Master Thesis
University of Amsterdam | Business Economics: Finance
What is the effect of bank concentration on financial stability? What are the policy implications? And can adequate trends be identified five years before the recent crisis?
Student J. P. van der Werf Administration number 6165796 Supervisor Prof. dr. E.C. Perotti Date 6 July 2015
Page | 1
Statement of Originality
This document is written by Student, Juliette van der Werf, who declares to take full responsibility for the contents of this document.
I declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it.
The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents.
Page | 2
Abstract
This paper investigates the effect of banking competitiveness on the financial stability In addition; this paper studies also the role of the regulators with respect to this issue and test whether there are other potential determinants related to financial stability. This is tested with the use of a logistic model that predicts the probability a systemic crisis will happen, conditional on certain independent variables, such as bank concentration, regulation policies, and other control factors. The model is using data of 40 countries for the period 1998-2011. The data is mainly collected from the World Data Bank and Bankscope database. The main findings of this paper are (i) when controlling for standard control factors, bank regulations, and policies in a concentrated banking environment, a systemic crisis is less likely to occur, (ii) when controlling for concentration, some regulation policies enhance financial stability, (iii) however, countries with better institutional developments have a lower probability of suffering a systemic crisis, and the level of concentration does not matter in this case and (iv) credit growth and trade balance are significant trends five years before the recent crisis. This paper overall supports the “concentration-stability” view. However, also suggests governance matters.
Page | 3
Agenda
Introduction 4 Literature review 6 Data 13 Methodology 20 Results 23 Conclusion 38 Reference list 41 Appendix 1 45 Appendix 2 46 Appendix 3 48 Appendix 4 49Page | 4
(1) Introduction
The consolidation of the banking system around the world has been the subject of a policy debate for many years. In 2011, The Economist started an active debate about the competitiveness of banks1. In this debate, Professor F. Allen delineated that there are just a few examples of highly competitive stable banking systems, compared to the numerous examples of stable concentrated financial systems. On the other hand, Professor Beck argued that the focus needs to be on the regulatory framework, which stimulates the risk taking incentives of the banking system, and therefore drives the stability and fragility, yet not the bank competition itself.
Over the past years the banking sector has become increasingly concentrated, and this is commonly considered an agreeable development. Specifically, more concentration in the banking system would boost the banks’ profits and enhance their market power. Higher profits and power in turn would increase the charter value of the banks, reducing the banks’ incentives to take excessive risks, and would also act as a buffer to external shocks. At the same time, regulators would be better able to closely monitor them. Overall, this would reduce the likelihood of a systemic financial crisis.
However, regulators are clearly expressing doubts about this traditional theory. The European Commission, for example, forces banks that required state support to divest (for example, Lloyds Banking Group and ING)2. The United Kingdom (UK) wants banks to put firewalls around their retail operations3, while banks in the United Stated (US) are forced into a more competitive environment4. These policy actions are more in line with the concentration-fragility view. This view suggests that the concentration-stability argument is ignoring the influences of market power on the behavior of firms and other industries. Having more market power allows banks to increase their interest charges for firms and clients. These higher charges are greater risks for the firms, and eventually lead to more uncertainty in the financial system. Additionally, regulators are more concerned about failing banks in a more concentrated environment and thus, banks tend to accept more subsidies through “too big to fail” policies, which actually reinforce the incentives of excessive risk taking and accessing fragility in the banking system.
Consequently, it is important to ask: are these regulators helping us by breaking down our big banks, or are they pushing the global economy into a more dangerous playing field? More generally, the financial stability was severely disrupted in 2008 and this instability has had significant repercussions for the entire global economy. These consequences suggest that profound knowledge of what promotes financial stability is highly essential, now more than ever. Also, the role of competition policy that can be envisaged as part of the broader reform of the financial market must be considered carefully. 1 http://www.economist.com/debate/days/view/706 2 http://europa.eu/rapid/press-release_IP-14-554_en.htm 3 http://www.bbc.com/news/business-13770741 4 http://www.bloomberg.com/news/articles/2012-10-25/banks-pushing-into-small-loans-compete-with-payday-shops
Page | 5
The aim of this study is to investigate the effect of bank competition on financial stability, and what the role of the regulators is with respect to this issue. In addition, other adequate trends from five years before the recent crisis are observed to draw conclusions about what promotes financial stability. Following this, the research questions are formulated as follows:
What is the effect of bank concentration on financial stability? What are the policy implications? And can adequate trends be identified five years before the recent crisis?
Following this enquiry, this study provides an empirical research into the field of banking concentration and competition. Similar research has been conducted earlier in this field, however, this study employs the most recent timeframe and thus also incorporates the most recent crisis. This study collects relevant data on 40 countries for the period of 1998 – 2011 and investigates the impact of banking concentration, banking regulation, and banking freedom on the probability that a country suffers from a systemic financial crisis.
While explained in more detail in the third section, this systemic financial crisis is defined according to the two following criteria set by Laeven and Valencia (2012): (i) there are significant symptoms of financial distress in the banking sector, and (ii) there is significant number of policy intervention measures in the banking sector (such as extensive liquidity support, bank restructuring gross costs, bank nationalizations, guarantees put in place, asset purchases, and deposit freezes or bank holidays). Besides focusing on the effect of banking concentration on financial stability, the impact of a country’s regulatory framework is also determined. These regulatory measures are mainly included in this research to (i) provide a robustness check between concentration and crises, and (ii) to simply test the relationship between the regulation of banks and financial stability. A reason why countries choose regulation is to stimulate stability. This model offers further insight into this relationship.
Following the introduction, section two provides an extensive literature review on the relationship between bank competition, concentration, and financial stability, both theoretically and empirically. Section three then describes the data employed for the analysis. The methodology used in the empirical tests is elaborated in section four. Finally, section five discusses the main findings and additional results of this research, and section six offers a succinct conclusion to this study.
Page | 6
(2)
Literature Review
This literature review outlines the broader picture of the relationship between bank competition, concentration, and financial stability, both theoretically and empirically. The first part of this literature review begins with a discussion of the existing theoretical literature in theory. Over the past few years numerous theoretical models have been developed to predict the relationship between competition, concentration, and financial stability. These models, however, draw on two different schools of thought with respect to this relationship. These divergent schools are examined further in this section. In the second part, the existing empirical literature is employed to make predictions for the results of this study.
This part is divided into two types of study levels, namely the country level and the cross-country level. The third part explores literature based on policy implications, which are associated with more competitive financial systems and stability. Finally, the last part reveals literature based on potential determinants of financial stability.
Theoretical existing literature
As mentioned above, numerous theoretical models have been developed to predict the relationship
between competition, concentration, and banks’ financial stability. However these models are rooted in two different schools of thought. The first part commences with the theories that support the concentration-stability view. The second part provides evidence for the opposing theory, the competition-stability view.Concentration-stability hypotheses
Under the traditional banking view, various theoretical models predict that concentrated banking systems are stable. This is because more concentration in the banking system would boost the banks’ profit and enhance their market power. Banks could then use these profits as a cushion against external shocks and possible bank runs. Additionally, higher profits and more market power in the more concentrated banking environments would increase the charter value5 (also called franchise
5 To be more specific about this charter value, Guttentag and Herring (1983) define the charter value as "the present value of the net income
the bank would be expected to earn on new business if it were to retain only its office, employees, and customers”. Generally, the charter value is the present value of future profits that banks are expected to earn from operating in new markets, with economies of scale and superior information. This value could be seen as an intangible asset to the balance of the bank. So in charter value theory, it is the future profit that matters, and not the current profit (because current profit can be distributed, and in fact banks do practice this). So more profits gained in the future allow banks to remain solvent. Therefore, banks do not want to gamble too much, because this increases the risk of losing their future profits.
Page | 7
value) of the banks and reduce the banks’ incentives to take excessive risks. This traditional “charter value” view was developed by Marcus (1984) and Keeley (1990) during the mid-1980s. In these models, banks are seen as being able to choose the risk of their asset portfolio. In a world of limited liability, banks have the incentives to shift risks to depositors, as shareholders only participate in the up-side part of risk. The pressure of more and more competition in the banking sector influences the banks’ profits. Decreasing profitability levels could increase the banks’ incentives to take more aggressive risks to compensate for these lesser profits. In turn, these higher incentives of excessive risk taking would result in a higher level of fragility.
On the other hand, in an environment with restricted entry, and thus a more concentrated system, banks are more willing to take on profitable projects and have better capital buffers. To protect these (future) profits, these bank owners have fewer incentives to take on excessive risks, which results in more financial stability. Besides, regulators are better able to closely monitor the few banks in a concentrated banking system. In contrast, a more competitive environment reduces the banks’ incentives to thoroughly screen and monitor borrowers, due to fewer informational rents gleaned from their relationship between them and their borrowers (Allen and Gale 2000, 2004). This also leads to an increase in fragility risk.
Moreover, Allen and Gale (2000) showed that in a world of perfect competition, interbank emergency lending incentives are reduced. If all banks are price takers, and a peer is affected by a temporary liquidity scarcity, banks have less incentive to provide liquidity. This is because it reduces the value of future profits and thus, of remaining solvent. This can result in a failing bank and could have negative effects on the stability of the whole banking system. Conversely, when banks are located in a more concentrated environment, they have more incentives to act strategically and support a temporarily troubled bank. In short, these theoretical models predict that liberalization of the financial sector (resulting in more entry and competition) would result in a less stable banking system.
Competition-Stability hypotheses
According to the opposing view, a more concentrated banking environment leads to more bank fragility. Following Boyd and De Nicolo (2005), the traditional view (needing market power to boost the banks’ profits) ignores the impact of market power on firm behavior. Banks with more market power have the incentive to increase the interest rate they charge borrowers. These higher interest rates increase firms’ risks when they borrow money to invest in profitable projects. On the other hand, higher levels of competition in the banking sector stimulate banks’ incentives to lower their interest rates. This means lower costs for borrowers, resulting in more certainty and higher firm profits. These higher profit levels lead to more stable firms. These more stable firms are better able to pay back their loans; therefore, the risk of these borrowers in their portfolio is reduced, which eventually results in more stable banks. In addition, this also leads to reducing adverse selection and moral hazard problems.
Page | 8
Another argument, which supports the “competition-stability” view, is that in more concentrated banking systems there are fewer yet larger banks. As shown by Mishkin (1999), policymakers are more concerned about banking failures in a financial system where there are only a few large banks. This is because of the increasing possibility of the contagion risk. Under the assumption of banks being “too big to fail”, regulators tend to give large subsidies during desperate times. The banks, knowing this, reinforce their aggressive risk taking incentives and, thus, a positive relationship between bank concentration and fragility is achieved.
In conclusion, these two opposing views express that a more concentrated banking system enjoys higher profits and increases the charter value, reduces the incentives of excessive risk-taking, and could act as a buffer against possible banking failures. This results in more stable banks. On the other hand, these more concentrated banks are able to increase the interest rates they charge, which leads to more uncertainty for borrowers. This increase in risk for firms further leads to lower firm profits, and thus, an increase in the risk for banks to not be repaid back. This results in a more unstable banking system or to less credit. Increasing the competitiveness in such a banking system causes lower interest charges, more stable firms, and eventually, more stable banks.
Empirical existing literature
Several empirical studies show that there is no direct relationship between the level of competition and financial stability. This section is divided into two types of empirical studies: (i) one (or two) country studies, and (ii) cross-country studies.
One (or two) country studies
There are several bank-level studies that either focused on one country or compared two countries. Keeley (1990) found that higher competition levels in the US results in reduced charter values and the increased fragility levels of banks. The relaxation of state branching restrictions in the 1980s in the US resulted in more competitive banks, where banks’ capital buffers were declined and risk premiums were raised. This view supports Dick (2006), who argued that the deregulation policies in the 1990s had an increasing effect on the loan-loss provisions and charge-off losses. Also, Jimenez, Lopez, and Saurina (2007) revealed, with their study on Spanish banks in the period 1988 – 2003, that a higher market power has a decreasing effect on non-performing loans. This means that there is a negative relationship between market power and banks’ risk, which thus supports the charter value view.
Furthermore, several studies that focused on bank sizes and stability found a negative relationship between bank scale and bank failures in the US (Boyd and Runkle (1993) & Calomiris and Mason (2000)). However, De Nicolo (2000) found evidence that there is a positive relation between bank size and banking failures in countries such as Japan, the US, and a few European
Page | 9 countries.
Additionally, evaluating the effect of mergers in the banking system, several empirical studies again found contradicting evidence with respect to market structure and bank fragility. On the one hand, Paroush (1995) discovered that an increase in market power, due to diversification gains after mergers, has a positive effect on bank stability. This finding is consistent with several other studies, which emphasize the positive relationship between mergers and diversification, resulting in a higher level of bank stability (Benston, Hunter and Wall (1995) & Craig and Santos (1997)). At the same time, Chong (1991) and Huges and Mester (1998) found contrasting evidence that demonstrated that consolidation in the banking system creates higher risk of bank portfolios.
By observing different banking structures across countries, several studies again yielded contradictory results. Comparing the Canadian with the US banking system, Bordo, Redish, and Rockoff (1996) noticed higher stability levels in Canada than in the US. They linked this to the market structure of these two countries, where Canada experienced an oligopoly banking system and the US’ system demonstrated a higher degree of competitiveness. On the other hand, Hoggarth, Milne, and Wood (1998) studied the differences in the banking systems of the UK and Germany, and found higher levels of bank competitiveness and less stability in the UK. At the same time, Staikouras and Wood (2000) found that the banking environment in Spain is more competitive and more stable than the Greek banking system.
Cross-country studies
More recently developed cross-country studies show the same ambiguous picture when testing the validity of the different theories with respect to the relationship between bank concentration, competition, and bank stability. This ambiguous picture is discussed below.
Boyd, De Nicolo, and Jalal (2006) focused on the bank concentration and financial fragility of an individual bank. Using the Z-score as indicator and the Herfindahl-Hirschman index (HHI) as a market concentration indicator, they found that countries with more concentrated banking systems are closer to insolvency and more likely to fail.
Inconsistent with this are the results of Berger, Klapper, and Turk-Ariss (2008). Their results implied that banks with higher levels of market power enjoy less overall risk exposures, which leads to more stable financial systems. The data indicates one component of the competition–stability view. That is that market power is positively related to loan portfolio risks.
Moreover, Beck, de Jonghe, and Schepens (2009) found a significant cross-country heterogeneity in the relationship between bank competition and financial stability due to a variety of factors, such as activity restrictions, deposit insurance, a better-developed stock exchange, and more effective credit information sharing systems.
Conversely, the study by Beck, Demirguc-Kunt, and Levine (2006) supports the concentration–stability view. Similar to this paper, the authors used panel logit models to study the
Page | 10
impact of a concentrated financial system, regulation, and national institutions on the likelihood that a country will experience a systemic banking crisis6. Controlling for several macroeconomic and institutional factors, they concluded that a more concentrated banking environment is less likely to suffer from a systemic crisis. This bulk of evidence is robust, having been subjected to a range of sensitivity checks, and their data was found not to support the concentration–fragility view. However, they also concluded that less regulatory restrictions on banks result in less banking fragility. Also, the data indicates that countries with national institutions that ease the banking competitiveness have a lower probability of experiencing a systemic crisis. In other words, the data indicates that competitive banking systems are also associated with less banking fragility when controlling for concentration. Thus, as mentioned by Beck et al. (2006) the finding that both concentration and competitiveness of a banking system are positively related to banking stability suggests that bank concentration is an insufficient measure for competition. Similarly, Claessens and Laeven (2004) could not find evidence that inverse competition correlates with bank concentration.
At the same time, Schaeck, Cihak, and Wolfe (2009) also found an inverse relationship between bank competitiveness and banking failures. They showed that more competitive environments, measured with the H-statistics, are less sensitive to systemic crises. They did not find any evidence that significantly linked banking concentration and banking fragility.
Policy implications
As mentioned above, refurbishing a contestable financial system positively impacts the financial stability of banks. On the other hand, stimulating banks to grow and become more concentrated is beneficial with respect to risk diversification. Thus, extensive literature shows that policymakers and their decisions are crucial in creating and maintaining a stable financial environment.
In their study, Demirguc-Kunt and Detragaiache (1999) found that in a weak institutional environment with deregulated interest rates, the effect of deposit insurance is detrimental for stability and tends to increase the likelihood for financial crises. Similarly, many studies that observed the experiences of the liberalization of the banking system over the past years also revealed detrimental results. In liberalized systems, banks are able to shift risks to the taxpayers, leading to aggressive risk taking and often resulting in systemic financial distress. Dell’ Ariccia, Igan, and Laeven (2008) studied the most recent sub-prime crisis in the US. They found that at times of loose monetary policy and increasing financial developments (such as securitization), the entry of new competing lenders results in a decline in lending standards.
Moreover, in their observations of the effects of regulatory policies on bank stability, Caprio and Levine (2004)7 stated that restrictions to entry and to banks’ activities result in a less competitive banking system with a higher likelihood of suffering a crisis. Their cross-country evidence also
6
Rather than focusing on the individual banking fragility like Boyd et al. did.
7
Page | 11
indicates that capital regulations are not significantly linked with the probability of financial distress. Hellmann, Murdock, and Sitglitz (2000) investigated policy implications in a dynamic moral hazard model. They found deposit rate ceilings and other regulatory restrictions reduce risky gambling incentives and prevent unhealthy competition, which would otherwise lead to bank fragility. Recently, Kaufman et al. (1999) constructed six aggregate indicators that correspond to six underlying governance concepts. With this new empirical evidence, they suggest there is a strong causal relationship between good governance and better development outcomes, such as higher per capita incomes, lower infant mortality, and higher literacy. Claessens and Laeven (2004) studied the interaction of policy regulation with bank competitiveness, with a particular focus on entry of foreign banks. They found that foreign bank participation contributes to financial stability in the system. However, Cull and Martinez Peria (2007) revealed contradicting results in their cross-country study. They concluded that countries who suffer a systemic crisis tend to have increasing foreign bank participation levels.
Thus, in weak institutional environments competition results in financial distresses. Alternatively, copious literature showed that policies, which are stimulating bank competition (such as lower entry barriers and openness to foreign bank entry), are related to financial stability. So in other words, this literature suggests it is important to enhance the regulatory framework, rather than constraining bank competitiveness.
Other potential determinants
In addition, this study investigates other potential determinants of financial stability. Many earlier studies have investigated the potential determinants of the recent crisis. Especially, credit growth has been studied by several studies over the past few decades. Borio and Drehman (2009) demonstrated that episodes of banking crises were preceded by rapid expansion of domestic credit to the private sector.
Also Schularick and Taylor 2010 showed that credit growth appears to be an indicator of financial instability. Bernanke, Gertler and Gilchrist, (1996) found a similar result. They suggested that credit booms have been followed by busts. These busts are noticed by contractions of credit and economic downturns.
Amri, Prabha and Wihlborg (2012) found that in advanced economies relatively high credit growth over several years increases the probability a banking crisis will occur. High leverage, weak capital regulation and supervision, and cumulative asset price inflation strengthened this effect.
However, as above mentioned that financial crises often are preceded by high credit growth, Gourinchas, Valdes and Landerretche et al. (2001) and Caprio and Klingebiel (1997) found that rapid expansion of credit is not a strong indicator of a financial crisis.
Page | 12
Overall, the extensive literature on the relationship between bank concentration, competition, and stability is ambiguous and reveals that much more research is needed. Especially because the financial stability was severely disrupted in 2008, and this instability had enormous effects on the entire global economy, it is important to further investigate whether the market structure of the banking sector influences these disruptions. Also, the role of competition policy that can be envisaged within the broader reforms of the financial market must be considered carefully. This research is similar to earlier empirical models, however, it contributes to the existing literature as it employs the most recent timeframe and a different set of countries. This recent timeframe allows this research to incorporate the most contemporary developments concerning regulatory policies and countries’ market structures. Additionally, the most recent global financial crisis is included in this sample period. As such, this study provides further investigation into the impact of concentration and regulatory policies on financial stability in the banking sector. Also, potential trends from the time before the recent crisis are examined.
Following prior studies, the main hypothesis is formulated as follows:
Hypothesis 1: In a concentrated banking system, the financial environment is stable.
Moreover, the role of the regulators in the competitive system is assessed. Therefore, the following hypothesis is constructed:
Hypothesis 2: In a regulated framework, which impedes competition, the financial environment is stable.
To test whether the interaction between concentration and regulation is associated with financial stability, the following sub-hypothesis is formulated:
Hypothesis 2.1: In a regulated framework, the interaction between concentration and regulation affects financial stability.
Finally, the role of the other control factors are assessed, particularly, the trends of credit growth and trade balance five years before the recent crisis. With respect to this issue, the following hypothesis is established:
Hypothesis 3: Credit growth and trade of balance were adequate trends five years before the crisis.
Page | 13
(3) Data
To investigate the effect of competition in the banking sector on financial stability, a logistic analysis is used to predict the probability a country will experience a systemic crisis. More specifically, this model predicts the probability that a systemic crisis will occur, depending on certain independent variables, such as bank concentration, regulation policies, and other control factors. The model employs data from 40 countries for the period 1998-2011. These 40 countries consist of all the countries that are a member of the Organisation for Economic Co-operation and Development (OECD). The rest are non-OECD countries. Appendix 1 shows all the countries employed in this study. This sample is used as, according to the OECD Financial Report 2011, some OECD countries (such as Canada and Australia) have more concentrated systems than other OECD countries. Using mainly OECD countries to test the effect of market structure in the banking industry, which difference between these countries, can thus provide a solid result. The data is mainly collected from the World Data Bank, which includes collections of time series data on various topics.
This section discusses the collected data and the different measures that are used to test the effect of banking competition and concentration on financial stability. The following section then explains the methodology of the model in more detail.
Sample data and summary statistics
As previously mentioned, the sample period is 1998-2011. Table 1 provides the descriptive statistics for the whole sample period. Furthermore, this research study deals with missing values by a particular method. If the amount of missing values for each variable is under the 10% level of all the observations, the missing observation is replaced by a country’s average of the period of 1998-2011. This is a commonly used approach to deal with missing values in empirical studies, and is thus the most appropriate approach for this analysis. This is discussed in more detail per variable that experiences missing values. The remainder of this section discusses the composite of the dependent variable, followed by explanations of the independent variables.
Dependent variable
Systemic Banking Crisis
To test the probability of the occurrence of a systemic crisis, a crisis dummy variable is needed to predict the likelihood. The dependent variable is a binary variable that takes the value of one when a financial sector is situated in a financial crisis, and zero otherwise. This banking crisis variable is
Page | 14
based on two criteria set by Laeven and Valencia (2012). They interpret a banking crisis to be systemic if these two conditions hold true:
Condition 1: There are significant symptoms of financial distress in the banking sector8.
Condition 2: There are a significant number of policy interventions in the banking sector. These interventions are initiated in response to significant losses in the banking sector.
Table 1 Summary statistics
Variable Obs. Mean Std. Dev. Minimum Maximum
Systemic banking crisis 560 0.185714 0.389224 0 1
Initial crisis years 488 0.059426 0.236663 0 1
Concentration 560 0.677158 0.208718 0.21399 1
Entry barrier index 510 0.072882 0.157271 0 0.695652
Activity restrictions index 508 6.757874 1.957748 3 12
Capital regulation index 493 6.896552 1.79301 2 10
Financial freedom 560 64.66071 18.20847 30 90 Economic freedom 560 67.79196 8.457865 47.4 82.6 Quality of institutions 560 77.35464 19.6682 21.10325 99.75804 GDP growth 560 3.088983 3.461169 -14.7376 14.16239 Trade of balance to GDP 560 2.062055 7.075659 -17.02494 33.59216 Credit growth 560 0.047952 0.109902 -0.47919 1.450454 Inflation 560 4.678954 8.952971 -5.3903 137.9649 Real interest 546 5.201086 7.817081 -24.6002 78.78996 M2/reserves 560 42.40069 145.9751 1.162266 1637.684 GDP per capita 560 26027.91 20203.45 425.4453 113738.7
The dependent variable is a crisis binary variable that takes the value of one when a financial sector is situated in a critical environment and zero otherwise. The real GDP growth is the annual percentage growth rate of GDP. Trade is the sum of exports and imports of goods and services measured as a share of GDP. Credit growth is calculated as the sum of the private credit by both deposit money banks and other financial institutions, divided by GDP. Inflation is determined as the annual growth rate of the GDP implicit deflator. Real interest rate is the lending interest rate minus inflation as measured by the GDP deflator. M2/Reserves is determined by the sum of money and quasi money to the total reserve ratio. GDP per capita is the ratio of gross domestic product divided by the midyear population. Concentration is estimated by the share of assets held by the three largest banks to the total commercial banking assets. Entry barrier index is the number of entry applications denied as a fraction of the total of applications received form domestic and foreign countries. Activity restrictions index is measured as the degree to which national regulatory institutions allow banks to participate in activities such as securities, insurance, and real estate business. Capital regulatory index is calculated as the sum of initial capital stringency and overall capital requirements. Financial freedom is calculated as the measure of the efficiency of the banks as well as a measure of independence of these banks from government control. Economic freedom is a composite measure of ten policy indicators in the field of governance finances, government interventions, monetary policy, trade, banking and finance, regulation, property rights, wages and prices, capital flows and international investment, and black market activity. The quality of institutions is a composite measure of governance indicators. Bank data is collected from the Bankscope database of Fitch IBCA. These country level variables are collected from the World Bank’s World Development Indicator database (GDP growth, trade to GDP, inflation, real interest, M2/reserves and GDP per capita) and the Global Financial Development Database (credit growth). Regulatory data is collected from Barth. et al. surveys which in turn are collected from The Bank Regulation and Supervision Survey. Surveys from 2012, 2008, and 2003 are used. The freedom data is collected from the Heritage Foundation (2015). Governance indicators are collected from Worldwide Governance Indicators.
Page | 15
The first year where both conditions are met is treated as the starting year of the systemic banking crisis. The year before both real Gross Domestic Product GDP growth and real credit growth are positive for at least two consecutive years is treated as the closing year of the banking crisis. Laeven and Valencia (2012) further define the number of policy interventions to be significant if at least three of the following tools have been applied:
Extensive liquidity support (five percent of deposits and liabilities to nonresidents).
Bank restructuring gross costs (at least three percent of GDP).
Significant bank nationalizations.
Significant guarantees put in place.
Significant asset purchases (at least five percent of GDP).
Deposit freezes and/or bank holidays.
The data collected for the crisis dummy is from Laeven and Valencia’s (2012) Systemic Crisis Database file. For the period 1998-2011, the sample includes 40 countries and 104 (of the 560 observations) crisis episodes. It is very likely that some (or may be all) independent variables are affected by the crisis itself. To actually incorporate the behavior of the concentration level in the banking system and the other explanatory variables in the crisis observations, only the initial crisis year observations are used first to provide valid information about the relationship between concentration and the crises. This is accomplished by excluding the subsequent years of crisis. For the period 1998-2011, the sample includes 40 countries and 29 (of the 488 observations) initial crisis periods.
Independent variables
Concentration measure
The variable of interest is the concentration variable. To measure the degree of concentration in the banking sector, the three-asset bank concentration measure is applied. This measure is estimated as the share of assets held by the three largest banks divided by the total commercial banking assets9.
The unconsolidated data is obtained from the Bankscope database by Fitch-IBCA. This database includes bank balance sheet data from a large cross-section of countries (including OECD countries). As mentioned earlier, this paper replaces the missing value of bank concentration with a country’s average amount of observations. Thus, for this variable as well, the missing values are replaced with a country’s average concentration level. Table 1 demonstrates that the concentration level for this
9
Total assets include total earning assets, cash and due from banks, foreclosed real estate, fixed assets, goodwill, other intangibles, current tax assets, deferred tax, discontinued operations, and other assets.
Page | 16
sample broadly varies. The range varies from around 20% (the US) to 100% (African countries), with a mean of 68%.
Regulatory measures
To incorporate the impact of a regulatory framework on financial stability, the following three regulatory measures are employed10:
1. Entry barrier index. This index is the number of entry applications denied as a fraction of the total of applications received form domestic and foreign countries. This index measures the entry restrictions in the banking sector and thus, the contestability of the sector. To be more specific, this regulatory variable is related to policies that stimulate or restrict competition. On the one hand, this regulation tool can stimulate bank profits and thus stimulate financial stability. However, this variable could also induce inefficiencies in the banking system, and could result in less financial stability.
2. Activity restrictions index. This variable measures the degree to which national regulatory
institutions allow banks to participate in activities such as securities, insurance, and real estate business. In the surveys, the activity restriction degrees are categorized in the following four categories: activities are (1) unrestricted, (2) permitted, (3) restricted, and (4) prohibited. So the higher the value of the activity restriction index, the higher the restrictions on bank activities. These activity restrictions keep banks from entering excessively risky projects. In sum, this index measures the bank regulatory and supervision. These restrictions deter banks from opting for excessive risk taking projects and, thus, lead to more stable banks and financial systems. But at the same time, these restrictions also keep banks from diversifying outside the traditional patterns. This could decrease the probability of financial stability.
3. Capital regulation index. Basically, this index measures the amount of capital banks must maintain and the solvency of banks. The measure is calculated as the sum of initial capital stringency and overall capital requirements. The initial capital stringency incorporates whether certain funds may be used to initially capitalize a bank and whether this is accomplished officially. The overall capital requirements include the requirement that reflects on certain risk elements and deducts certain market value losses from capital before minimum capital adequacy is determined. Both these values are calculated by answering several questions about capital, which is explained in more detail in appendix 2. A higher value of the capital regulation index signals a higher level of stringency. Additionally, a higher level of capital regulation index further
10
Page | 17
stimulates the bank solvency, thus it is expected that in countries with higher capital regulation indexes, the financial systems are more stable.
Controlling for these regulations and restrictions on banks provides additional evidence of the relation between concentration, competition, and systemic financial stability. If these regulations encumber competition, then controlling for these regulation tools can eliminate the significance of concentration in the logit model and provide information on the potential significance of the concentration coefficient.
The data of these estimations are collected from the three surveys conducted by the World Bank. The Bank Regulation and Supervision Survey11 is updated over the years and is the only source of comparable global data on how banks are regulated and supervised. The latest review includes information on regulation and supervision for 143 jurisdictions and data accumulated since 2008. This means the current survey examines the latest state of bank regulation and supervision; thus, one can compare this to the pre-crisis situation. This study employs the survey data of the year 2003 for the years 1995-1999, the survey data of 2008 for the years 2000-2005, and the survey data of 2012 for the years 2006-2010.
Table 1 presents diverse variation in the regulations variables. Where some countries experience zero entry barriers, others deny seven out of 10 bank applications. Also, activity restrictions on banks vary from three to 12, and the capital regulation index varies from two to 10, which is again a very wide spectrum.
Freedom and quality of institutions variables
To capture the magnitude of freedom in the banking system and economic freedom in general, two additional variables are used to measure this. These variables are included in the model to test how freely banks can operate, in their financial environments, as well as in the general economy. Also, the effectiveness of the governments is incorporated in the model. This captures the magnitude of the effect of institutional development on financial stability.
Financial freedom. This variable is determined as the measure of the efficiency of the banks, as well
as a measure of independence of these banks from government control. This variable thus indicates the openness of the financial system. Higher values demonstrate that there are fewer restrictions within the financial system and thus, that banks are able to operate more freely. This could, on the on hand, stimulate banking efficiency which results in more stability. However, it also has the potential to induce destabilizing banking competition, which could be detrimental for financial stability.
11http://econ.worldbank.org/WBSITE/EXTERNAL/EXTDEC/EXTRESEARCH/0,,contentMDK:20345037~pagePK:64214825~piPK:64214
Page | 18
Economic Freedom. This indicator is a composite measure of ten policy indicators in the field of
governance finances, government interventions, monetary policy, trade, banking and finance, regulation, property rights, wages and prices, capital flows and international investment, and black market activity. This variable determines how a country’s policy is ranked in terms of providing general economic freedoms. Higher values suggest that policies are more conducive to competition and economic freedom. Again, this should stimulate efficiency in the financial banking system, which could result in more stable banking environments. However, greater freedoms also stimulate more risk taking, which could result in more fragile banking environments.
Quality of institutions. To control for the overall level of institutional development, this composite
indicator is used. This indicator is the weighted average of six composite indicators for broad dimensions of governance. The six aggregate indicators are (1) Voice and Accountability, (2) Political Stability and Absence of Violence/Terrorism, (3) Government Effectiveness, (4) Regulatory Quality, (5) Rule of Law, and (6) Control of Corruption. These six indicators capture governance perceptions as reported by survey respondents, non-government organizations, commercial business information providers, and public sector organizations worldwide. These underlying indicators are constructed by Kaufman et al. (1999). The overall indicators’ percentile rank ranges from 0 (lowest) to 100 (highest). Higher ranks correspond to better governance.
Both freedom variables are collected from the Heritage Foundation12. This foundation is specialized in formulating and promoting policies that are based on principles such as free enterprise, limited government, and individual freedom. The six indicators used to compose the quality of governance indicator, are collected from the World Bank’s Worldwide Governance Indicators.
There are no missing values for these variables and table 1 shows that there is significant variation in the observations. Some countries have highly financial free banking sectors, whereas others demonstrate the opposite. The range of economic freedom is narrower, but still widely varied. Also, the development of governances differs widely from country to country.
Control measures
In examining the relationship between competition, concentration, and systemic stability, this research study controls for a number of country level variables that are likely to affect the quality of the banking sector:
1. GDP growth. This variable is the annual percentage growth rate of GDP13 at market prices based on constant local currencies. It measures the macro-economic development.
12
http://www.heritage.org/research/reports/2014/04/using-the-index-of-economic-freedom-as-guide
13 GDP is the sum of gross value added by all resident producers in the economy, plus any product taxes, and minus any subsidies not
Page | 19
2. Trade Balance to GDP. Trade is the difference between exports and imports of goods and services measured as a share of GDP. This variable measures global integration.
3. Credit growth. This variable is based on the credit growth ratio of Beck et al. (2009). This
credit growth measure is calculated as the sum of the private credit by both deposit money banks and other financial institutions, divided by GDP. This control factor is included in the model to allow for the development of the banking sector.
4. Inflation. Inflation is determined by the annual growth rate of the GDP implicit deflator. This
indicator presents the rate of price change in the economy as a whole. The GDP implicit deflator is the ratio of GDP in a current local currency to GDP in a constant local currency, and allows for macro-economic development.
5. Real Interest. Real interest rate is the lending interest rate minus inflation, as measured by the GDP deflator. This real interest rate is incorporated to evaluate the costs of funds for banks. Also, interest rate may influence a bank’s profitability, thus increasing default rates.
6. M2/Reserves. M2/Reserves is determined by the sum of money and quasi money, which is
divided by the total reserve ratio. Money and quasi money includes the currency outside banks, demand deposits other than those of the central government, and the time, savings, and foreign currency deposits of resident sectors other than the central government. Total reserves cover the holdings of monetary gold, special drawing rights, reserves of International Monetary Fund (IMF) members held by the IMF, and holdings of foreign exchange under the control of monetary authorities. This variable captures the vulnerability of the banking system to an abrupt capital outflow created by foreign exchange reserves.
7. GDP per capita. GDP per capita is the ratio of GDP divided by the midyear population. GDP
includes the gross value added by all resident producers in the economy, plus any product taxes, and minus any subsidies not included in the value of the products. This variable incorporates the overall level of development of a country.
These country level variables are collected from the World Bank’s World Development Indicator database (GDP growth, trade to GDP, inflation, real interest, M2/reserves, and GDP per capita) and the Global Financial Development Database (credit growth). The missing values for these control values are handled in the same way as the missing values of the concentration variable. If the amount of missing values for each variable is under the 10% level (which holds for all of these control variables), the missing observation is replaced by a country’s average from the period of 1998-2011. Finally, table 1 shows the wide range of variation in the observations for the control variable.
Page | 20
(4) Methodology
The logistic model
This study uses a logit probability model to test whether bank concentration affects systemic banking stability. This regression calculates the probability that a systemic banking crisis could take place in a specific country at a specific time, as a function of certain variables. The model is robust to heteroscedasticity. The likelihood function of the model is equal to the following equation:
equation (1) 𝐿𝑛 𝐿 = ∑𝑡=1,..,𝑇∑𝑖,..,𝑛{𝑦(𝑖, 𝑡)𝑙𝑛[𝐹(𝛽′𝑋(𝑖, 𝑡))] + (1 − 𝑃(𝑖, 𝑡))𝑙𝑛[1 − 𝐹(𝛽′𝑋(𝑖, 𝑡))]}
In equation (1), 𝑖 represents the specific country and 𝑡 the time period, and the equation sums up these two variables. 𝐿𝑛 𝐿 is the log-likelihood that a crisis will occur. This estimation is a log function because it is more convenient and easier to work with14. 𝑦(𝑖, 𝑡) in equation (1), represents the dependent dummy variable that takes the value of one when a banking crisis occurs in country i at time t. 𝑙𝑛[𝐹(𝛽′𝑋(𝑖, 𝑡))] is the cumulative probability distribution function where 𝛽′ is a vector of the unknown coefficients and X can be replaced by:
𝑋 = 𝛽0+ 𝛽1𝐵𝑎𝑛𝑘 𝐶𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛𝑖,𝑡+ ∑ 𝛽2𝑅𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 𝑃𝑜𝑙𝑖𝑐𝑦𝑖,𝑡+ ∑ 𝛽2𝐶𝑜𝑛𝑡𝑟𝑜𝑙 𝑓𝑎𝑐𝑡𝑜𝑟𝑠𝑖.𝑡 + 𝜀𝑖,𝑡 It must be noted here that the estimated coefficients for each particular explanatory variable do not represent the magnitude of the effect of a marginal change in the explanatory variable regarding the likelihood of a systemic crisis. It is just the sign and the significance level of the coefficient that show whether an increase of the explanatory variable has an increasing or a decreasing effect on the dependent variable. The slope coefficient (𝛽′) can be interpreted as the rate of change in the "log
odds" (which is ln (1−𝑦(𝑖,𝑡)𝑦(𝑖,𝑡) )) when an explanatory variable changes. To estimate a more intuitive effect, the marginal effect of the independent variables is calculated. This depends on the slope of the cumulative distribution function. These marginal coefficients estimate the magnitudes of the relationship between each independent variable and the likelihood of a systemic banking crisis. Throughout this research, the marginal effects of the variables are represented.
As mentioned earlier, this study tries to answer the following questions:
What is the effect of bank concentration on financial stability? What are the policy implications? And can adequate trends be identified five years before the recent crisis?
14 The log-likelihood is a function that increases monotone and accomplishes its maximum value at the same peak as the normal likelihood
Page | 21
To find valid answers to these research questions, four hypotheses are constructed:
Hypothesis 1: In a concentrated banking system, the financial environment is stable.
For this hypothesis, the model tests whether the marginal effect of concentration is equal to: 1. 𝐻0 = 𝛽𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛= 0 or,
2. 𝐻1 = 𝛽𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛< 0.
Hypothesis 2: In a regulated framework, which impedes competition, the financial environment is stable.
For this hypothesis, the model tests whether the regulation marginal effect is equal to: 1. 𝐻0 = 𝛽𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛= 0 or,
2. 𝐻1 = 𝛽𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛< 0.
Hypothesis 2.1: In a regulated framework, the interaction between concentration and regulation is affecting financial stability.
For this sub-hypothesis, the model tests whether the interaction variable is equal to: 1. 𝐻0 = 𝛽𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 ∗ 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛= 0 or,
2. 𝐻1 = 𝛽𝑟𝑒𝑔𝑢𝑙𝑎𝑡𝑖𝑜𝑛 ∗ 𝑐𝑜𝑛𝑐𝑒𝑛𝑡𝑟𝑎𝑡𝑖𝑜𝑛 < 0.
Finally, the roles of the other control factors, including the trends of credit growth and trade balance five years before the recent crisis, are also examined. The following hypothesis is investigated with respect to this issue:
Hypothesis 3: Credit growth and trade of balance are adequate trends five years before the crisis.
For this hypothesis, the model tests whether the marginal effect of credit growth and/or trade of balance is equal to:
1. 𝐻0 = 𝛽𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑓𝑎𝑐𝑡𝑜𝑟 = 0 or,
2. 𝐻0 = 𝛽𝑐𝑜𝑛𝑡𝑟𝑜𝑙 𝑓𝑎𝑐𝑡𝑜𝑟 ≠ 0.
To test the first three hypotheses, a one tailed test is used. For the last hypothesis, a two tailed test is used. Linking this to the logit probability model, if the concentration marginal effect is significant and negative, then there is enough evidence to conclude that hypothesis 1 is supported and H0 can be rejected. The same holds for the marginal effect of regulation and the interaction between regulation and concentration. If the particular tested marginal effect of these variables are negative and
Page | 22
significant than there is enough evidence to conclude hypothesis 2 and 2.1 are supported by this study and H0 can be rejected. For the last test, if the effects of credit growth and/or trade balance are significantly related to financial stability, there is enough evidence to support hypothesis 3.
As mentioned above, many crises run for many years. So the behaviors of some of the independent variables, including the concentration variable, are likely to be affected by the crisis themselves. By excluding the subsequent crisis years after the initial year, this reverse causality problem is minimized. It is the reserve causality problem which would undermine this study. Another concern is the possibility of unobserved or omitted variables that could also affect financial stability. This concern is more difficult to deal with. While the standard control factors are based on existing literature, this paper is fully aware that this analysis is potentially missing additional control variables.
Page | 23
(5) Results
This section discusses the results of this paper. The first and the second part show the main evidence of the relationship between concentration, regulation and financial stability. The third part shows how concentration interacts with regulation. In the fourth part discusses the differences with the paper of Beck. et al (2006). In this part a sub sample period is used. Finally, other potential determinants are tested in the last part.
Supported by evidence. Concentration and crises
Table 2 presents the main results of this study. Regression 1 presents the marginal effects of the standard control variables on the initial crisis year observations15. The growth rate of real GDP, balance of trade, the inflation rate, interest rate, and GDP per capita are highly significant. Real GDP growth, balance of trade, and GDP per capita are negatively related, while inflation and real interest are positively related to financial fragility. This is consistent with the findings of earlier empirical literature.
Adding all the subsequent crisis observations in the model, the significance levels of the control variables change to a certain degree. As shown in regression 2, credit growth becomes significant, while the real interest variable is no longer significant.
Regression 3 tests the marginal effects of the control variables and the concentration variable on the initial crisis years. The significance level of these standard control variables remains the same, similar to the ones in regression 1. Regression 3 calculates the concentration variable at the 5% significance level. This effect indicates that a higher degree of concentration in the banking sector reduces the probability that a crisis will occur. This estimation also suggests that the economic impact is significantly large. One standard deviation increase in the concentration variable (0.21, see table 1) results in a decrease of the probability that a systemic crisis will occur by 1,02% (-1,02%=0,21*-0,0484). The average probability of a crisis is around 6%, so a 1,02% decrease is substantial.
Adding together all crisis year observations in the model, the regression still presents a negative and significant marginal effect of the concentration variable, as shown in regression 4. The result shows a significant relationship between bank concentration and all the crisis observations at the 5% significance level. Also, the economic impact appears to be large in this regression, predicting a decrease in the probability of a crisis of around 3%. This is large compared to the crisis mean of 19%. Thus, the main result supports the concentration-stability view. In addition, the control factors GDP growth, trade balance, credit growth, and GDP per capita are highly significant in this regression. Inflation and M2/reserves are significant at the 5% level. The real interest variable proves to be insignificantly related to financial stability when observing all the crisis years.
15
Page | 24
Table 2. Banking crises, concentration, and controls. (1998-2011)
(1) (2) (3) (4) (5) (6) Real GDP growth -0.0090*** -0.0381*** -0.0084*** -0.0376*** -0.0082*** -0.0348*** (0.0027) (0.0056) (0.0034) (0.0058) (0.0026) (0.0057) Trade Balance -0.0015** -0.0051*** -0.0014*** -0.0047*** -0.0013* -0.0038** (0.0007) (0.0016) (0.0007) (0.0016) (0.0008) (0.0017) Credit growth 0.0205 -0.5283*** 0.0301 -0.4952*** 0.0227 -0.4835*** (0.0343) (0.1534) (0.0353) (0.1562) (0.0370) (0.1394) Inflation 0.0022*** 0.0042** 0.0020*** 0.0041** 0.0019*** 0.0038** (0.0006) (0.0022) (0.0007) (0.0021) (0.0007) (0.0020) Real Interest 0.0013*** 0.0008 0.0011** 0.0005 0.0012*** 0.0013 (0.0009) (0.0022) (0.0005) (0.0018) (0.0005) (0.0018) M2/Reserves -0.0000 -0.0001 -0.0001 -0.0002** -0.0001 -0.0001 (0.0001) (0.0001) (0.0001) (9.12E-05) (0.0001) (8.74E-05) GDP per Capita 0.000*** 2.78E-06*** 1.34E-05*** 3.49E-06*** 1.27E-06*** 3.62E-06***
(3.58E-07) (8.58E-07) (3.21E-05) (8.77E-07) (3.75E-07) (8.32E-07)
Concentration -0.0484** -0.1432** (0.0293) (0.0661) Concentration*Q1 -0.0809 -0.6483** (0.1472) (0.3356) Concentration*Q2 -0.0346 -0.2972 (0.0995) (0.2092) Concentration*Q3 -0.0456 -0.3378** (0.0842) (0.1743) Concentration*Q4 -0.0456 -0.2527** (0.0695) (0.1456) Concentration*Q5 -0.0594 -0.3415** (0.0603) (0.1297) No. of crises 29 104 29 104 29 104 No. of obs. 475 546 475 546 475 546 Pseudo r2 0.2070 0.2849 0.2175 0.2944 0.2254 0.3223
The logit probability model is calculated as follows:
Crisis(country=j,time=i)= α + β1Real GDP growthj,i+ β2Trade Balance to GDPj,i+ β3Credit growthj,i+ β4Inflationj,i+ β5Real Interest ratej,i+ β6M2/Reservesj,i+ β7GDP per Capitaj,i+ β8Concentrationj,i+ εj,i.
The dependent variable is a crisis binary variable that takes the value of one when a financial sector is situated in a critical environment, and zero otherwise. The real GDP growth is annual percentage growth rate of GDP. Trade is the sum of exports and imports of goods and services measured as a share of GDP. Credit growth is calculated as the sum of the private credit by both deposit money banks and other financial institutions, divided by GDP. Inflation is determined as the annual growth rate of the GDP implicit deflator. Real interest rate is the lending interest rate minus inflation, as measured by the GDP deflator. M2/Reserves is determined by the sum of money and quasi money to the total reserve ratio. GDP per capita is the ratio of GDP divided by the midyear population. Concentration is estimated by the share of assets held by the three largest banks to the total commercial banking assets. These country level variables are collected from the World Bank’s World Development Indicator database (GDP growth, trade to GDP, inflation, real interest, M2/reserves, and GDP per capita) and the Global Financial Development Database (credit growth). Bank data is collected from the Bankscope database of Fitch IBCA. Regressions (1), (3), and (5) exclude the crisis observations after the first crisis year. Regressions (2), (4), and (6) include all crises observations. Regressions (5) and (6) are piecewise regressions that are used where the concentration variable is divided into five quintiles and enters five times, each time multiplied by a dummy quintile. This table shows the marginal effects (dydx) of the logit regressions. White’s heteroskedasticity consistent standard errors are given in the parentheses. This model uses the one tailed t-test.
Page | 25
Furthermore, regressions 5 and 6 are piecewise regressions, used to check for a non-linear relationship between the crises and the concentration variable. In these regressions, the concentration variable is divided into five quintiles. These five quintiles are concentration dummies that are identical to one if the concentration value falls into the particular quintile range, and are identical to zero otherwise. In other words, there are five concentration quintile dummies for each observation with the value of one or zero. To run the regression, the concentration variable is multiplied by each of the quintile dummy variables. Following this, for each observation, one concentration*quintile variable becomes the value of the concentration variable, where the other four concentration*quintile dummies are set to zero.
Regression 5 observes the marginal effects of these quintile variables and shows that none of the concentration quintiles are significant. One could thus assume that there is not enough evidence to conclude that a relationship between concentration and crises exists.
However, including all the crisis observations leads to a significant effect of all quintile variables, except for the second quintile variable, as shown in regression 6. This last regression suggests that the stabilizing effect of concentration is statistically significant for each concentration quintile and thus, indicates that there is a negative relationship between concentration and fragility in the banking system. Moreover, in this last regression, the control factors are significantly related to financial stability, except for real interest and M2/reserves. The Pseudo R2 varies from 20% to 33%
depending on the regression and estimates the fit of the model.
Furthermore, the significance level of several control factors is of great importance. Where GDP growth, trade balance, inflation, and GDP per capita are significant in all regressions, there are others depending on the number of crises observations.
The significance level of credit growth is particularly notable. The even numbered regressions show that credit growth becomes negative and significant when including all the crisis years. This suggests that this control variable is not necessarily causing significant systemic crises. However, when adding the subsequent crisis years, this factor is related to financial stability. So when a country is suffering from a crisis, credit growth is related to the length of the crisis. However, this model does not show whether credit growth has a causal effect on the size of the crisis periods. Although recent related literature (for example, Alessi and Detken, 2014) suggests that excessive credit growth often leads to systemic risks, it seems possible that a longer period of crisis can have an impact on the reduction in the amount of credit provided by banks and other financial institutions.
Where credit growth becomes significant when adding all the crisis observations together, real interest shows the opposite result. This variable becomes insignificant when the model includes all the crisis observations. Consequently, this model suggests that real interest rate could be one of the causes of a crisis, however, it is not related to the length of a crisis.
Later in this study, it is tested in more detail whether these control variables are potential determinants for the recent crisis. This is achieved by running regressions with observations of five years before the recent crisis.
Page | 26
Overall, the main suggestion of the model is that concentration in all regressions is significant (except for the fourth regression). So it is appropriate to conclude that the concentration marginal effect appears to have a significant and negative effect. This acts as evidence to reject H0 and supports the main hypothesis.
Supported by evidence. Concentration, regulation, freedom, and crises
This part tests the relationship between concentration, regulation, and the likelihood that a crisis will occur. Three different variables are included in the model, namely regulation variables, freedom variables, and the overall quality of institutions indicator. These variables are included because they provide a robustness check between concentration and crises. Also, they are used to provide additional information on the concentration-stability relationship. Controlling for regulatory systems eliminates the significance level in the concentration variable, if concentration is simply a proxy for regulatory policies that encumber competition in the system. If the concentration variable becomes insignificant, then controlling for these additional regulatory variables explains the negative and significant effect of concentration in table 2. A third reason is to simply test the relationship between the regulation of banks and financial stability. Thus, a reason why countries choose regulation is to stimulate stability. This model suggests information that sheds further light on this relationship.
Table 3 includes the additional variables of the model. This table shows that, in the first five regressions, concentration enters negatively and significantly. The significance value of the effect of concentration varies from -0.12 to -0.17. This again suggests that concentration has a substantial decreasing economic effect on the probability that a crisis will occur. More specifically, an increase in concentration level can decrease the probability of suffering a crisis from 2,5% to 3,4% according to this model. This suggests there is a negative relationship between concentration and the probability that a systemic crisis will occur, even when controlling for regulatory and freedom variables.
However, it is interesting that in the last regression the concentration variable is insignificant and the level of institutional development is significant. This means that more developed institutions enhance financial stability and the concentration level of the banking system is not related. This evidence supports the view that governance matters and thus, it is important to enhance the regulatory framework, rather than constraining bank competitiveness.
Furthermore, the regulatory variables, capital regulatory, financial freedom, economic freedom, and the quality of governance enter the model significantly and negatively. This means that improved capital regulations stimulate financial stability. In addition, these results suggest that countries with a greater financial and general economic openness to competition are less likely to experience systemic downturns in the banking sector. In addition, better institutional developments are associated with financial stability. Overall, these results support the view that regulation enhances financial stability.
