Bank competition, stability, and intervention quality

• Published date: 01 January 2019
• Author: Hussein A. Hassan Al‐Tamimi,Mohamed Albaity,Ray Saadaoui Mallek,Angelos Kanas
• DOI: http://doi.org/10.1002/ijfe.1680
• Date: 01 January 2019

RESEARCH ARTICLE

Bank competition, stability, and intervention quality

Angelos Kanas

1

| Hussein A. Hassan Al‐Tamimi

2

| Mohamed Albaity

2

|

Ray Saadaoui Mallek

2

1

University of Piraeus, and Parliamentary

Budget Office, Hellenic Parliament,

Greece

2

Department of Finance and Economics,

University of Sharjah, Sharjah, United

Arab Emirates

Correspondence

Angelos Kanas, Department of Economics

University of Piraeus, and Parliamentary

Budget Office, Hellenic Parliament.

Email: akanas@unipi.gr

JEL Classification: G21; G28

Abstract

We use a flexible semi‐parametric estimation approach and a sample of 7,227

U.S., U.K., and Canadian banks for 2009–2015 to provide evidence that bank-

ing stability is non‐linearly determined by competition. We show that stability

is not monotonic against competition, and may increase and decrease at high

competition, has a mixed behaviour at medium competition, and increases at

low competition. This non‐monotonic stability behaviour at different competi-

tion levels is attributed to the intervention quality, which is found to be an

important determinant of the competition–stability relation. It is non‐linearly

related to and being revised at different competition levels. As intervention is

a policy variable, its level can be adjusted to reduce the competition effects

on stability. We illustrate that for the U.S. banking sector, the intervention

quality has to hedge these competition effects. Regulators should treat inter-

vention quality as a “hedging instrument”against the destabilizing competi-

tion effects.

KEYWORDS

banking stability, competition, intervention quality, non‐linearity, semi‐parametric estimation

1|INTRODUCTION

Recently, research on banking stability and its determi-

nants has focused on two areas: bank competition and

government intervention (Abedifar, Molyneux, & Tarazi,

2013; Ariss, 2010; Aysan, Disli, Duygun, & Ozturk, 2017;

Beck, 2008; González, Razia, Búa, & Sestayo, 2017; Mili

& Abid, 2017). The impact of competition on stability is

controversial. On the one hand, higher competition causes

lower stability because market competition erodes market

power and decreases profit margins. This makes banks less

able to endure demand or supply shocks and encourage

undue risk‐taking (Kabir & Worthington, 2017; Leroy &

Lucotte, 2017). On the other hand, because competition

pulls interest rates down, deterring moral hazard and

adverse selection behaviour among borrowers, and

decreasing loans, it may lead to greater stability (Fu, Lin,

& Molyneux, 2014; Noman, Gee, & Isa, 2017). Empirical

research has found evidence supporting a non‐linear

U‐shaped relation (Hakenes & Schnabel, 2014; Kabir &

Worthington, 2017; Kasman & Kasman, 2015; Tabak,

Fazio, & Cajueiro, 2012). The second determinant, govern-

ment intervention, has implications for bank competition

(Acharya & Yorulmazer, 2007; Ahamed & Mallick, 2017)

and for the competition–stability relation. It distorts

banks' incentives by extending liquidity support and

injecting capital or nationalizing banks. Intervention

policies against competition increase banking fragility:

Weak banks do not exit the market and evolve as

unviable banks, which may crowd out their healthy

competitors (Beck, Demirgüç‐Kunt, & Levine, 2006;

Calderon & Schaeck, 2016). Thus, a link is established

between intervention and bank stability through the

mechanism of unintended effects of intervention on com-

petition.

1

This raises the issue of the intervention quality:

Increasing intervention quality may, on the one hand,

Received: 24 May 2018 Revised: 9 July 2018 Accepted: 9 September 2018

DOI: 10.1002/ijfe.1680

568 © 2018 John Wiley & Sons, Ltd. Int J Fin Econ. 2019;24:568–587.wileyonlinelibrary.com/journal/ijfe

entail effective bank rescues but, on the other hand, may

cause destabilizing effects in the banking sector because

of its role in preserving unhealthy banks and increasing

competition.

The motivation for this study refers to employing a

new flexible non‐linear semi‐parametric econometric

model, focusing on a new large data set for three

developed countries (the United States, the United King-

dom, and Canada) from 2009 to 2015 (40,400 bank‐year

observations from 7,227 banks), and explicitly modelling

government intervention into the new econometric

approach. Thus, the present study contributes to the

existing literature in three different ways. First, the cur-

rent study looks at the non‐linearity of the link between

competition and stability without imposing any restric-

tion on the type of the non‐linearity; that is, there is no

a priori assumption set before investigating the competi-

tion–stability nexus. This is different from previous

studies that explored the non‐linearity up to the squared

level. For example, Jiménez, Lopez, and Saurina (2013),

Liu, Molyneux, and Wilson (2013), Tabak et al. (2012),

and Kabir and Worthington (2017) are among the

researchers who suggested either a U‐shaped or inverted

U‐shape relationship between competition and stability.

Hence, this study contributes to the literature by explor-

ing beyond U‐shape non‐linearity using a flexible semi‐

parametric estimation approach. This is done by using a

semi‐parametric generalized additive model (GAM) to

obtain the underlying relations between competition

and stability and between intervention quality and

competition from the data. We base our motivation to

use the GAM on the following grounds. First, in contrast

to previous empirical studies on these relations, the GAM

does not impose any restrictive linear or non‐linear

parameterization of the model and offers a flexible

semi‐parametric approach to modelling bank stability

(Kanas, Vasiliou, & Eriotis, 2012). The adoption of the

semi‐parametric non‐linear approach is motivated further

by Berger, Klapper, and Turk‐Ariss (2009), Gomez and

Ponce (2014) Pino and Araya (2013), and Boyd and De

Nicolo (2005) who argued that there are neither compel-

ling theoretical arguments nor robust empirical evidence

on the shape of the relation between banking stability

and the degree of competition. This implies that a

semi‐parametric approach is useful because it allows

the underlying relation to be obtained from the data. A

further motivation is provided by prior evidence

documenting that intervention may create complex

non‐linearities. Indeed, Claessens (2009) argues that

factors driving banking competition make for complex

competition structures and competition–stability rela-

tions. Parametric techniques try to capture such complex

non‐linearities using piecewise regression models or

adding higher order polynomials. The key issue with

these approaches is that one has to exogenously impose

the turning points of the relation and then estimate these

models (Florackis, Kanas, & Kostakis, 2015; Keele, 2008).

The GAM avoids the exogenous imposition of turning

points. Another motivating factor for using the GAM is

it is not subject to misspecification problems (Man,

2014; Man, 2015; Racine, 2008). GAM allows for linking

an outcome variable (stability) with one or more indepen-

dent variables (competition and intervention) and relies

on using smoothing splines, namely, piecewise polyno-

mials where individual pieces meet knots (Finch, 2015).

An important advantage of this methodology is it does

not pre‐specify ad hoc cut‐off points and, hence, a highly

restrictive parametric form in the relation between

competition, stability, and intervention. This avoids

severe misspecification problems that previous paramet-

ric approaches faced (Racine, 2008). A final motivating

factor for using the GAM is the large number of

bank‐year observations in the sample. This approach

accommodates in a unified model two components,

namely, a non‐parametric component and a linear

component. The former contains variables such as com-

petition and intervention quality, whose effect on bank

stability is unclear and open to alternative interpretations

(positive or negative relation). Importantly, as the “true”

relationship between each of these variables and bank

stability is “unknown,”a flexible non‐parametric

approach is more suitable than the linear model or the

restricted U‐shaped model for addressing the problem of

inferring relationships from data. The latter component

is utilized to allow for linearly‐related (control) variables

in the unified model. Using a formal testing procedure,

the semi‐parametric approach is found to be superior to

the linear parameterization. We recover and graphically

portray the empirical relation between stability and

competition, and stability and intervention, along with

the 95% confidence interval band for statistical inference.

We identify diverse relational patterns of stability at

various levels of the determinant variables of competition

for each uncovered relation, thereby signalling the exis-

tence of non‐linearity in each relation and justifying the

adoption of the non‐parametric approach. The evidence

of non‐linearity is by‐and‐large interpreted as a thresh-

old‐type behaviour of stability: When the corresponding

determinant exceeds a certain level, this variable affects

stability in a different, that is, asymmetric, way. This find-

ing is important for policy makers because, in managing

the stability of the financial system, they should target

not only specific determinant variables but also specific

threshold levels of these variables.

Second, the present paper contributes to the debate on

the relation between competition and stability and on the

KANAS ET AL.569