What is the impact of competition on Bank Systemic Risk?

What is The Impact of Competition on Bank Systemic Risk?

ELBERT S. WESSEMIUS∗

elbert.wessemius@student.uva.nl

29 June 2015

ABSTRACT

This paper analyzes the relationship between bank competition and systemic risk. The Contingent Claims Analysis (CCA) is discussed along with other ways to monitor systemic risk in banking. Also stated are the potential drawbacks of the CCA and Option-iPoD framework. Examining the evolution of competitive bank conditions in a cross-country level setup using firm-level balance sheet information from the period 1993 to 2013 shows that pricing power can evaporate quickly. This is signaled by the decline in the Lerner index between 1998 and 2001, and the decline in the onset of the financial crisis.

JEL classification: C51, G21, L11, L13

∗

Student BSc Economics and Business, specialization Economics and Finance, Faculty of Economics and Business, University of Amsterdam, Amsterdam, The Netherlands.

Statement of Originality

This document is written by Elbert Wessemius 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.

AFTER THE FINANCIAL CRISIS OF THE 1930s competition was restricted in many industrial-ized countries. Financial liberalization during the 1970s and 1980s however resulted in unimpeded competition, which often is named for a reason of the recent boom and bust of the sub-prime mortgage market (Beck, 2008, p.4). The global financial crisis of 2007-2008 has caused turmoil to the banking sector, debates are fuelled on issues as risk-taking, moral hazard, adverse selection and financial stability. While competition leaded to innovation and efficiency, some empirical questions on these subjects remained without concession.

One of these subjects: the influence of bank competition on financial fragility, has always been of active debate in academic and regulatory circles, however no consensus arose. The public interest also grew tremendously, questions on the topic gained in interest: do we need additional measures to ensure more competition? Or should competition be slowed down? What menaces the financial stability? gained in interest.

In order to judge the implications of the developments or the lack thereof, the question “is competition desirable for the stability of the banking sector?” is a relevant one. Especially since “one cannot manage what one does not measure” is compelling for financial stability since the empirical evidence on this issue remains mixed. Policymakers, regulators, and practitioners have to reach a consensus on how to define “systemic risk”. There is a growing similarity nevertheless that from a regulatory perspective of ensuring systemic stability, the correlation in the risk taking behavior of banks is much more relevant than the absolute level of risk taking in an individual institution.

Another inconclusive (ongoing) debate being held is the effect of the larger institutional and regulatory framework on competition in banking. In the EU banking industries for example, the process to foster competition started by deregulation of capital flows, removal of legal entry barriers, the single currency; all in favor cross-border interaction. Whether these efforts contributed to greater competition however is a view that is not generally empirically supported in favor of one of the the viewpoints (Weill, 2013).

In this paper both sets of issues are addressed. I will re-examine the relationship between systemic risk and competition. Unlike the extant literature which focused on risktaking on the bank-level, the focus lies on systemic risk. The absolute level of risk in individual banks is to examine the correlation in risk taking behavior. Secondly, the purpose of this paper is to show the evolution of bank competition. Thirdly discussed is the effect of changes and difference between countries in the regulatory and institutional framework on competition.

The Merton’s (1974) contingent claim approach pricing framework is used to model bank default risk and its contribution to systemic risk. Thereafter matured by making extensions using implied volatilities and a different framework; Option-iPoD. To investigate the issue of competition and its dependents in the banking-sector, I will use a sample of 2700 publicly traded banks in 108 countries over the period 1997 to 2013. In line with recent studies on bank competition (e.g. Claessens and Laeven, 2004, Berger et al., 2009 and Ariss, 2010), I will estimate non-structural measures: the Lerner index and I will pose the H-statistic provided by the Panzar-Rosse (1987)

model. These indicators have an advantage to measure bank behavior directly rather than inferring it from proxies. The impact of the larger institutional and regulatory environment is examined in addition.

The evolution of competitive bank conditions, calculated in a cross-country level setup, shows that pricing power can evaporate quickly. This is signaled by the decline in the Lerner index between 1998 and 2001, and the decline in the onset of the financial crisis.

The purpose of this paper is to contribute to a larger literature in order to fill the gaps in banking literature. Economy theory is conflicted on both the dependents of competition and on the impact of banking structure on the financial stability. On the one hand there could be significant stability costs of competition. Berger et al. (2004) provides an overview. Monopoly rents could give banks higher incentives to invest in relationships with smaller and more opaque borrowers (as cited in Beck, 2008) and claimed likewise: more competition leads to less incentive to monitor (Boot and Thakor, 1993; Boot,2000,p.18). Diversifying benefits the big banks and leads to the ancillary less fragility (as cited in Anginer, Demirguc-Kunt Zhu, 2014, p.2). These are examples that can be lined among the Competition-Fragility hypotheses. A main argument here is the so-called ”franchise value” hypothesis: greater market power allows banks to protect their franchise value by generating larger capital buffers, which in turn makes them act more sensible and emulate low-risk strategies (Schaeck, Cihak and Wolfe, 2009, p.2; Jimnez, Lopez, Saurina, 2013, p.1).

On the other hand, reports by the Bank for International Settlements (2001) expressed con-cerns of the high pace of consolidation and the following increasing concentration in the banking sector. The next arguments presented are classified as being in favor of the Competition-Stability hypotheses. The size and complexity of the new formed organizations could undermine regulation and supervision. And could set the government safety net into service, “too big to fail” and “too important to fail” (Miskin, 1999) are terminology that is closely related to this argument. The “quiet life hypothesis”: managers enjoying monopoly profits and tend to operate less (cost) efficient (Zhang, Jian, Qu Wang, 2013, p. 2). The higher interest set due to the pricing power -by the big banks causes the adverse selection and moral hazard to worsen, as in Stiglitz and Weiss (1981). Even in favor of competition is that it is a resolution to the hold-up problem (remark: if hold-up problem is existing: see Boot, 2000, p.20). It could mitigate the bilateral monopoly of the bank (Boot, 2000, p.8 ; Kracaw and Zenner, 1998). Ex post competition (multiple simultaneous bank relationships by the borrower) is an option, but the viability of the relationship banking may suffer (Boot,2000, p.8 p.18; Thakor, 1996; Ongena and Smith, 2000).

There could also be imperfect correlation between competition and systemic risk. Martinez-Miera and Repullo (2010, p.2) allowed for this, and found a U-shaped relationship. This theory assert that at first, the probability of a bank default decreases but to just an extent, after this point, new entry increases the probability of default.

Theoretical and empirical studies has not yet come to a conclusive finding on the impact of bank competition on financial stability (Anginer, Demirguc-Kunt Zhu, 2014, p.2) and also (logically) the topic remains widely debated and controversial, both among policymakers and academics (Claessens

Laeven, 2004, p. 218; Martinez-Miera and Repullo, 2010; Wagner, 2010). One of the biggest difficult in the discussion is the operationalizing of both stability and competition (Beck,2008,p.5; Boyd De Nicolo, 2005, pp. 1333), which could be a reason why the conclusive finding is not there yet. Large body of literature exists investigating the relationship between competition in banking markets and various proxy measures of bank risk exposure are used, drawback is that they employ often risk proxy measures that are only indirectly related to the probability of bank-failure (Boyd De Nicolo, 2005, pp. 1333).

Now, there are two different concepts that studies consider actual systematic banking distress vs. the probability of individual bank distress; the last one may not imply the first (Beck, 2008, p.16). The correlation (systematic risk) between banks has been less of a focus in the past (Anginer, Demirguc-Kunt Zhu, 2014, p.3). Two approaches to talk about systemic risk found in the literature are the single or cross-country setup. The cross-country setup is still in an early stage (Claessens Laeven, 2004, p. 564), data was previously a hindrance, which has become less of a reason to not implement, and the setup allows to examine the impact of the institutional and regulatory environment (Anginer, Demirguc-Kunt Zhu, 2014, p.3). The relationship is blurred by other variables that indirectly measure systemic risk. E.g. non-interest is positively correlated with greater bank systemic risk (Brunnermeier, Dong, and Palia (2011) and incentive compensation that is association with systemic risk (Van Bekkum, 2010). These papers do not address issues of bank competition and only consider US banks (Anginer, Demirguc-Kunt and Zhu, 2014). Competition should be measured with respect to actual behavior, not only to market structure, but also entry barriers (foreign ownership), activity restrictions, and others that limit degree of competition in an industry. Other forms of financial intermediation play a role too. Measuring competition though using concentration, is potentially not the best idea, research from the past indicate that it is a bad proxy for competition in the banking environment (Demirgui-Kunt, Laeven, and Levine (2004) as cited in Berger et. Al. (2004); Schaeck, K., Cihak, M., Wolfe, S. (2009), as cited in Claessens, Laeven, 2004). Testing the competition-stability relation requires a structural approach (Claessens Laeven, 2004, p. 219).

The other inconclusive debate has been led on the effect of the regulatory framework on banks risk-taking incentives and ultimately bank stability. A reason for abolishing restrictions on com-petitions in the 1970s and 1980s may be to the common view of markets being more effective under heavy competition. However, the relationship may vary between different regulatory and institutional framework (Beck, De Jonghe and Schepens, 2013).

This paper can contribute in the following ways. First, most of the extant literature has focused on the impact of competition and the absolute level of risk taking of individual banks, whereas this paper examines the correlation in the level of risk taking behavior of banks. Second, the causal effect of competition is discussed by investigating a different template to measure the level of risk, which is constructed by constraints derived from option market prices. Third, is possible to examine competition at the bank level, by country, and the evolution at the general level. By adopting variables proven to be associated and augment competition or systematic risk it is possible

to better the structural approach in estimating the risk-competition relationship. As a last, the cross-country data set allows to investigate the impact of regulatory and institutional environment on systemic stability.

The remainder of the paper is organized as follows. Section I discusses the data and empirical method to estimate the relationships between systemic risk, competition and the regulatory institu-tional framework and describes that which is used in this study. Section II practise the theoretical underpinnings, discusses the findings and drawbacks of the models. Section III for the concluding remarks.

I.

Empirical method and Data

A. Sets and the Sample

Boyd and De Nicolo in 2005 already highlighted that it would be ”a good idea to allow for the issuance of bank equity claims” (pp.1339-1340). They didn’t consider the extension in their research, but named it important for two reasons. First , equity claims are not insured. Their expected returns depend on default probabilities. Second, the claims are traded in the ”equity market”, and, to a first approximation, the number of competitors in this market is independent of the number of banks (Boyd De Nicolo, 2005, pp. 1339-1340). The data sources would be based on markets where claims on equity are traded: the equity, option and future market. Frequency of data is exquisite for such approach; the second moment of data series is needed. Variance of the historical prices of stocks can give a proxy for the probability of default, making the default probability indirectly observable (Wang and Xie, 2014). This approach to measure the stability-fragility relationship can be bettered, option data can provide details of contracts, which is then used to construct the constraint equations to derive the ”Option-implied Probability of Default”. Expiration time of options, prices of options are required in the algorithm. The CCA approach (as in: Anginer, Demirguc-Kunt Zhu, 2014) requires many assumptions, such as a constant rate and distribution function of assets. The Option-iPoD approach is constructed to release these assumptions. Too many assumptions is a barrier to derive suitable empirical solutions. In the next section described are the ways to derive the default risk.

The data used come from the BankScope database, obtained are bank level financial statements, data which other cross-country studies also used. There is panel data for the years 1997-2013, including all banks: commercial, saving, cooperative banks and bank holding companies. Some restrictions are imposed to avoid double-counting, that is e.g. only data from consolidated accounts. The complete sample I got hold of results in bank-year observations of 2700 banks. Winsorizing would better the results, banks for which no data is available for important variables as personnel expense or interest expense are then excluded. Next are some new selection criteria. Countries have to have an amount bank-year observations. The first sample consists of 30672 bank-year observations included are the years from 1993 to 2014, which is then reduced to 28759 to get to the final panel data for the years 1997-2013. Sample size does vary however, because not all variables

are available for all bank year observations. In one of the applications the banks in a given country should be included in the world bank DS survey. In competition and its dependents the aggregates for every country is used, consisting of 1658 country-year observation. The annual accounting data from Bankscope could be combined with stock market daily information from a database. Both international banks and US banks could be included by consulting Compustat and CRSP respectively. These two databases provide daily data for both active and delisted companies. B. Systemic default risk measures

The models to asses the credit risk or probability of default can be classified as structural or as reduced form models. The first category of structural models is based on the framework of Merton (1974) and ancillary using a model to price options. I will discuss these models in the first subsections of this section and develop the BSM option pricing methodology to suggests better attainment of the probability of default and the accompanying systemic risk measure. Reduces form models however are modeled independently of firms structure, but lacks economic intuition behind the default event; in contrast to the structural approach, the default is exogenous by a random variable.

B.1. Merton’s distance-to-default

In a leading structural approach, based on Merton (1974), the distance-to-default is measured and calculated by making use of the BSM model, assuming a stochastic process of the firm’s value. In the resulting framework, also known as the Contingent Claims Analysis (CCA), equity is treated as a call option whose underlying assets is the firms value, with a strike price equal to the promised payments. The difference between these two values is scaled by the standard deviation of the firms’ assets value. Instead of the book value of assets as approximation of a firms value, the value that the company - or in this case bank - can be sold for is used.

Balance sheet information is available on an annual basis, but the stock market information is updated daily. It is one of the forward looking measures of systemic risk. Therefore a computed distance-to-default probability is the reflected market perception of agents on bank soundness in the (near) future.

VE = VAN (d1)e−δT − Xe−rTN (d2) + (1 − e−δT)VA d1 = log(VA/X) + (r − δ + σA2/2)T σA √ T ; d2= d1− σA √ T (1)

In eq.1 VE is the market value of equity, VA value of the bank’s assets, X the face value of debt expiring in time T , r is the risk-free rate, sA the standard deviation of banks asset value, which is related and calculated by calculating the volatility of the market value of equity. Since equity itself is a function of asset volatility, the optimal hedge ratio converges, and is derived from the

Itˆo’s lemma, which gives the following equation:

σEVE = VAe−δTN (d1)sA (2)

This eq.2 is jointly calculated with the equation above to find values for VA and sA. The market value of equity and the total liabilities expiring at expiration date T are used. Since the first dataset is only available on an annual basis, one often chooses to linearly interpolate between beginning and end-year account values; something worth mentioning. The dividend rate is expressed in terms of VA. In calculating the standard deviation one requires a bank in the sample to have at least a few dozen days of non-missing returns on equity in the previous 12 months. After the asset value VAis determined, assigned is an equity premium to be equal to 6 %. Merton’s distance to default is then computed as follows:

dd = log(VA/X) + (m − d − σ 2 A/2)T σA √ T (3)

With F as the cumulative distribution function of a standard normal distribution, one can transform the distance-to-default measure to a default probability with P D = F (−dd). The 1 year US Treasury yield is taken as the risk-free rate. Two variables left which are not directly observable, the implied asset value Ai,t and the implied assets volatility sA_i,t. These two are estimated with standard iterative techniques, the system of two equations are nonlinear which makes iteratively derivation necessary. Both using the market value of equity Ei,t and its return volatility sEi,t.

However, the CCA is just one of the two main systemic risk measurement methods based on financial market data. The Option-iPoD method is also widely used. Despite their similarities there are also differences in the design, in and output parameters, empirical researches, sector applica-bility, assumptions etc. The general idea is that the two methods are better than the traditional Financial Stability Indicators (FSIs) at solutions to the problem in this paper. Investigating the relationship using z-score e.g. would be problematic since the Lerner index is also calculated using profitability measures, the positive correlation is potentially spurious (Anginer, Demirguc-Kunt Zhu, 2014). The Merton (1974) distance-to-default is already shown a good predictor of defaults, outperforming account-based models (Hillegeist, Keating, Cram, and Lundstedt 2004; Campbell, Hilscher and Szilagyi 2008; Bharath and Shumway 2008). The CCA is applying the thinking of option and the option-iPoD the price information of options.

In the next subsection I discuss the option-iPoD model and suggest improvements of the CCA-approach by releasing some of the assumptions of the CCA. The last section in this section discusses the way to examine the correlation in risk-taking; some ways to empirically estimate the systemic stability in a given country.

B.2. Option-iPoD

Christian Capuaono (2008)’s Option Implied Probability of Default (iPoD) model releases two mean assumptions: the PDF of Assets and the default barrier. The principle in this model is using maximum entropy and the related minimum cross-entropy principle to extract market-based default probabilities from prices of equity options, which is the result of the Principle of Maximum Entropy(PME).

As said, an advantage of this approach is that, except for the prior probability distribution, all constraints come from real option market prices, which are meaningful if option markets price information is effective. As options are forward-looking instruments, it brings the advantage of extracting information market participants expectations (Zer, 2013, p.101). One of the main em-pirical analysis of option-iPoD is what Capuano (2008) has done with Citi Bank. His approach with the option-iPoD is used to estimate how the default barrier and leverage ratio cultivated trough financial crisis in 2008 and therewith showing that the option-iPoD is suitable for systemic risk measurement.

The probability of default (PoD) is defined as: Z X

0

f v dv (4)

The PDF of the value of the asset is fv and the default threshold is X. To be solved is the following:

lim D f (VlimT) Z ∞ VT=0 f (VT) log f (VT) f0_(V T) dVT (5)

Both D and f (VT) is the minimization problem. Prior PDF f0(VT) represents researcher prior knowledge on f (VT), the posterior density. The cross-entropy _ff (V0_(VT)

T) is between f (V ) and f

0_{(V ),} the prior and posterior. This represents the degree of uncertainty around f (V ). f0(VT) is easily included but also excluded and not required. Constraints that determine PDF are solely driven by observable information. The minimization problem in (5) is subject to three constraints: The balance sheet constraint:

E0 = e−rT Z ∞ VT=0 max(VT − D; 0)f (VT)dVT = e−rT Z ∞ VT (VT − D)f (VT)dVT (6)

This says, that PV stock price at T should be equal to stock price observed today E0. In addition to this first constraint. Constraint f (V ) by asking posterior density to be able to price observable option prices. The second is imposed next.

The observable options pricing constraint(s): C₀i = e−rT Z ∞ VT max(VT − D − Ki; 0)f (VT)dVT = e−rT Z ∞ VT=D+Ki  VT − D − Ki  f (VT)dVT (7)

today C₀i. The number of available option contracts is indicated by i = 1, 2, ., n. If K0 then call option payoff is zero, K is the strike price. Each option price is weighted by the volume divided by the total volume of all option contracts. One trivial constraint on posterior density function, the PDF must integrate to 1. Normalization constraint:

1 = Z ∞

VT=0

f (VT)dVT (8)

So we have an objective functions and constraints, the problem is solved sequentially. First F (VT). This f (VT) will be in function of the free parameter D. Second, given optimal f (VT), we will solve for D. The Lagrangian is solved by taking the Frecht (Capuano, 2008) derivative. We end up with:

f (VT, λ) = 1 µf 0_(V T)exp[λ1e−rT1VT>D(VT − D) + n X i=1 wiλ2,ie−rT1VT>D+Ki(VT − D − Ki)] (9) µ(λ) = Z ∞ VT=0 f0(VT)exp[λ1e−rT1VT>D(VT− D) + n X i=1 wiλ2,ie−rT1VT>D+Ki(VT− D − Ki)]dVT (10) since exp[λ0− 1] = 1 µ(λ (11)

the 1x>y is corresponding to the indicator function, taking value 1 when x > y and 0 if not. To evaluate (9), the λs need to be calculated from the constraints (which are incorporated). These are given by the following system, equivalent to solving the following system:

1 µ(λ) ∂µ(λ) ∂λ1 = E0 (12) 1 µ(λ) µ(λ) ∂λ2,i = C₀i f or i = 1, 2, ...., n (13)

Solved by Newton minimization, system is nonlinear and to be solved numerically. Once a solution is found to the system of equations, I could plug the solution back into (11), to obtain:

f∗(VT, D) (14)

The default barrier, the free parameter D, is the parameter where f∗(VT, D) depends on. One can substitute eq. 17 back into original Lagrangian (9); L(f∗(VT, D)). The optimal D is determined by:

lim ∆→0

L(f∗(VT, D + ∆)) − L(f∗(VT, D))

D + ∆ = 0 (15)

This eq. 15 is also solved numerically.

Two option contracts is the minimum amount of contracts needed to solve for the density. One contract to pin down D, the other shapes density f∗(VT, D). So a data constraint is that I need

two option contracts, suppose I would have had them, on the same stock and the same expiration date, then the implementation algorithm is presented in Appendix A.

The derived PDF however is solely by observation; no preconceptions or assumptions. Therefore maximum entropy distribution closest to the true distribution. As long as the market can judge the soundness of the company (bank) in the future on aggregate in the right way. Researcher can also incorporate prior information. In this case maximum entropy is extended into the principle or minimum cross-entropy.

Some remarks with this model in comparison with the fist model. As said, the CCA approach has more input parameters and constraints; not derived from real derivative prices, which leads to a deviation from real market. The CCA approach however can be bettered by relaxing more assumptions, while the option-iPoD approach could incorporate more constraints from the market. For sectors with trading equities and options, option-iPoD provides a more accurate and effective systemic measurement approach (Wang and Xie, 2014, p.355) therefore sensible in the reseach ara of this paper. The CCA approach has a wider range of application but the option-iPoD has less manual assumptions, which makes it closer to reality and more ideal for assessing the soundness of the financial sector.

B.3. Systemic stability measures

Both models can be used examine the correlation in risk-taking of banks, which is measured here as the total variation of changes in default risk of a given bank explained by changes in the same risk of all other banks in a given country. Systemic stability is the R-squared obtained from regressing changes in the bank i default risk on changes in the average default risk of all the other banks in a particular country excluding bank i. To calculate this for each bank i in country j in week w of year t, a weekly distance-to-default is firstly computed (ddi,j,t,w). After this, I run a time series regression of for each bank i in year t.

∆ddi,j,t,w= αi,j,t+ βi,j,t 1 n n X k=1,k6=i ∆ddk,j,t,w+ εi,j,t,w (16)

Afterwards the logistic transformation of the R-squared is obtained from the regression above, as follows: log( rsqi,j,t

1−rsqi,j,t). This to measure the systemic risk posed by bank i. This is only sensible

if banks have at least an x amount weeks of changes and therefore should only be computed in this case.

Systemic risk can measured with respect to country average, a cross-country study allows to examine the impact of competition - so actually one has to do this - each country differs in their regulation and supervision, financial development etc, so to judge the implications, this is a good manner. Also often, the US is overrepresented in the sample, which is not the case if the final regressions is done only with country averages. Government safety nets implicit or explicit nets -has an impact on correlated risk-taking. And this sign of this effect is often suggested negative.

measure (CoVaR). The functional relationship among variables at different quantiles is estimated by quantile regressions. In stess periods the co-dependence of credit risk is taking account in a nonlinear way. Using a number of state variables, as in Adrian and Brunnermeier (2009), it is following runs over the sample period:

∆ddi,t= αi+ γiMt−1+ εi,t4

∆systemddt= αsystem|i+ βsystem|i∆ddi,t+ γsystem|iMt−1+ εsystem|i,t

(17)

The change in value-weighted distance-to-default in a given country is displayed as ∆systemddt. Co-depende may be best described by the R-squared measure. The ∆CoV aR may be under-estimating the fragility in good times. Billio, Pellzion, Lo and Getmansky (2012) explains that trough financial innovation, the risk codependence of financial institutions could be high but not yet reflected in substantional losses. This results may suggest that ∆CoV aR is not accurately capturing the theory.

C. Competition measures

Measuring competition though using concentration, is potentially not the best idea, research from the past indicate that it is a bad proxy for competition in the banking environment (Demirguc-Kunt, Laeven, and Levine (2004) as cited in Berger et. Al. (2004); Schaeck, K., Cihak, M., Wolfe, S. (2009), as cited in Claessens, Laeven, 2004). Hence the Lerner index is used to measure the (lack) of competition. This index is a proxy for the profits that a bank can gain as result of the pricing power, calculated by a translog cost function. The total operating cost as a function of three input prices, w_itj, with j ∈ 1, 2, 3 and an output proxy, Qi,t.

ln(Cit) = α0+ α1ln Qit+ α2(ln Qit)2+ 3 X j=1 βjln wjit+ 3 X j=1 3 X k=1 βj,kln witj ln w k it+ 3 X j=1 γjln wjitln Qit + Θ · Y earDummies + Ω · SpecializationDummies + εit (18) In eq. 24 Cit is the total operating costs (interest expenses, personnel and other administrative or operating costs). Qit the total assets of bank total assets for bank i at time t. The three input prices here capture the price of fixed assets, the price of labor and price of borrowed funds. Constructed with w1 as the share of interest expenses to total assets, w2 the ratio of personnel to total assets, and w3 is the ratio of administrative and other operating expenses to assets. The cost function is estimated for each country in the sample over the sample period from 1997 to 2013. Included are time dummies, to capture varying business cycles, and bank specialization dummies. The requirement homogeneity of degree one in input prices is imposing five restrictions on the regression: 3 P j=1 βj = 1, 3 P j=1 γj = 0 and ∀k ∈ {1, 2, 3} : 3 P j=1

βj,k = 0. The needed marginal cost is

year t: M Cit = ∂Cit ∂Qit = ∂ ln Cit ∂ ln Qit Cit Qit = Cit Qit  ˆ α1+ 2 ˆα2ln(Qit) + 2 X j=1 ˆ γjln wj_it w3 it  (19) The Lerner index is thereafter calculated:

Lernerit= (Pit− M Cit)/Pit (20)

The Pit is the price of assets and equal to total revenue divided by total assets.

An additional measure could be the Panzar-Rosse (1987) H-statistic. The Panzar and Rosse (1987) is one of the most widely used techniques to study competitive conditions in banking. The reduced-form revenue for each country is estimated for each calendar year as in Claessens and Leaven (2004). ln(Pi) = α + 3 X j=1 βjln wj_it+ 3 X j=1 γjln Y_itj+ Ω · SpecializationDummies + εi (21)

In this estimation Piis the total interest income (interest income on loans and other interest income) and divided by total assets and measures the output price of loans. Again W1,i as ratio of interest expenses to total assets, measuring input price of loans. W2,i the input price of labor; personnel expenses divided by total assets. W3,i the input price of capital, the ratio of administrative and other operating expenses to total assets. Thereby are the exogeneous control variables at the bank level, the vector xi,t, calculated by Y1,i: the ratio of equity to total assets, Y2,i: the ratio of net loans to total assets, and Y3,i: total assets in millions of USD. The natural log of every variable and ancillary regression using ordinary least squares (OLS). The statistic and conventional Panzar-Rosse H-statistics for every bank i are measured as the elasticity of revenue with respect to factor input costs - calculated by β1+ β2+ β3. The interval runs from −∞ till 1, under monopoly or the banking market in short-run disequilibrium, P RH < 0, under perfect competition, P RH = 1; and under monopolistic competition, 0 < P RH < 1.

The long-run equilibrium condition of the Panzar-rosse model needs to be checked, therefore as follows: ln(1 + ROAi) = α + 3 X j=1 βjln w_itj + 3 X j=1 γjln Y_itj+ εi (22)

The bank’s returns on assets is indicated by ROAi. The long-run equilibrium holds when the input prices doesn’t effect the return on assets, which is the case if β1+ β2+ β3 is equal to zero.

In addition, following the Perare et al. (2006), a second specification of equation (...) is estimated. Total revenue in this modified version also incorporates commission and discount income, forex income and other operating income. The original model is interest-based market model and the second as a total market model.

D. Institutional and regulatory variables

The regulatory and institutional framework of a given country is one of the variables blurring the relationship between systemic risk and a lack or excess of competition. The data to create variables to proxy this framework can come from the World Bank Doing Business Survey. Barth et al. (2008) conducted three surveys in the years 1999, 2002 and 2005 and 2010. A group can consist of a measurement of bank regulation and supervision. The Capital stringency index measures the obliged amount of capital a bank needs to keep on to, impeded by regulations. The Activity restrictions index measures the degree of allowed activity in the housing market, insurances, securities and other derivatives. The supervisory power index indicates the power of authorities to take corrective (or preventive) actions. Diversification index then measures whether there are certain explicit, doable, guidelines to which a bank needs to obey in asset management and loaning outside the borders of its country. The second group is formed to use data on credit information sharing and thereby solving adverse selection, if not shared, bank high default and the degree of investor and depositor protection. The deposit insurance coverage ratio then is derived by the amount of deposit insurance coverage divided by deposits per capita. However: this index didn’t change much up until the financial crises of 2008-2009. The third group is related to policies that restrict competition. There are surveys available regarding legal submissions to obtain banking licenses, the percentage of applications that is denied to set up a bank is the operationalization. Government ownership measures the fraction of banks that are 50% or more owned by the government.

E. Control variables

Size, concentration, behaviour, incentive structure, transparency, leverage, interconnectedness, regulation, substitutability and institution structure, cross-jurisdictional, complexity, transparency and liquidity are all systemic risk factors for securities markets (IOSCO Research Department as in: Bisias, Flood, Lo Valavanis, 2012).

In order to examine the relationships of competition, the regulatory and institutional framework, and the systemic risk, I control for a number of bank and country level variables. Variables to control for bank size, funding structure, business model and profitability are included in this essay. Bank level controls come from BankScope. For each bank, each year, I will calculate the bank size (natural logarithm of total assets), reliance on st funding, profitability (net income dividided by total assets), market-to-book ratio, non-interest income share, and provisions (loan loss provisions divided by total assets).

The actual behavior should be related not only to banking markets structure but also to entry barriers, including barriers on foreign ownership, and the severity of activity restrictions since those can limit the degree of intra-industry competition. Furthermore, the degree of competition from other forms of financial intermediation(capital markets, non-bank financial institutions, insurance companies) will play a role in determining banking system competitiveness. To date, however, few cross-country tests have taken this approach.

Therefore country level controls are collected from a number of sources. Economic development measures come from the World Bank’s World Development Indicator (WDI) database. The natural logarithm of GDP per capita in US dollars is used to measure economic development, the growth rate and the variance of it to measure economic stability, and imports plus exports and services divided by GDP to measure global integration (as in: Anginer, Demirguc-Kunt, Zhu 2014).

II.

Application of the theoretical underpinnings and

empiricability of the Model

Here I discuss the application and describe the empirical results obtained. The evolution of the average banking market power is showed in figure 1 over the years 1993-2014. This is the Lerner index calculated at the bank-year level using three input prices and then averaged per country on

−.3 −.25 −.2 −.15 −.1 −.05 0 .05 .1 .15 .2 .25 .3 Lerner 1993 1996 1999 2002 2005 2008 2011 2014

Figure 1. Evolution of bank competition: from 1993 to 2014, showed with the Lerner index over the sample period. The Lerner index is at the outset calculated at the bank-year level and thereafter averaged by country on a yearly basis. The lines connect the yearly averages of these cross-country averages.

a yearly basis. The plotted dots and the lines between them correspond to the yearly averages of these cross-country averages and the linear interpolation between them, respectively. For the purpose of showing evolution between 1993 and 2013 instead of between 1997 and 2013. In this way it is possible to parallel the evolution of banking market power to other studies. Shown is comparable to Anginer, Demirguc-Kunt Zhu (2014) and is for that reason also comparable to Beck et al. (2013) (as cited in: Anginer, Demirguc-Kunt Zhu, 2014).

The increasing of the Lerner index starting in 2001, suggests a rise in power and a decrease in competition. A reason for this could be the consolidation which concerned the Bank for Inter-national Settlements (2001) in 2001. The pace of consolidation was high and the concentration increased. Information technology and the ancillary rise in fixed costs increased economies of scales

and decreased the possibilities for entry or competition by smaller banks. The shift from tradi-tional intermediation to complex financial transactradi-tional banking also had its impact. In the more prevalent complex banking the products differ to a greater extent from person to person, this allows heterogeneity and the supplementary less price competition. Non-traditional activities are next as-sociated with new financial products as Credit Default Swaps and Collaterized Debt Oblligations, which in turn also make balance sheets more complex and creates another round of interconnected-ness. In the derivative market is rivalry with hedge funds and insurance companies, but do require high economies of scale. The complexity and the lack of transparency of the new products and the gained risk of the counterparty needed to weight in computation, makes that economies of scale and implicit government safety nets are also a necessity.

The table in the appendix gives an overview of the variables in this study and their definitions. Panel A of table I provides summary statistics for the variables used in this paper. An average bank in the sample has log total assets value of 7.272 and Lerner index of 0.146. These two numbers are comparable to those in the extant literature.

Table I Summary statistics.

Variables N P25 Mean Median P75 STD

Panel A: Summary statistics

Year 1881 2001 2005 2005 2010 6

Lerner 1199 0.134 0.146 0.233 0.324 1.144

Log (total assets) 1658 6.166 7.272 7.744 8.991 2.880

Non-deposit ST funding 1363 0.035 0.156 0.094 0.209 0.189

ROA 1655 0.009 0.020 0.017 0.030 0.047

Non-interest income share 1654 0.305 0.415 0.389 0.500 0.289

Provisions 1845 0.002 0.009 0.005 0.010 0.016

Log (GDP per capita) 1850 7.838 8.867 8.954 10.131 1.507

GDP growth rate 1623 1.700 3.690 3.800 5.900 4.160

GDP growth volatility 1642 2.124 3.683 3.148 4.415 2.168

Trade/GDP 1568 56.400 91.771 80.550 106.050 59.252

Inflation 1512 1.900 5.810 3.600 7.250 7.992

A. Competition and systemic stability: the baseline

For the purpose of estimation of the relationship, the impact of competition on systemic stability, bank and country level variables are matched and added to a regression.

riskijt= β0+ Ω · bankcontrolsijt−1+ Θ · countrycontrolsjt−1 + β1· competitionijt−1+ αi+ λt+ εijt

Table I (continued)

Year Frequency

Panel B: Sample distribution

1997 601 1998 1000 1999 1473 2000 1603 2001 1596 2002 1892 2003 1997 2004 2027 2005 2047 2006 1821 2007 1850 2008 1843 2009 1841 2010 1857 2011 1839 2012 1512 2013 1658

Sample includes banks with financial information available in Bankscope.

Panel A shows summary statistics; definitions of all those variables are in Appendix B. The sample distribution by calendar year is reported in Panel B

System risk of bank i (in country j in year t) riskijt is the dependent variable and is derived after regressing bank i’s weekly changes in probability of default on country j’s average weekly changes in probability of default in year t, hereby excluding bank i itself in the country j’s average. riskijt is then equal to the logistic transformation of the R-squared from this regression. The main explanatory variable of interest is competition, measured at first with the Lerner index of market power. A vector of bank-specific characteristics(controls) - with the reason for using this control between the brackets - include bank size (economies of scale), square term of bank size, bank non-deposit short term funding (funding structure), profitability, market-to-book ratio (bank growth opportunities), provisions to net interest income (loan riskiness), non-interest income share (business model and the efficiency).

Second a set of control variables in the regression-financial environment factors and macroeco-nomic environment factors, the countrycontrol variables. They intend to capture financial develop-ment effects. A set of the countrycontrols could include financial deepening; measured by the ratio of domestic credit to the private sector to GDP and additionally stock market turnover rate, which

is expected to capture the effect of competition from the non-banking financial sector on bank competition. The other set of the countrycontrols aims to control for the different macroeconomic environments between the countries across countries, and includes the natural logarithm of GDP per capita, variance of GDP growth rate, imports plus exports of goods and service divided GDP, and could include concentration measure: the log number of banks in each country and year. The definitions of the (country) control variables are given Appendix B. All explanatory variables are lagged by 1 year. In the regression, possibly also included are bank fixed effects, αi, to control for time invariant bank heterogeneity and use calendar year fixed effects, λt, to control for time varying global business cycle effects and are intended to capture crisis periods. Some papers follow Caprio and Klingebiel (2003) and Laeven and Valencia (2008) in the respect of adding this variable to control for time invariant bank heterogeneity.

A.1. Estimation of competition and its dependents

In this essay the focus is on competition, therefore the second step is to establish an empirical model which is suitable to study banking competition in host economies. In this subsection, ex-amined is how bank competition is impacted after controlling for bank and country level variables. The regression is specified as follows:

Competitionjt = β0+ β1Cjt+ Ω · bankcontrolsjt−1+ Θ · countrycontrolsjt−1+ λt+ εjt (24)

Different than described before, the dependent variable is competition where j indexes the country and t indexes year; competitionjt. I use the same controls as described in the beginning of the subsection. However, in a different way indicated is the country control variable of concentration. It is presented separately here from the other country variables, namely as (Ci,t). The variables included in the regression, their coefficients and p-value below their coefficients are presented

B. Competition and Regulation

The impact of competition on systemic risk may depend on the larger institutional environ-ment and may be mitigated through regulation. Anginer, Demirguc-Kunt, Zhu (2014) found that systemic risk is higher in countries with weak supervision and private monitoring, greater gov-ernment ownership of banks, and public policies that restrict competition. Investor protection and diversification regulation named also important. Therefore one can also specify the following equation:

Competitionjt = β0+ β1Cjt+ Ω · bankcontrolsjt−1+ Θ · countrycontrolsjt−1 +β2· regulationjt−1+ λt+ εjt

(25)

As in the subsubsection of the previous subsection, the dependent variable would be the aggregate bank pricing power in country j and year t, competitionijt. The same controls used in this regression

Table III Results regression

Variables (1) (2)

Log (total assets) 0.020 0.045

(0.036) (0.038) Log (total assets) squared -0.001 -0.003

(0.002) (0.002) Non-deposit ST funding -0.808∗∗∗ -0.569

(0.222) (0.232)

ROA 9.075∗∗∗

(0.854)

Non-interest income share -0.023 0.108 (0.145) (0.152)

Provisions -14.732∗∗∗ -27.185∗∗∗

(3.173) (3.109) Log (GDP per capita) -0.123∗∗∗ -0.162∗∗∗

(0.031) (0.033) GDP growth rate -0.023 -0.018 (0.010) (0.010) Trade/GDP 0.000 -0.000 (0.000) (0.001) Constant 1.323∗∗∗ 1.826∗∗∗ (0.308) (0.321) Observations 1003 1003 R-squared 0.188 0.096

Regression results of model of Competitionjt= β0+ Ω · bankcontrolsjt−1+ Θ · countrycontrolsjt−1+ λt+ εjt

*** Significante at 1% two tailed level.

as described in the previous regression. Difference is that country level bank regulation variables (regulation) is added.

III.

Concluding remarks

This paper hopefully contributes to better measurement of competition and systemic risk, by providing insight into cross-country difference and level of competition worldwide. The comparisons and suggestions for alternative measures could generate new material to develop and answer the question whether bankcompetition is good or bad for global long-term economy. The papers on the subject, also this paper, each potentially have important policy implications. It is important for antitrust authorities, supervisory authorities, politicians and policy makers to follow critically the

most recent findings, which can be different than most of the earlier literature. Anginer, Demirguc-Kunt and Zhu (2014) have results that suggest competition isn’t associated with greater systemic fragility, but also stress the importance of regulatory and institutional framework.

Problem in the whole estimation is that variables are often influencing eachother; there is the endogeneity problem in measuring bank competition. Economic growth, financial stability is not only impacted by competition, but also at their turn bank competition. Financial instability could cause defaults, which is in turn profiting other banks and thereby changing the market conduct and relationships. The endogeneity is ignored for the most part in the extant literature (Beck et al., 2010; Fernndez de Guevara en Maudos, 2011; Schaeck et al., 2009), which could be due to problems choosing the right instrument. The instrument variable should optimally be exogeneous, not containing omitted variables and should be relevant and correlate with bankcompetition. Also the connection between competition and financial stability is possibly stronger in downturns. If the economy is in upturn then competition between banks is possibly less. Also if compared to a downturn, wherein banks are competing a lot to survive, may default some of the banks, so the relationship is not equal in up and downturns and possibly stronger in a downturn.

Appendix A.

In estimating iPoD, for each bank, the following inputs are needed to generate its iPoD for a given time interval (Bisias, Flood, Lo and Valavanis, 2012).

i. European option Prices on a given date t for a given maturity T, as well as their volumes and their corresposing strike prices

ii. The stock price at the given date t.

iii. The imposed value of Vmax. e.g. calculated by projecting the time-t book value of assets to time T by using the average growth rate of the past four quarters plus some multiple of its standard deviation.

iv. The prior distribution f0(V T ), which is assumed to be uniform in N DS0 and to be zero in DS0.

The implementation algorithm is presented below:

i. Calibrate Vmax, based on book value of assets, average growth rate of these book values in the year, and its standard deviation.

ii. Start by using guess D0, dividing the domain of V in two intervals, DS0 = [0; D0] for the default state and N DS0_{= [D}

0+ ; Vmax]

iii. When VT ∈ DS0 we discretize that VT can take two values: 0 or D0, starting by f0(VT = 0) = 0 and f0(VT = D0), requires P oD(D0) = 0.

iv. When VT ∈ N DS0 we discretize by dividing domain in 100 spaces equally spaced. Prior distribution is uniform, so equal chance for all these values.

v. Solve numerically for new D, calling it D’, and f (VT, D0) for which f (VT = 0) = 0. P oD(D0) = f (VT = D0), P r(VT ∈ N DS0) = 1 − P oD(D0) = 1 − f (VT = D)

vi. Using a new NDS interval, because there is a mass of probability on the left end. Since new D, D’, is obtained, system of equations in (11)-(12) can be numerically solved for a new density So: once the solution is obtained, repeat (3)-(5), V can now only take values between DS = [0; D0] and N DS0= [D0+ ; Vmax], then find new values for f (VT, D00) and D00. vii. Only when D00= D0 = D∗ this will come to an end, so that also f∗(VT, D0) = f∗(VT, D00) =

f∗(VT, D∗) and P r(VT ∈ N DS∗) = 1 − P oD(D∗) = 1 − f∗(VT = D∗) viii. The option-implied Probability of Default corresponds to P oD(D∗)

We can also adjust this approach (Vilsmeier, 2011). The procedure of estimating P oD could have drawbacks and alternatively one could proceed by an approach based on the Lagrange multipliers. The search for the roots of a highly non-linear system of equations is transformed (in Vilsmeier, 2011) into a more stable and computationally efficient minimization problem for a strictly convex scalar function.

Appendix B.

Table III: Variable definitions

Variables Definitions

Bank level control variables

Log (total assets) Log value of total assets in millions of US dollars

Provision Loan loss provision divided by net interest income

Non-interest income share Non-interest income divided by total operating income

Non-deposit short-term funding non-deposit short term funding divided by sum of deposits and non-deposit short term funding

Return on Assets (ROA) Total income before taxes divided by total assets

Competition variables

Lerner Lerner index is qual to the difference between asset price and marginal cost (MC),

divided by asset price

H-statistic H-statistic is measured the total elasticity of revenue with respect to the input prices

Country level control variables

Log (GDP per capita) Log value of GDP per capital in nominal constant US dollars (WDI)

GDP growth volatility Variance of GDP growth over the sample period

Trade/GDP Exports plus imports of goods and services divided by GDP

REFERENCES

Anginer, D., Demirguc-Kunt, A., Zhu, M. (2014). How does competition affect bank systemic risk?. Journal of Financial Intermediation, 23(1), 1-26.

Ariss, R. T. (2010). On the implications of market power in banking: Evidence from developing countries. Journal of Banking Finance, 34(4), 765-775.

Bank for International Settlement (2001). The banking industry in the emerging market economies: Competition, consolidation, and systemic stability. BIS Paper 4.

Beck, T. (2008). Bank competition and financial stability: friends or foes?. World Bank Policy Research Working Paper Series, Vol.

Beck, T., De Jonghe, O., Schepens, G. (2013). Bank competition and stability: cross-country heterogeneity. Journal of financial Intermediation, 22(2), 218-244.

Berger, A. N. (1999). The” big picture” about relationship-based finance. InFederal Reserve Bank of Chicago Proceedings (No. 761).

Berger, A. N., Demirg-Kunt, A., Levine, R., Haubrich, J. G. (2004). Bank concentration and competition: An evolution in the making. Journal of Money, Credit and Banking, 433-451. Berger, A. N., Klapper, L. F., Turk-Ariss, R. (2009). Bank competition and financial stability. Journal of Financial Services Research, 35(2), 99-118.

Bharath, S. T., Shumway, T. (2008). Forecasting default with the Merton distance to default model. Review of Financial Studies, 21(3), 1339-1369.

Billio, M., Getmansky, M., Lo, A. W., Pelizzon, L. (2012). Econometric measures of connectedness and systemic risk in the finance and insurance sectors. Journal of Financial Economics, 104(3), 535-559.

Bisias, D., Flood, M. D., Lo, A. W., Valavanis, S. (2012). A survey of systemic risk analytics. US Department of Treasury, Office of Financial Research, (0001).

Boot, A. W. (2000a). Relationship banking: What do we know?. Journal of financial intermedia-tion, 9(1), 7-25.

Boot, A. W., Thakor, A. V. (2000b). Can relationship banking survive competition?. The journal of Finance, 55(2), 679-713..

Boyd, J. H., De Nicolo, G. (2005). The theory of bank risk taking and competition revisited. The Journal of finance, 60(3), 1329-1343.

Campbell, J. Y., Hilscher, J., Szilagyi, J. (2008). In search of distress risk.The Journal of Finance, 63(6), 2899-2939.

Capuano, C. (2008). The Option-iPoD: The probability of default implied by option prices based on entropy. IMF Working Papers, 1-29.

Caprio, G., Klingebiel, D. (2002). Episodes of systemic and borderline banking crises. Managing the Real and Fiscal Effects of Banking Crises, World Bank Discussion Paper, 428, 31-49.

Claessens, S., Laeven, L. (2004). What drives bank competition? Some international evidence. Journal of Money, Credit and Banking, 563-583.

Dijkstra, M. A., Randag, F., Schinkel, M. P. (2014). HIGH MORTGAGE RATES IN THE LOW COUNTRIES: AN INTRODUCTION. Journal of Competition Law and Economics, nhu027. Diallo, B. (2015). Bank competition and crises revisited: New results. Economics Letters, 129, 81-86.

Geppert, G., Kamerschen, D. R. (2008). The effect of mergers on implied volatility of equity options. International Review of Financial Analysis, 17(2), 330-344.

Hellmann, T. F., Murdock, K. C., Stiglitz, J. E. (2000). Liberalization, moral hazard in banking, and prudential regulation: Are capital requirements enough?. American economic review, 147-165. Hillegeist, S. A., Keating, E. K., Cram, D. P., Lundstedt, K. G. (2004). Assessing the probability of bankruptcy. Review of accounting studies, 9(1), 5-34.

Huang, X., Zhou, H., Zhu, H. (2009). A framework for assessing the systemic risk of major financial institutions. Journal of Banking Finance,33(11), 2036-2049.

Jimnez, G., Lopez, J. A., Saurina, J. (2013). How does competition affect bank risk-taking?. Journal of Financial Stability, 9(2), 185-195.

Kracaw, W. A., Zenner, M. (1998). Bankers in the boardroom: good news or bad news. Laeven, L., Valencia, F. (2012). Systemic banking crises database: An update.

Martinez-Miera, D., Repullo, R. (2010). Does competition reduce the risk of bank failure?. Review of Financial Studies, 23(10), 3638-3664.

Merton, R. C. (1974). On the pricing of corporate debt: The risk structure of interest rates*. The Journal of Finance, 29(2), 449-470.

Mishkin, F. S. (1999). Financial consolidation: Dangers and opportunities. Journal of Banking Finance, 23(2), 675-691.

Mohammad-Djafari, A. (2001). A Matlab program to calculate the maximum entropy distributions. arXiv preprint physics/0111126.

Ongena, S., Smith, D. C. (2000). What determines the number of bank relationships? Cross-country evidence. Journal of Financial intermediation,9(1), 26-56.

Panzar, J. C., Rosse, J. N. (1987). Testing for” monopoly” equilibrium. The journal of industrial economics, 443-456.

Perera, S., Skully, M., Wickramanayake, J. (2006). Competition and structure of South Asian banking: a revenue behaviour approach. Applied Financial Economics, 16(11), 789-801.

Schaeck, K., Cihak, M., Wolfe, S. (2009). Are competitive banking systems more stable?. Journal of Money, Credit and Banking, 41(4), 711-734.

Stiglitz, J. E., Weiss, A. (1981). Credit rationing in markets with imperfect information. The American economic review, 393-410.

Thakor, A. V. (1996). Capital requirements, monetary policy, and aggregate bank lending: theory and empirical evidence. The Journal of Finance, 51(1), 279-324.

Van Bekkum, S. (2010). Downside risk and agency problems in the US financial sector: Examining the effect of risk incentives from 2007 to 2010. Available at SSRN 1682139.

Vilsmeier, J. (2011). Updating the option implied probability of default methodology.

Wagner, W. (2010). Loan market competition and bank risk-taking. Journal of Financial Services Research, 37(1), 71-81.

Wang, P., Xie, M. (2014). Analysis of Two Methods of Systemic Risk Measurement Based on Option. In Digital Services and Information Intelligence(pp. 346-357). Springer Berlin Heidelberg. Weill, L. (2013). Bank competition in the EU: How has it evolved?. Journal of International Financial Markets, Institutions and Money, 26, 100-112.

Zer, I. (2014). Disclosure Practices and Option Implied Probability of Default. Available at SSRN 2335717.

Zhang, J., Jiang, C., Qu, B., Wang, P. (2013). Market concentration, risk-taking, and bank performance: Evidence from emerging economies. International Review of Financial Analysis, 30, 149-157.