselaki &
Tagkalakis, 2013). The unemployment rate affects the ability of the borrower to repay their loan(s). Rinaldi & Sanchis-Arellano (2006) and Babihuga (2007) all found a negative relationship between the unemployment rate and NPLs: if the unemployment rate is high, borrowers are more likely to default on their loans. Hence, the expected relationship between the unemployment rate and NPls to total gross loans is negative. Lastly, =
%corrects for the unobserved year effect
5and >
$%is the idiosyncratic error term.
12 According to the Hausman test, a fixed effects model is preferred. Robust standard errors are used to correct for heteroscedasticity.
15 3.2.3 Return on equity
Building on the work of Babihuga (2007) and Kasselaki & Tagkalakis (2013), the following fixed effects model with panel robust standard errors is estimated:
13#BE
$%= '
$+ '
)*+,-./0
$%+ '
1!+02-03,43.+0
$%+ '
5!+67-3.3.+0
$%+ '
8?@A
$%+ '
:!,-9.3
$%+ '
C!+F3F
$%+ '
D;
$%+ =
%+ >
$%#BE
$%is the return on equity (aggregated at the country level) of country i at year t. '
$is the unobserved country effect and *+,-./0
$%is the percentage of foreign bank assets among total bank assets for country i at year t. To control for the banking system’s structure,
!+02-03,43.+0
$%and !+67-3.3.+0
$%are included. ?@A
$%is the nonperforming loans to total gross loans of country i at year t. Babihuga (2007) includes NPL as a proxy for the risk exposure of the banking system and finds that higher NPLs, on average, decrease the profitability of the banking system. On the other hand, Kasselaki & Tagkalakis (2013) are not able to replicate the results of Babihuga (2007). !,-9.3
$%captures the financial development and !+F3F
$%are the bank overhead costs of country i at year t. The relationship between ROE and bank overhead costs is expected to be negative since higher costs reduce profitability. ;
$%are the macroeconomic variables included, namely the annual GDP growth, the GDP deflator based inflation rate, the change in real effective exchange rate, and the real interest rate.
14Lastly, =
%corrects for the unobserved year effect
15and >
$%is the idiosyncratic error term.
4. RESULTS
Tables 2 to 4 shows the results of the regression models. The tables report the results for each time period, namely 2008-2013 (models 1, 4 and 7), 2008-2010 (during the GFC - models 2, 5 and 8) and 2011-2013 (after the GFC - models 3, 6 and 9). Furthermore, the regressions are run for the pooled dataset (subsamples 1 and 2 - columns 2, 5 and 8), subsample 1 (Latin America and developed countries - columns 3, 6 and 9) and subsample 2 (CEE and developing countries - columns 4, 7 and 10).
13 According to the Hausman test, a fixed effects model is preferred. Robust standard errors are used to correct for heteroscedasticity.
14 See appendix 3 for the variable definitions.
15 Time-fixed effects are only included if the coefficients for all year dummies are not equal to zero.
16 4.1 Regulatory Capital to Risk-Weighted Assets
Table 2 shows the regressions results for the CAR. To begin with, I do not find a significant relationship between the CAR and foreign bank participation (foreign) between 2008 and 2013 (model 1 column 2) for the pooled dataset and neither for both subsamples (columns 3 and 4). Model 1 shows a negative and statistically significant relationship for the pooled dataset between GDP and the CAR, similar to Wong, Choi & Fong (2005), Babihuga (2007) and Kasselaki & Tagkalakis (2013). The result, however, is only significant at the 10% level.
Looking at the results per subsample, the relationship between GDP and CAR is only statistically significant for CEE and developing countries and not for Latin American and developed countries. Hence, on average, banking systems have a higher CAR when economic growth is low. Similar to GDP, the coefficients for the inflation rate are only significant for the pooled dataset and CEE and developing countries. For model 1, I do not find a significant relationship between the CAR and the other explanatory variables.
During the GFC (model 2), foreign bank participation, on average, results in a lower CAR for the pooled dataset: if foreign bank participation decreased by 1%, the CAR increased by 0.11% (column 5). For Latin American and developed countries, the relationship is not statistically significant, although the sign is also negative. Interestingly, the impact of foreign bank participation is especially strong in CEE and developing countries: if foreign bank participation decreased by 1%, the CAR increased by 0.19%. Hence, the results indicate that, on average, foreign bank participation in CEE and developing countries led to financial instability and a less resilient banking system. The results, however, have to be interpreted with care. Between 2008 and 2010, the banking system as a whole was unstable.
Subsequently, banks had to use their capital to absorb losses. Hence, the decrease in the CAR might imply that capital was used to absorb losses and hence promote financial stability instead of financial instability. As long as the CAR is above the required minimum capital ratio, a decrease in CAR does not necessarily imply that the banking system is unstable.
Contrary to model 1, model 2 finds a negative and statistically significant relationship between the inflation rate and the CAR and the real interest rate and the CAR for Latin American and developed countries (column 6) but not for CEE and developing countries (column 7). Similar to model 1, the inflation rate and the CAR are negatively related for the pooled dataset.
The coefficient for foreign banks is not significant in model 3. Although not
significant, note that, in comparison with model 2, the signs switch from negative to positive
17
TABLE 2
Determinants of the Capital Adequacy Ratio
2008-2013 (model 1) 2008-2010 (model 2) 2011-2013 (model 3)
Column 2 3 4 5 6 7 8 9 10
Variables CAR CAR (1) CAR (2) CAR CAR (1) CAR (2) CAR CAR (1) CAR (2)
Foreign (%) -0.0288 0.0442 -0.0750 -0.110** -0.0481 -0.193*** 0.00495 0.0141 0.0275
(0.0395) (0.0279) (0.0573) (0.0478) (0.0782) (0.0603) (0.0193) (0.0148) (0.0594)
Concentration (%) -0.0140 0.0453 -0.0111 -0.0172 -0.00703 0.0613 -0.00509 0.00991 -0.00644
(0.0172) (0.0315) (0.0206) (0.0374) (0.0410) (0.0591) (0.0126) (0.0209) (0.0299)
Competition 0.825 0.905 0.356 2.859 4.085 3.217 0.532 4.570 -0.0823
(1.170) (1.793) (1.079) (2.020) (4.616) (2.931) (0.778) (3.442) (1.388)
Credit -0.0547 -0.0332 -0.0852 -0.0699 -0.0318 -0.101 -0.0187 0.0425*** -0.0774
(0.0334) (0.0294) (0.0510) (0.0585) (0.0460) (0.0616) (0.0266) (0.0142) (0.0480)
GDP (%) -0.104* 0.120 -0.221** -0.0921* -0.0422 -0.192 -0.0478 -0.0745* 0.00156
(0.0594) (0.129) (0.104) (0.0503) (0.0590) (0.135) (0.0381) (0.0388) (0.0815)
Inflation (%) -0.230*** -0.248 -0.311** -0.240*** -0.652** -0.334 0.155*** 0.106* 0.357
(0.0459) (0.155) (0.130) (0.0413) (0.242) (0.198) (0.0456) (0.0550) (0.225)
REER (%) 0.0102 -0.0105 0.0109 0.0155 -0.0452 0.0443 0.0283 0.0517* 0.0327
(0.0219) (0.0230) (0.0456) (0.0309) (0.0455) (0.0601) (0.0217) (0.0295) (0.0341)
Interest (%) -0.170** -0.163 -0.330 -0.195*** -0.494** -0.355 0.0911** 0.0310 0.397
(0.0706) (0.118) (0.223) (0.0715) (0.195) (0.334) (0.0346) (0.0361) (0.270)
Deposit (%) -0.138 -0.0308 -0.0423 -0.0447 0.358 -0.120 -0.304** -0.260 -0.357**
(0.0933) (0.197) (0.125) (0.151) (0.407) (0.242) (0.142) (0.196) (0.133)
Constant 26.10*** 15.31*** 31.20*** 30.82*** 27.13*** 34.94*** 18.41*** 12.65*** 18.42***
(3.496) (3.587) (5.310) (4.778) (6.335) (7.473) (2.110) (2.025) (4.706)
Year FE No Yes No No No No Yes Yes No
Observations 228 108 120 114 54 60 114 54 60
R-squared 0.316 0.415 0.411 0.354 0.409 0.457 0.207 0.497 0.209
Number of Countries 38 18 20 38 18 20 38 18 20
Note: Significance at ***1%, **5% and *10% levels respectively. The panel robust standard errors are reported in parentheses. The dependent variable is the capital adequacy ratio. Detailed descriptions and sources of the independent variables can be found in appendix 3. Columns 2, 5 and 8 show the results for the pooled dataset.
Columns 3, 6 and 9 show the results for Latin American and developed countries (subsample 1). Lastly, columns 4, 7 and 10 show the results for Central and Eastern European and developing countries.
18 for each regression (columns 8, 9 and 10). Notably, the relationship between the inflation rate and the CAR, and the real interest rate and the CAR after the GFC is positive for the pooled dataset, whereas the relationship was negative during the GFC. However, the results for the inflation rate have to be interpreted with care, as there is not much variation in the inflation rate after 2010.
16The same holds for the real interest rate.
4.2 Nonperforming loans to total gross loans
Table 3 reports the regressions results for NPLs to total gross loans. The results indicate that foreign banks reduced NPLs to total gross loans and hence improve asset quality for CEE and developing countries between 2008 and 2013 (column 4). Although I find similar results as Barth et al. (2002) and Boudriga, Taktak & Jellouli (2009), namely a negative effect between foreign banks and NPLs (to total gross loans), the countries in the samples differ. Barth et al.
(2002) use data for 70 countries across developed, emerging and transition economies in 1999. They find that foreign banks improve asset quality for each subsample and the pooled dataset.
17Boudriga, Taktak & Jellouli (2009) employ a dataset of 59 developed and developing countries between 2002-2006. When Boudriga, Taktak & Jellouli (2009) divide their sample into developed and developing countries, they find opposite results: foreign banks worsens asset quality in developed countries whereas foreign banks strengthen asset quality in developing countries. Furthermore, they find that foreign banks reduce NPLs for the pooled dataset. I only find a significant result for CEE and developing countries and not for the pooled dataset, which would be more similar to the composition of the dataset used by Barth et al. (2002) and the total sample used by Boudriga, Taktak & Jellouli (2009). However, I do not find a significant relationship between foreign banks and NPLs to total gross loans during and after the GFC. Surprisingly, model 4 finds opposite results for the effect of the CAR on NPLs to total gross loans for Latin American and developed countries and CEE and developing countries, where the latter is found to be negative (at the 1% level) and the former is found to be positive (at the 5% level). This result implies that for CEE and developing countries, the CAR serves as a cushion to absorb losses. Furthermore, the CAR might be able to mitigate banks excessive risk taking (Sinkey & Greenawalt, 1991; Fries, Neven &
Seabright, 2002; Boudriga, Taktak & Jellouli, 2009).
16 Because the fixed effects model depends on within-country variation, a reasonable amount of variation for each explanatory variable is needed. It is difficult to assess the effect of the explanatory variables if there is little within-country variation. This is a known limitation of the fixed effects model.
17 Barth et al. (2002) split their sample based on the structure of the bank supervision. Thus, their subsamples are not comparable to the subsamples that I am using.
19
TABLE 3
Determinants of Nonperforming Loans to Total Gross Loans
2008-2013 (model 4) 2008-2010 (model 5) 2011-2013 (model 6)
Column 2 3 4 5 6 7 8 9 10
Variables NPL NPL (1) NPL (2) NPL NPL (1) NPL (2) NPL NPL (1) NPL (2)
Foreign (%) -0.121 0.000113 -0.287** -0.0259 0.00997 -0.127 -0.0175 -0.0131 -0.0456
(0.0790) (0.0119) (0.132) (0.0823) (0.0521) (0.165) (0.0193) (0.00811) (0.0923)
Concentration (%) 0.0362 0.0251 0.0448 -0.00808 0.00519 0.0551 -0.0236* -0.0157 -0.0741**
(0.0350) (0.0179) (0.0312) (0.0768) (0.0220) (0.140) (0.0125) (0.0180) (0.0269)
Competition 0.0291 -0.488 1.542 5.785** 0.795 10.98*** -0.739 -4.765* -1.120
(2.744) (0.454) (3.174) (2.270) (1.221) (3.191) (2.416) (2.486) (1.840)
Credit -0.0409 -0.00238 -0.0373 -0.0603 -0.00571 -0.137 0.00407 0.00417 0.000747
(0.0273) (0.00636) (0.0457) (0.0393) (0.0101) (0.0983) (0.0311) (0.00782) (0.0684)
CAR (%) -0.147 0.341*** -0.311** -0.273* 0.182** -0.514* 0.200 0.136* 0.302*
(0.110) (0.0934) (0.134) (0.159) (0.0725) (0.272) (0.153) (0.0685) (0.174)
ROA (%) -1.754*** -0.334 -1.704*** -1.467*** -0.851*** -1.217** -0.885** 0.108 -1.275***
(0.377) (0.278) (0.379) (0.337) (0.256) (0.460) (0.395) (0.186) (0.384)
GDP (%) 0.161* 0.0144 0.152* 0.0241 -0.0536* 0.0116 0.0340 -0.00273 0.0621
(0.0830) (0.0338) (0.0766) (0.0528) (0.0284) (0.0857) (0.0530) (0.0168) (0.104)
Inflation (%) 0.159** -0.0567 0.104 0.0241 -0.143** 0.0119 0.0191 0.0926*** -0.549**
(0.0613) (0.0413) (0.0989) (0.0592) (0.0496) (0.105) (0.0729) (0.0292) (0.230)
REER (%) 0.0269 0.0178 0.127* 0.0422 -0.0170 0.151** 0.0342* 0.0118 0.113***
(0.0402) (0.0129) (0.0641) (0.0402) (0.0191) (0.0654) (0.0173) (0.0102) (0.0394)
Interest (%) 0.119** -0.0276 0.0826 0.0393 -0.110** 0.130 -0.0171 0.0463** -0.669**
(0.0555) (0.0272) (0.135) (0.0485) (0.0450) (0.113) (0.0705) (0.0184) (0.258)
Unemployment (%) 0.738*** -0.138 0.866*** 0.628*** -0.234 0.667** 0.605** -0.0192 0.793**
(0.240) (0.130) (0.282) (0.213) (0.193) (0.314) (0.272) (0.248) (0.308)
Constant 6.539 -2.388 18.94** 11.80 4.518 21.53 -0.0524 0.341 9.185
(4.567) (2.013) (7.594) (8.692) (3.123) (16.31) (2.757) (2.291) (6.857)
Year FE Yes Yes No No No No No No No
Observations 228 108 120 114 54 60 114 54 60
R-squared 0.584 0.588 0.664 0.718 0.638 0.798 0.280 0.471 0.449
Number of countries 38 18 20 38 18 20 38 18 20
Note: Significance at ***1%, **5% and *10% levels respectively. The panel robust standard errors are reported in parentheses. The dependent variable is nonperforming loans to total gross loans. Detailed descriptions and sources of the independent variables can be found in appendix 3. Columns 2, 5 and 8 show the results for the pooled dataset. Columns 3, 6 and 9 show the results for Latin American and developed countries (subsample 1). Lastly, columns 4, 7 and 10 show the results for Eastern European and developing countries.
20 However, for Latin American and developed countries, the CAR increases when NPLs to total gross loans increases. A possible explanation for this finding might be that in Latin American and developed countries NPLs are written-off slowly. This is especially true in Europe where banks struggle with high levels of NPLs. Subsequently banks might increase their CAR as a precautionary measure. The coefficient for the ROA has the expected sign, namely negative, and shows robust results for the different specifications. Similar to Rinaldi
& Sanchis-Arellano (2006), Babihuga (2007), I also find a negative relationship between NPLs to total gross loans and the unemployment rate. Although not significant, note that the sign is negative for Latin American and developed countries for all specifications (columns 3, 6 and 9). Furthermore, model 6 indicates that a highly concentrated banking system has lower NPLs to total gross loans for the pooled dataset and CEE and developing countries.
According to model 5, asset quality decreases when banking systems are less competitive - indicated by an increase in the Boone indicator - for the pooled dataset and CEE and developing countries (columns 5 and 7). On the other hand, more competitive banking systems in Latin American and developed countries have higher NPLs to total gross loans after the GFC (model 6 column 9). Nevertheless, since there is not much variation in the Boone indicator, these results have to be interpreted with care.
4.3 Return on equity
Table 4 reports the regressions results for the ROE. Foreign banks, on average, had a positive effect on the ROE in Latin American and developed countries between 2008 and 2013 (column 3). The opposite, however, holds for CEE and developing countries; when the share of foreign bank assets increases by 1%, ROE decreases by 0.46% (column 4). The coefficient is not significant during the GFC for all specifications (model 8), but is negative and statistically significant for CEE and developing countries after the GFC (model 9 column 10).
Thus, foreign banks decreased profitability in CEE and developing countries between 2008
and 2013 and after the GFC. Notwithstanding, Barth et al. (2002) found that foreign banks
have a positive impact on ROE for all countries in 1999. However, I find that the impact of
foreign banks differ for each subsample. As expected, the results show a strong negative and
statistically significant relationship between NPLs to total gross loans and the profitability of
the banking system models 7 and 8 (columns 2 to 7). Hence, in more risky banking systems -
banking systems with high NPLs to total gross loans - profitability is lower. Moreover,
profitability increases when economic conditions strengthen, measured by an increase in
GDP.
21
TABLE 4
Determinants of Return on Equity
2008-2013 (model 7) 2008-2010 (model 8) 2011-2013 (model 9)
Column 2 3 4 5 6 7 8 9 10
Variables ROE ROE (1) ROE (2) ROE ROE (1) ROE (2) ROE ROE (1) ROE (2)
Foreign (%) -0.180 0.122** -0.464** 0.0107 0.153 -0.136 -0.0920 0.0495 -0.492*
(0.132) (0.0573) (0.213) (0.143) (0.331) (0.229) (0.121) (0.0601) (0.239)
Concentration (%) -0.0498 -0.0365 -0.0624 -0.127 -0.222 0.141 -0.116** -0.0630 -0.125
(0.0635) (0.0927) (0.0512) (0.183) (0.203) (0.376) (0.0515) (0.108) (0.0901)
Competition 1.093 0.668 3.477 18.48* -3.831 33.21*** -11.39 -22.21 -12.63
(7.487) (8.797) (9.752) (9.836) (6.463) (8.678) (11.72) (16.38) (10.23)
Costs 0.0443 0.731 0.00834 0.161*** 2.095 0.109 -0.551 0.0997 -1.499
(0.0305) (0.924) (0.0361) (0.0375) (1.966) (0.0752) (0.782) (0.749) (1.858)
Credit -0.182*** -0.0141 -0.184* -0.0871 -0.0643 -0.240 -0.170 -0.0188 -0.374
(0.0597) (0.0359) (0.0995) (0.125) (0.119) (0.390) (0.157) (0.0535) (0.298)
NPL (%) -1.415*** -1.883** -1.399*** -2.161*** -5.256** -2.116*** -0.923 -0.0453 -0.959
(0.225) (0.842) (0.257) (0.225) (2.182) (0.283) (0.766) (1.399) (0.581)
GDP (%) 0.705*** 0.0483 0.860*** 0.450** -0.130 0.584** 0.265 0.0334 0.727
(0.242) (0.106) (0.260) (0.188) (0.231) (0.270) (0.240) (0.164) (0.501)
Inflation (%) 0.917*** 0.363 0.699** 0.900*** -0.549 0.810** 0.394** 0.464 -0.0449
(0.207) (0.240) (0.267) (0.234) (0.537) (0.305) (0.168) (0.350) (0.800)
REER (%) -0.0865 -0.0687 0.0397 -0.101 -0.299 -0.0326 -0.0639 -0.156 0.0327
(0.0782) (0.0856) (0.102) (0.118) (0.172) (0.217) (0.0812) (0.105) (0.130)
Interest (%) 0.484* 0.200 0.139 0.551* -0.487 0.431 0.153 0.306* -0.435
(0.277) (0.193) (0.424) (0.296) (0.415) (0.550) (0.167) (0.164) (0.951)
Constant 32.13*** 16.67* 51.20*** 28.07 46.76* 28.41 40.02** 16.93 77.03**
(8.603) (7.982) (11.16) (17.75) (24.78) (32.62) (15.05) (11.98) (29.55)
Year FE Yes No No No No No No No No
Observations 228 108 120 114 54 60 114 54 60
R-squared 0.622 0.311 0.719 0.752 0.540 0.839 0.238 0.240 0.416
Number of countries 38 18 20 38 18 20 38 18 20
Note: Significance at ***1%, **5% and *10% levels respectively. The panel robust standard errors are reported in parentheses. The dependent variable is return on equity.
Detailed descriptions and sources of the independent variables can be found in appendix 3. Columns 2, 5 and 8 show the results for the pooled dataset. Columns 3, 6 and 9 show the results for Latin American and developed countries (subsample 1). Lastly, columns 4, 7 and 10 show the results for Eastern European and developing countries.
22 This paper hypothesized that foreign banks had a positive impact on financial stability after the GFC. However, as indicated by the regression results in tables 2 to 4, I do not find support for this hypothesis: I only find a significant but negative relationship between foreign bank assets and ROE for CEE and developing countries. Hence, foreign banks decreased profitability in CEE and developing countries after the GFC. Furthermore, during the GFC, I find a negative and statistically significant relationship between the CAR and foreign banks for CEE and developing countries and the pooled dataset. In other words, foreign banks reduced the CAR between 2008 and 2010. As previously stated, this finding does not necessarily imply that foreign banks led to financial instability. To conclude, this paper finds no support for hypotheses 1 and 2 and partial support for hypothesis 3.
5. SENSITIVITY TESTS
To begin with, the author acknowledges that an important stream of literature is completely neglected, namely the effect of bank regulation and supervision on financial stability. Klomp
& De Haan (2012) found that “supervisory control, capital regulations, and market entry regulations have a significant effect on ‘capital and asset risk’ while supervisory control and regulations on activity restrictions, private monitoring, market entry restrictions, and liquidity have a significant effect on ‘liquidity and market risk’” (p.12). Furthermore, using data on supervisory powers from the survey by Barth, Caprio & Levine (2008), Shehzad & De Haan (2015) found that “giving more powers to supervisory bodies to hire and fire bank managers and to change banks’ organizational setup” (p.2) reduces NPLs to total gross loans (used as a proxy for bank risk-taking). Hence, bank regulation and supervision likely has an effect on the CAR, NPLs to total gross loans and ROE. Unfortunately, in the survey by Barth, Caprio &
Levine (2008, 2013) data are missing or there is no time-variation regulation data for many of the countries used in this study.
Barth, Caprio & Levine (2008, 2013) constructed the ‘capital regulatory index’, which measures “the amount of capital banks must hold and the stringency of regulations on the nature and source of regulatory capital” (Barth, Caprio & Levine, 2013: 14).
18Capital restrictions likely have an impact on the CAR, NPLs to total gross loans and the ROE.
18 The capital regulatory index refers to the ‘capital regulatory all index', which is calculated only if responses to all corresponding questions are available.
