International Review of Applied Economics

ISSN: 0269-2171 (Print) 1465-3486 (Online) Journal homepage: http://www.tandfonline.com/loi/cira20

Empirical evaluation of ‘structure-conduct-

performance’ and ‘efficient-structure’ paradigms in
banking sector of Pakistan

Mahmood ul Hasan Khan & Muhammad Nadim Hanif

To cite this article: Mahmood ul Hasan Khan & Muhammad Nadim Hanif (2018): Empirical
evaluation of ‘structure-conduct-performance’ and ‘efficient-structure’ paradigms in banking sector
of Pakistan, International Review of Applied Economics, DOI: 10.1080/02692171.2018.1518411

To link to this article: https://doi.org/10.1080/02692171.2018.1518411

Published online: 21 Sep 2018.

Submit your article to this journal

Article views: 5

View Crossmark data

Full Terms & Conditions of access and use can be found at

http://www.tandfonline.com/action/journalInformation?journalCode=cira20
INTERNATIONAL REVIEW OF APPLIED ECONOMICS
https://doi.org/10.1080/02692171.2018.1518411

Empirical evaluation of ‘structure-conduct-performance’ and

‘efficient-structure’ paradigms in banking sector of Pakistan
Mahmood ul Hasan Khana and Muhammad Nadim Hanifb
a
Monetary Policy Department, State Bank of Pakistan, Karachi; bResearch Department, State Bank of
Pakistan, Karachi

ABSTRACT ARTICLE HISTORY

Historically, the banking sector of Pakistan has been character- Received 19 July 2017
ized by credit ceilings, directed and subsidized credit, controlled Accepted 20 August 2018
interest rates and lacking competition. This scenario prevailed KEYWORDS
till the financial sector reforms in 1990. However, after 1990, the Structure-conduct-
banking sector changed significantly and the reforms led to performance; efficient-
notable improvement in the indicators of market structure of structure; competition;
the banking industry of the country. In literature, the changes in banking sector
indicators of market structure are interpreted in the context of
JEL CLASSIFICATION
either structure-conduct-performance (SCP) or relative market
Codes: D40; E50
power (RMP) paradigm, and/or in relation to the efficient struc-
ture (ES) hypothesis. This paper studies relevance of SCP, RMP
and ES paradigms for banking industry of Pakistan. This study
uses (balanced) panel data, spanning 1996 to 2015, of 24 com-
mercial banks to estimate banks’ profit function by using fixed
effects model. Findings suggest that: (a) there is a weak associa-
tion between the indicators of market structure and banks’
performance; (b) there is no empirical evidence to support
SCP or RMP paradigms; and (c) the ES paradigm is relevant in
the case of banking sector of Pakistan. Thus, the focus of policy-
makers should be to improve the efficiency of the banking
sector in Pakistan. Focus on improving indicators of market
structure, like concentration ratios, to encourage competition
in the banking sector may not be productive.

Introduction
The traditional theory of industrial organization based on the seminal work of Mason
(1939), refined by Bain (1956), postulates a causal link from the market structure to price
setting behavior of firms (conduct), and from the conduct to efficiency of the market
participants (performance). The proponents of this ‘Structure-Conduct-Performance
(SCP)’ paradigm argue that high concentration impairs competition, which leverages
firms (banks in this study) to pocket excess profits by exercising their market power
[Mason (1939), Stigler (1964), Clark (1986), Arby (2003), Sathye (2005), Samad (2008)
and Al-Muharrami and Matthews (2009)]. The traditional paradigm suggests that small
numbers of banks in a country encourages collusive behavior by reducing the cost of
coordination, and subsequent implementation of the agreement.

CONTACT Muhammad Nadim Hanif muhammadnadeemhanif@yahoo.com

© 2018 Informa UK Limited, trading as Taylor & Francis Group
2 M. U. H. KHAN AND M. N. HANIF

Support for the SCP paradigm came from Rhoades (1997), who presented the
concept of ‘Relative Market Power (RMP)’ based on the work of Shepherd (1972).
RMP implies that higher market shares increases profitability of big institutions. Like
SCP, RMP also predicts a positive correlation between market share and profitability,
although slightly differently. Both SCP and RMP assume one-way causation from
market structure to performance or competition in the banking sector. At a policy
level, this strand of literature recommends the use of anti-concentration policies to
reign in market power and promote competition.
The analysis based on SCP and RMP assumes that market structure is exogenously
determined, and that indicators of market structure such as concentration ratios and
the Herfindahl-Hirschman index (HHI) can be used to assess the level of competition.
The most severe criticism of this approach came from the proponents of the ‘Efficient
Structure (ES)’ paradigm (Demsetz 1973; Peltzman, 1977), which took into account
causality from firms’ (banks’) performance to market structure. The proponents of the
ES hypothesis argue that efficient banks benefit from a competitive environment and
grow at a faster pace as compared to peers. As a result, the market share of efficient
banks increases over time, which ultimately contributes to market concentration
(Demsetz 1973). Like the SCP approach, it also implies that there should be a positive
association between the efficiency (profitability) and concentration in the banking
sector while the pass-through mechanism is different to the SCP approach. The policy
prescription based on the ES hypothesis also points toward the opposite direction, as
the anti-concentration policies suggested by the SCP paradigm would impair efficiency
of the banking sector by distorting the competitive environment.
Given the different interpretations of changes in market structure, it is important for
policy makers to understand the links between market structure and the performance of
the banking sector in the context of both SCP and ES paradigms. This is important for
the banking sector of Pakistan which has witnessed notable changes since the implemen-
tation of financial sector reforms. Specifically, broad-based financial sector reforms
initiated in the early 1990s paved the way for the privatization of government-owned
commercial banks and strengthened the role of the private sector. These reforms pro-
moted competition by creating a level playing field for all market players; implementing a
market based monetary management; and liberalizing the credit market by slashing
directed credit schemes and eliminating the cap on lending rates. Moreover, an effective
supervisory and regulatory framework supported the growing role of the private sector
(in line with international best practices), and an auction-based system for the sale of the
government securities was developed.
The banking sector of Pakistan witnessed significant structural changes after these
financial sector reforms (Hanif 2003). Different measures of concentration (namely the
concentration ratio, and the HHI) indicate that the market structure of the banking
sector significantly improved, moving toward a competitive environment [SBP (2003),
and (2006); Khan (2009)]. Not only has the number of banks increased considerably,
but the big five banks have conceded market share to second-tier private sector banks.
Whether these changes in the structure of the banking sector of Pakistan should be
interpreted in the context of SCP or the ES paradigm is an open research question. This
study contributes by evaluating both paradigms in the case of the banking sector of
Pakistan by using annual data from 1996 to 2015 for 24 commercial banks.
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 3

The paper is structured as follows: following this introduction, Section 2 provides a brief
review of the banking sector of Pakistan. A review of the literature is the subject of section 3.
Section 4 elaborates the methodology used to test the relevance of SCP and ES paradigms.
The data description and discussion on definitional issues relating to construction of
different variables used in this study is the subject of section 5. A descriptive analysis of
the data is provided in section 6, which is followed by estimation of a fixed effects model
and interpretation of findings in section 7. The final section concludes.

2. Banking sector of Pakistan

The banking sector in Pakistani has witnessed revolutionary changes since the partition
of India in 1947. The banking sector at that time comprised of 195 branches, primarily
dominated by non-Indian foreign banks, as one of the only two commercial banks
operated by Muslims was in Pakistani territory. The new-born country had no central
bank for almost a year, and commercial banks were facing an acute shortage of trained
human resource. To develop an efficient and sound banking system and maintain
monetary stability, the State of Bank Pakistan (SBP) was established in July 1948. The
following decades could be termed as a period of expansion for the banking sector. The
number of domestic commercial banks increased more than threefold from 5 (in 1951)
to 17 (in 1973). The branch network of these domestic banks expanded from 97 to 3123
(SBP 2004). This was followed by a period of nationalization during 1974–1978. A
number of small commercial banks merged to create big banks, owned and operated by
the government. As a result, the number of domestic banks operating in the country fell
from 17 in 1973 to 8 in 1975.
Following this nationalization of commercial banks, the structure of the banking
sector in Pakistan remained almost unchanged for the next one and a half decades.
Entry of domestic private banks was banned, to protect the nationalized banks from
competition. Administered changes in the branch networks of public sector banks, and
the rare entry of foreign banks in the financial sector, were the only sources of
expansion. Moreover, overall business activities of nationalized banks were controlled
by the Pakistan Banking Council (PBC), which was operating as a holding company of
nationalized banks (Hanif, 2003). Consequently, the number of Pakistani banks
remained unchanged at 9 from 1977 to June 1989, while their branch network saw an
expansion of only 487 branches during this period to 7,188 branches.
By the late 1980s, the performance of Public Sector Commercial Banks (PSCBs) was
significantly impaired by: (a) political interference in banking business, especially in
allocation of credit, and loan recovery; (b) a ceiling on lending rates and a floor on
deposit rates which undermined incentives for price competition among the banks; (c)
massive government borrowing at administered interest rates; and (d) other institutional
inefficiencies like unclear demarcation of regulatory and supervisory responsibilities
between PBC and SBP (Janjua, 2004). In this context, a broad-based financial reform
process was considered necessary for the development of an efficient and competitive
financial sector.
The first notable development on this front took place in 1990, when the
Nationalisation Act was amended to pave the way for the privatization of public sector
financial institutions, and to create legal provisions for opening of new private banks. In
4 M. U. H. KHAN AND M. N. HANIF

August 1991, 10 domestic private banks were granted permission to start their businesses.
In the subsequent years, three foreign and eight new domestic private banks were also
granted license to operate as commercial banks. These developments led to a sharp
increase in the number of scheduled banks operating in Pakistan during the early
1990s. Specifically, the number of local private banks jumped from 10 in June 1990 to
25 by June 1995: more than a two-fold increase in just five years.
In addition to the opening of new private banks, the government also initiated the
privatization of PSCBs. At present, four of five PSCBs operating in 1990 are privatized.
The fifth public sector bank is also partially privatized as the government has divested a
quarter of its shares to the general public and the bank is managed by a Board of
Directors. In aggregate, more than 80 percent of banking sector assets are currently
managed and controlled by the private sector.
In addition to the above developments, the banking sector of Pakistan has gone through
a wave of mergers and acquisitions since 2000. SBP (2006) noted that over 150 mergers and
acquisitions had taken place among banks and non-bank financial institutions by the end of
December 2006. Besides these developments, Islamic banks emerged as an important player
in the banking arena in the 2000s, accounting for around 10 percent of total assets of the
banking sector (Ahmed et al., 2010). Regardless of the underlying factors, these mergers
and acquisitions along with the emergence of new players have implications for the
structure, performance, and competition in the banking sector. With this backdrop, the
banking sector of Pakistan is an excellent case study to test the relevance of the SCP and ES.

3. Review of the literature

Given the strong theoretical underpinnings of both SCP and ES hypotheses, researchers
have tried to resolve this issue empirically, leading to a vast literature, as the empirical
findings differed, primarily due to: (a) differences in methodology, (b) period of study or
estimation period, (c) definitional issues, and (d) country-specific factors. In addition,
while some studies used cross-country data, many focused on a single country.
Goldberg and Rai (1996) explored the links between concentration and profitability
in the context of the SCP paradigm, using bank-wise data of 11 European countries
from 1988 to 1991 to estimate a reduced-form profit function, and two additional
equations to model the market power in the context of the ES paradigm. The results
supported one of the two versions of the ES hypothesis over the estimation period, and
not the positive relationship between market structure and profitability as envisaged by
the SCP paradigm. The results of Chortareas, Garza-Garcia, and Girardone (2007)
support the findings of Goldberg and Rai (1996). The authors estimated a profit
function by using data of 3000 banks operating from 1997 to 2005 in the 10 largest
Latin American countries, and found that market concentration lost its significance in
the presence of technical and scale efficiency in the profit function.
In contrast, Mirzaei, Liu, and Moore (2011) used bank-level data of 40 (emerging
and advanced) economies from 1999 to 2008, to explore the impact of market power on
the performance (profitability) and stability of banks, estimating a profit function
specified under the SCP paradigm. Specifically, profitability of banks (proxied by
average returns on assets, and returns on equity) was regressed on a number of
country-specific, bank-specific, and market structure indicators to disentangle the
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 5

impact of the market power. The results suggested that ‘a greater market power leads to
high bank performance’ especially in advanced economies. On the other hand, the
results did not support any paradigm (SCP or ES) in the case of emerging countries.
Given the inconclusive evidence from cross-country studies, it is necessary to review
country-specific studies. Park and Weber (2006) tested the SCP hypothesis against the
ES hypothesis by using panel data of Korean banks from 1992 to 2002. The results
supported the ES hypothesis, as the efficiency measures significantly impacted the
profitability of Korean banks. Similarly, Samad (2008) tested the validity of SCP and
ES hypotheses by using pooled data of Bangladeshi banks from1999 to 2002. The results
indicated that changes in the market structure of Bangladeshi banks could be explained
in the context of the ES paradigm. Sami and Zouari (2011) investigated the relevance of
the SCP hypothesis by using panel data comprising 10 Tunisian banks for the period
1990 to 2005. The results suggested that the market structure of the Tunisian banking
sector could be explained within the context of the RMP hypothesis. On the other hand,
the results did not support the SCP and ES hypotheses. The authors concluded that
‘Tunisian banks do not exert a monopoly power entailing the exploitation of customers,
yet they are able to extend their market share and generate profits – thanks to a
diversification of products’. Tetsushi, Yoshiro, and Hirofumi (2012) investigated the
relationship between banks’ growth and their efficiency, which can be termed as a direct
test of the ES hypothesis. The authors used annual data of 26 Japanese banks from 1997
to 2005 to estimate growth and the efficiency functions of banks. The results indicated
that efficient banks grew at a faster rate relative to other banks, which supported the
intuition of the ES hypothesis.
As results from country-specific studies differed, similarly the outcomes of past studies
conducted on the banking sector of Pakistan were divided. Arby (2003) analyzed the
structure and performance of Pakistani banks under the SCP paradigm, estimating a
profit function by using panel data of 36 banks in Pakistan from 1990 to 2000. The results
indicated that loans and capital-to-asset ratios positively impacted the profitability of
banks. The author also found ‘the absence of competitive environment in its true sense’ in
the banking industry of Pakistan. The findings related to competitive environment were
based on indicators of inequality (concentration). The study offered no insights on the
underlying factors impacting the behavior of inequality indicators. Moreover, although
the analysis was carried out under the SCP paradigm, none of the inequality (structural)
indicators was included in the profit function. In contrast to Arby (2003), Bhatti and
Hussain (2010) analyzed the impact of market structure upon the profitability of the
banking system in Pakistan by using panel data of 20 commercial banks from 1996 to
2004. The authors tested both the SCP and ES hypotheses. Their results indicated that the
profitability of banks was positively affected by the concentration in the banking system.
However, their results did not support the ES hypothesis over the estimation period. At
best, the results from Pakistan-specific studies remained inconclusive.
This paper analyzes the links between performance and market structure of the
banking sector of Pakistan. Second, the estimation is based on the most recent and the
longest panel data of the banking sector of Pakistan (for 1996–2015). And thirdly,
detailed information from the banks’ annual audited accounts allowed for refining
various indicators used for empirical analysis.
6 M. U. H. KHAN AND M. N. HANIF

4. Methodology
The traditional SCP hypothesis seems a good starting point for empirical analysis,
especially relevant as the banking sector of Pakistan gradually moved from a point of
high concentration toward a competitive market structure. Prior to the 1990 financial
sector reforms, the sector was characterized by lack of competition, high concentration,
inefficiency, political lending, lack of proper regulation and a weak supervisory frame-
work (SBP, 2000). Thus an ES explanation seems less likely in the pre-reform era, but
quite relevant in the post reform period, as a number of newly established banks
progressed to second tier banks, increasing their market share in the recent past,
while the big five banks were on the losing end, with the asset share of the top 5
banks declining from over 80 percent in the pre-reform period, to around 50 percent by
2015.1 The starting point of this study is thus the SCP paradigm, which postulates that
profitability (π) of a bank depends on the market structure:
π ¼ f ðMarket StructureÞ
Market structure may be proxied by a measure of concentration (i.e. concentration ratio
(CR) or HHI):
π ¼ f ðCRÞ
While the CR captures the potential impact of collusive behavior that facilitates banks
to earn abnormal profits, market share (MS) could be included to explore the relative
market power of the banks:
π ¼ f ðCR; MSÞ
This is expanded to include measures of efficiency (EF), to test the ES hypothesis
(Doyran 2013):
π ¼ f ðCR; MS; EF Þ
Bank-specific developments are controlled for by including relevant factors (Z) which
impact bank profitability:
π ¼ f ðCR; MS; EF; Z Þ
Following Samad (2008), Doyran (2013), and Mirzaei and Moore (2014), a reduced
form of the profitability equation utilized in this study is given below.
XK
π it ¼ β0 þ β1 CRt þ β2 MSit þ β3 EFit þ γ Z þ εit
j¼0 j jit

If β1 > 0 and β2 ¼ β3 ¼ 0, this implies that the traditional SCP holds, as market con-
centration positively impacts profitability. In the literature, testing these restrictions is
also interpreted as ‘pure collusion hypothesis’ (Sami and Zouari, 2011). If
β1 ¼ β2 ¼ 0 and β3 > 0, it indicates that efficient banks enjoy higher profitability, sup-
porting the ES hypothesis. Finally, a test of β1 ¼ β3 ¼ 0 and β2 > 0 suggests that banks
with a relatively higher market share, as compared to their peers, are more profitable,
supporting the RMP paradigm, and indirectly supporting the ES hypothesis, as efficient
banks grow faster, increasing market share.
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 7

5. Data and estimation

The primary source of data is annual audited accounts, available in the SBP library.
‘Banking Statistics of Pakistan’—a periodical publication of SBP—is the second major
source of data, available on the SBP website. In this study, a panel data of 24
commercial banks operating in Pakistan from1996 to 2015 is used to estimate the profit
function specified above.2 The use of panel data helps analyze both the time and
individual effects (Kennedy, 2008). Collecting bank data in Pakistan is a challenging
task because: (i) a number of banks merged; (ii) banks changed their names following
change in ownership; (iii) some banks were liquidated; (iv) a number of new banks
emerged in the banking sector; and (iv) some foreign banks switched from branch-
mode to locally incorporated subsidiaries. We thus assigned great importance to the
preparation of a consistent time series of key balance sheet indicators.
A comprehensive list of mergers and acquisitions among the scheduled banks
operating in Pakistan was prepared by collecting information from various issues of
Banking Statistics of Pakistan, Financial Sector Assessment Report, Banking System
Review, Financial Stability Review, and History of the State Bank of Pakistan. A
consistent time series of key balance sheet indicators were used to analyze the market
structure of the banking sector and to formally test the relevance of SCP and ES
paradigms. To ensure consistency across-banks, ‘outlier’ observations were excluded
to compile a consistent data set. Banks with negative administrative expenses, over 20
percent cost of funding, and unreasonably high labor costs, were excluded from the
sample. Most of the banks not meeting the criteria were foreign banks operating in
‘branch mode’, small private banks, and specialized banks.3 We were left with a
balanced panel of 24 banks, accounting for (on average) 90.8 percent of bank total
assets (with a range of 89.6 to 92.7 percent during the sample period).4 We used the
following definitions to proxy the variables of interest.
Profitability: A bank’s profitability is primarily measured by average return on assets
(ROA), defined as the ratio of a bank’s profit before tax to the average of assets at the
beginning and end of the year. Another measure which is widely used by investors is
the return on equity (ROE): a ratio of profit before tax to average bank equity. In this
study both the ROA and ROE are used ‘before tax’. These are denoted by ROABT and
ROEBT respectively.
Concentration: Although there are various measures of concentration in the
literature, this study relies upon CR and the HHI for the estimation of the profit
function. CR is defined as the share of k largest banks in the overall banking sector,
P
i.e. CRk ¼ ki¼1 mi where mi is the market share of ith bank. The CRk ranges from
zero (if there is a large number of equal size banks in a country) to one (if a few big
banks constitute approximately the entire banking sector). In this study k = 5. In
addition to other weaknesses, CRk assigns equal weights to the k largest banks. HHI
—which is also known as full information index—is an improvement on CRk . It
is defined as the sum of square of the market share of each bank, i.e.
P Pn
HHI ¼ ni¼1 mi  mi ¼ 2
i¼1 mi . Its value also varies from zero to one. Use of
more than one measures of market structure helps the robustness of results.
Market Share: Market share is calculated as the share of a bank’s assets in the total
assets of the banking system. A key issue was whether to use the total assets of 24 banks
8 M. U. H. KHAN AND M. N. HANIF

included in our sample or the overall assets of the banking system of Pakistan. The
study used overall assets as it ensures that the market share of ith bank is influenced by
the changes in the assets of all other banks operating in the country.
Efficiency: As the focus of this paper is to test the SCP and ES hypotheses, not to rank
banks in term of their efficiency per se, the study relies on an indirect measure:
following Samad (2008), and Mirzaei, Liu, and Moore (2011), cost efficiency of banks
is proxied by a ratio of administrative cost to average assets. Moreover, (a log of) a
bank’s ‘total assets’ is also used to analyze economies of scale (i.e. scale efficiency).
Control variables: a number of bank-specific variables are used to control for the
impact of portfolio, capitalization, funding structure etc. on the bank’s profitability,
including (i) the share of non-performing loans (NPLs) to gross loans (to take into
account the infection in bank asset portfolios), (ii) share of non-remunerative deposits
to total deposits (to account for the impact of access to low cost funding), (iii) ratio of
equity to average assets (to gauge a bank’s capitalization), (iv) share of earning assets in
total assets (as a measure of productive assets), and (v) net interest margin (to take
account of banking spreads).

6. Descriptive analysis
Table 1 reports descriptive statistics of key variables used for the estimation, showing a
wide variation in each variable over the estimation period.
ROA varies from negative 19.1 to positive 9.4 percent, which could be attributable to
one-off factors like strong provisioning requirements during a year due to a sudden
jump in (or recognition of old) non-performing loans, which itself could be driven by
factors like floods etc., and on the positive side could be windfall gains from strategic
investments, or other positive shocks. These extreme values would be difficult to explain
through bank-specific, industry-specific, or some other variables, which systematically
impact the profitability of banks. The same is the case for return on equity (ROE).5
However, it is encouraging to note that such values are few and generally related to
smaller banks.6
Net interest margin (NIM) ranges from −4.4 to 20.1 percent (Table 1). Only 14 of
480 observations (2.9 percent) are less than zero, and these pertain to a small foreign
bank operating in a branch mode. Only 2 observations have the value of NIM greater

Table 1. Descriptive statistics.

Mean Maximum Minimum Std. Dev.
Return on assets (ROA) 1.3 9.4 −19.1 2.6
Return on Equity (ROE) 12.4 244.6 −862.0 62.2
Non-performing loans to total loans (NPL) 13.0 95.5 0.1 13.8
Share of Non-remunerative deposits (SNRD) 25.6 84.0 0.3 12.8
Earning assets to total asset ratio (EATTAR) 0.8 0.9 0.3 0.1
Equity to average asset ratio (EQTTAR) 0.1 0.6 −0.2 0.1
Admin expense to asset ratio (ADETAR) 2.8 7.9 0.0 1.2
Log of total assets (LTA) 11.0 14.6 6.1 1.8
Net Interest Margin (NIM) 3.7 20.1 −4.4 2.1
HHI 0.1 0.1 0.1 0.0
Concentration Ratio (CR) 56.8 65.0 50.9 5.1
Market Share (MS) 3.8 23.8 0.0 4.9
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 9

than 10 percent. However, 108 of 480 observations (22.5 percent) indicate NIM of
greater than 5 percent. This could be one of the reasons that the banking sector of
Pakistan is generally characterized by high spreads.
The share of NPLs in gross loans varies from 0.1–95.9 percent. Around 14 percent of
observations are greater than 25 percent, which reflects significant incidence of NPLs.
This could be one of the contributory factors toward higher NIM, as provisioning
against NPLs is an expense for a bank.
Admin expense to average asset ratio (ADETAR) varies from 0.0 percent–7.9 percent,
with an average of 2.8 percent. There are only 74 observations (15.4 percent) in
ADETAR data with the ratio higher than 4 percent. In the literature, this indicator is
used as a measure of efficiency (being the cost of financial intermediation).
The correlation coefficients among the variables shown in Table 2 indicate that the
concentration ratio CR and HHI are highly correlated with each other, implying that both
the CR and HHI cannot be used in a single regression. Moreover, the log total assets (a scale
variable) have a slightly higher correlation with HHI, CR and MS. This is understandable,
as three measures of concentration are derived from bank assets. However, none of the
variables show multicollinearity, and all the correlation coefficients have the expected signs.
For example, NPLs is negatively correlated both with the ROA and ROE. This is in line with
expectations, as the increase in NPLs will undermine the profitability of the banking sector
by reducing the revenues of the banks, either by increasing provisions or by direct write-
offs. Both of which are expenses and routed through profit and loss accounts.
All indicators of market structure are positively correlated with both ROA and ROE,
implying that higher concentration is positively associated with bank profitability, as
predicted by both SCP and ES hypotheses. Similarly, the MS is also positively associated
with bank profitability; both SCP and ES hypotheses postulate a positive association
between profitability and MS. The next section analyzes which of these hypotheses
explain this positive association between profitability and MS for Pakistan.

7. Estimation and empirical findings

The regression equation specified in the methodology section is estimated by using panel
data analysis techniques. The first step is to decide on a fixed or random effects model,
driven by the nature of the variability within the subjects, the potential impact of omitted
variables, and considerations about the impact of time in-varying variables (Wooldridge,
2002, Hsiao, 2003). This study relies on the fixed effects method because: (a) it allows for
controlling unobserved time in-varying characteristics of banks; (b) the subjects (banks) in
the data remains unchanged over the estimation period; (c) data entails considerable
variation within the banks (or subjects) as the big 5 banks have significantly lost their
market share to small banks over the estimation period; and (d) all banks in the panel data
face the same macroeconomic environment, and the same regulatory and supervisory
framework. Despite strong arguments in favor of fixed effects model as discussed above,
this paper explicitly tests the relevance of fixed or random effect models by using Hausman
(1978) test. This test evaluates the null hypothesis of no misspecification in the model by
using random effects as the base model and applying Chi-square statistic. Significantly
higher estimated value of Chi-square (27.4) with 8 degree of freedom indicates strong
 10

Table 2. Correlation coefficients.

ROABT ROEBT NPL SNRD EATAR EQTAR ADETAR LTA NIM HHI CR MS
ROABT 1.0000
ROEBT 0.4725 1.0000
11.4219
NPL −0.4171 −0.2597 1.0000
−9.7785 −5.7302
M. U. H. KHAN AND M. N. HANIF

SNRD 0.1657 0.1191 0.1621 1.0000

3.5793 2.5568 3.4998
EATTAR 0.0980 0.0876 −0.0164 0.0225 1.0000
2.0979 1.8728 −0.3495 0.4789
EQTTAR 0.1086 −0.0054 −0.0937 −0.0954 −0.5343 1.0000
2.3278 −0.1156 −2.0051 −2.0418 −13.4674
ADETAR −0.3224 −0.1484 0.1513 0.1477 −0.1450 0.0798 1.0000
−7.2560 −3.1982 3.2604 3.1827 −3.1215 1.7068
LTA 0.0961 0.0791 −0.0996 0.0299 0.4471 −0.3767 −0.0584 1.0000
2.0564 1.6918 −2.1318 0.6368 10.6506 −8.6660 −1.2474
NIM 0.4815 0.2267 −0.1919 0.1349 −0.0277 0.0734 0.1789 0.1995 1.0000
11.7072 4.9595 −4.1655 2.9004 −0.5903 1.5678 3.8738 4.3390
HHI 0.0117 0.0186 0.0539 0.1702 −0.2824 −0.2214 −0.1287 −0.5099 −0.1726 1.0000
0.2501 0.3962 1.1495 3.6808 −6.2727 −4.8367 −2.7645 −12.6290 −3.7327
CR 0.0148 0.0138 0.0510 0.1632 −0.2764 −0.2123 −0.1300 −0.5058 −0.1650 0.9936 1.0000
0.3145 0.2950 1.0880 3.5247 −6.1281 −4.6279 −2.7946 −12.4944 −3.5654 187.260
MS 0.0857 0.0787 0.0659 0.1481 0.1178 −0.2910 −0.0521 0.6767 0.2052 −0.0011 −0.0008 1.0000
1.8336 1.6828 1.4066 3.1899 2.5279 −6.4799 −1.1126 19.5821 4.4679 −0.0242 −0.0173
Values beneath the correlation coefficients are t-statistics.
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 11

evidence against the null hypothesis, which is a random effects model in the Hausman test.
Thus, the subsequent empirical analysis is based on a fixed effects model.
The GLS estimates are obtained by using cross-section weights as the residual variance
for each bank will be different due to huge variation in bank size.7 Similarly, cross-section
weights are also used to estimate adjusted standard errors of the parameters to account for
the heteroskedasticity. Cognizant of strong association between HHI and CR, separate
regressions are estimated for both CR and HHI. Moreover, there is a strong correlation
between the market share and overall assets of banks as the former variable is derived from
the bank assets. However, both variables are included in the regression to test the validity of
SCP and ES hypotheses, as the total assets are included to analyze the impact of economies
of scale (efficiency), while the market share captures the impact of market power. Overall, 4
different versions (Model 1 to Model 4) are estimated and results are reported in Table 3.
Diagnostics of estimated regression equations indicate that the models are parsimo-
nious: (a) around 65 percent variation in ROA is explained by the factors included in
the regression equation, (b) standard error of regression is not only on the lower side,
but also remained largely unchanged across the different specifications for ROA and (c)
estimated coefficient of key indicators remained largely unchanged across different
specifications, indicating the robustness/stability of the results. The same is the case
for the ROE equation: explanatory power is even higher than the estimated ROA
equation and there is no visible change in results across both specifications.
The parameter estimates in Table 3 indicate that the coefficient of administration
expenses to average assets (ADETAR) is statistically significant in all four regressions.
Being a proxy for cost efficiency, it implies that cost-efficient banks enjoy higher
profitability. The coefficient of scale variable (log of total assets- LTA) is also statistically
significant for two of four regressions, providing weak evidence in favor of scale
efficiency as well. A formal test for the presence of both cost and scale efficiency
(combined restrictions on the coefficients of ADETAR and LTA) cannot be rejected
in all four specifications (Table 4). This suggests that both cost and scale efficiency play
an important role in determining the profitability of banks in Pakistan.

Table 3. Parameter estimates.

ROA ROE
Model 1 Model 2 Model 3 Model 4
Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic Coefficient t-Statistic
C −0.398 −0.141 1.574 0.449 13.289 0.447 15.296 0.422
NPL −0.069 −5.578 −0.069 −5.539 −0.656 −5.022 −0.649 −4.984
SNRD 0.021 2.668 0.021 2.789 0.205 2.426 0.207 2.437
EATTAR 4.030 3.798 3.953 3.587 43.095 3.546 43.787 3.681
ADETAR −0.926 −10.683 −0.935 −10.732 −6.986 −6.479 −7.079 −6.522
LOG(TA) 0.000 −0.001 −0.097 −0.605 −3.263 −2.141 −3.650 −2.531
NIM 4.014 7.775 4.023 8.045
EQTTAR 9.072 5.019 8.254 4.636
HHI 8.677 0.641 99.871 0.723
CR −0.001 −0.021 0.179 0.508
MS −0.082 −2.188 −0.072 −1.897 0.115 0.248 0.174 0.409
Redundant fixed 8.58* 8.81* 3.75* 3.86*
effects test
Adjusted R Sq 0.650 0.650 0.692 0.688
SE of Regression 1.944 1.944 51.796 51.555
12 M. U. H. KHAN AND M. N. HANIF

Table 4. Results of wald test.

Restrictions Model 1 Model 2 Model 3 Model 4
Coefficient of ADETAR = Coefficient of LTA = 0 63.99* 67.72* 21.78* 23.34*
Coefficient of HHI = Coefficient of MS = 0 2.52 2.48 0.62 0.44
*: Significant at 1 percent

The results also indicate that parameter estimates of both the HHI and the CR are
statistically insignificant across all specification. These results do not support the
traditional SCP paradigm, as the concentration has no statistically significant impact
on bank profitability over the estimation period. These results are in sharp contrast to
the findings of Bhatti and Hussain (2010) according to which a positive relationship
exists between commercial bank profitability and concentration in the case of Pakistan.
However, findings of this study are in line with the analysis based on simple correlation
coefficients presented above.
The coefficient of market share (MS) is statistically significant in only one of four
regressions at conventional level of significance. The negative coefficient of MS implies
that higher market share is associated with low bank profit over the estimation period.
This result is in contrast to the positive association envisaged in the RMP paradigm.
This result should be interpreted with caution as the big five banks have been losing
their market share to second tier banks in the country over the estimation period, which
may have concealed a positive effect of market power on bank profitability.
A joint test of zero restrictions on the parameter estimates of concentration and
market power variables does not support either SCP or RMP paradigms Table 4.

8. Conclusion
This study evaluates the relevance of the SCP and ES paradigms for the banking sector
of Pakistan, using a balanced panel data of 24 commercial banks from 1996 to 2015.
The descriptive analysis of data and formal tests of hypotheses suggest: (a) there is a
weak association between the indicators of market structure and bank performance; (b)
the formal tests of the hypotheses do not provide meaningful support to the SCP or
RMP paradigms; and (c) the ES paradigm proves more relevant. Indicators of cost
efficiency play a statistically significant role in determining bank profitability. These
results suggest that ‘changes in the indicators of market structure’ cannot be unam-
biguously used to analyze the nature of competition in the banking sector of Pakistan.
These results suggest that the focus of policy makers should be to improve the
efficiency of the banking sector, which plays a key role in determining overall
performance. Excessive focus on the traditional indicators of market structure such
as the concentration ratio or the HHI to improve competition in the banking sector
could prove counterproductive. As the big five banks have been losing market share
to second tier private sector banks, this implies that these big banks have been
unable to use their market power to expand their banking business. Given the more
than 50 percent share of big five banks in total assets of the banking sector, the
relatively poor performance of these banks (compared to the mid-sized private
banks) could partially conceal the impact of market power on banking business.
This study provides evidence to the banking sector regulator (SBP) that market
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 13

structure indicators cannot be unambiguously used for exploring competition in the

banking sector of Pakistan. Commercial bank management should focus on improv-
ing efficiency to increase market share.

Notes
1. While there is no consensus on the number of banks to include for calculating the
concentration ratio, in Pakistan the banking sector has been dominated by 5 banks since
the nationalization in the early 1970s. It is, therefore, instructive to calculate the concen-
tration ratio for the big 5 banks: National Bank of Pakistan, Habib Bank Limited, United
Bank Limited, MCB Bank Limited, and Allied Bank Limited.
2. Selection of time period is based on the availability of consistent data set on the subject.
3. Three specialized banks owned by the public sector are not included, as these primarily
rely on equity or borrowing (instead of deposits) for their lending activities. Moreover,
government policies heavily influence their business activities in contrast to the develop-
ments taking place in the banking system.
4. These banks include 4 public sector commercial banks, 4 foreign banks, and 16 private
commercial banks. None of the Islamic banks are included in the profit function estima-
tion as the estimation period starts before their arrival. Indirectly, all the banks operating
in Pakistan are taken into account for the calculation of the concentration ratio and HHI
since ‘overall’ banking assets has been used in the denominator. Names of individual banks
are reported in Annex 1 (Table A1).
5. Unlike ROA, the return on equity (ROE) indicates higher dispersion. It is understandable
as banking is a highly leveraged business. Specifically, equity to assets ratio is less than 10
percent (in Pakistan) over the estimation period.
6. 92.1 percent of ROA data are ± 5.0 percent. In case of ROE, only 7.5 percent of
observations are −25 percent or lower.
7. As of 31 December 2015, total assets of the biggest bank were 666 times the total assets of
the smallest bank, implying residual variance of these banks would be substantially
different, making it necessary to account for heteroskedasticity by using residual variance
of each bank as weight.

Disclosure statement
No potential conflict of interest was reported by the authors.

References
Ahmad, A., M. I. Malik, and A. F. Humayoun. 2010. “Banking Developments in Pakistan: A Journey
from Conventional to Islamic Banking.” European Journal of Social Sciences 17 (1): 12–17.
Al-Muharrami, S., and K. Matthews. 2009. “Market Power versus Efficient-Structure in Arab GCC
Banking.” Applied Financial Economics 19 (18): 1487–1496. doi:10.1080/09603100902845478.
Arby, M. F. 2003. “Structure and Performance of Commercial Banks in Pakistan”, MPRA Paper
No. 4983.
Bain, J. 1956. Barriers to New Competition. Cambridge: Harvard University Press.
Bhatti, G. A., and H. Hussain. 2010. “Evidence on Structure Conduct Performance Hypothesis in
Pakistani Commercial Banks.” International Journal of Business and Management 5 (9): 174–187.
Chortareas, G., J. G. Garza-Garcia, and C. Girardone. 2007. “Performance-Structure and Market
Power versus Efficiency in Latin American Banking.” Working Papers, Presented in confer-
ence in Department of Accounting, Finance and Management. University of Essex.
14 M. U. H. KHAN AND M. N. HANIF

Clark, J. 1986. “Single-Equation, Multiple-Regression Methodology: Is It an Appropriate

Methodology for the Estimation of the Structure-Performance Relationship in Banking?”
Journal of Monetary Economics 18: 295–312. doi:10.1016/0304-3932(86)90042-5.
Demsetz, H. 1973. “Industry Structure, Market Rivalry and Public Policy.” Journal of Law and
Economics, no. 16: 1–10. doi:10.1086/466752.
Doyran, M. A. 2013. “Financial Services Consolidation and Performance in New York State Savings
and Loan Associations, 2000–2011.” International Journal of Economics and Finance 5 (1): 8–22.
Goldberg, L. G., and A. Rai. 1996. “The Structure-Performance Relationship for European
Banking.” Journal of Banking and Finance 20: 745–771. doi:10.1016/0378-4266(95)00021-6.
Hanif, M. N. 2003. “Restructuring of Financial Sector in Pakistan.” The Journal of the Institute of
Bankers in Pakistan, no. January: 70(1): 43–74.
Hausman, J. 1978. “Specification Tests in Econometrics.” Econometrica 46: 1251–1271.
doi:10.2307/1913827.
Hsiao, C. 2003. Analysis Of Panel Data, Second ed. Cambridge: Cambridge University Press.
Janjua, A. 2004. History Of The State Bank Of Pakistan (1988-2003). Karachi: State Bank Printing
Press.
Kennedy, P. 2008. A Guide to Econometrics, 6th ed. Malden, MA: Blackwell Publishing.
Khan, M. H. 2009. “An Analysis of Degree of Competition in Banking Sector of Pakistan through
a Direct Measure of Market Contestability.” SBP Research Bulletin 5 (2): 37–52.
Mason, E. 1939. “Price and Production Policies of Large-Scale Enterprise.” The American
Economic Review 29 (1): 61–74.
Mirzaei, A., G. Liu, and T. Moore. 2011. “Does Market Structure Matter on Banks’ Profitability
and Stability? Emerging versus Advanced Economies,” Economics and Finance Working
Paper Series, Working Paper No. 11–12.
Mirzaei, A., and T. Moore. 2014. “What are the Driving Forces of Bank Competition across
Different Income Groups of Countries?” Journal of International Financial Markets,
Institutions & Money 32: 38–71. doi:10.1016/j.intfin.2014.05.003.
Park, K. H., and W. L. Weber. 2006. “Profitability of Korean Banks: Test of Market Structure
versus Efficient Structure.” Journal of Economics and Business 58: 222–239. doi:10.1016/j.
jeconbus.2005.09.003.
Peltzman, S. 1977. “The Gains and Losses from Industrial Concentration.” Journal Of Law and
Economics 20: 229–263.
Rhoades, S. A. 1997. “Research on IO Topics in Banking: An Introduction and Overview.”
Review of Industrial Organization 12: 1–8. doi:10.1023/A:1007769127555.
Samad, A. 2008. “Market Structure, Conduct and Performance: Evidence from the Bangladesh
Banking Industry.” Journal of Asian Economics 19: 181–193. doi:10.1016/j.asieco.2007.12.007.
Sami, M., and A. Zouari. 2011. “Banking Industry, Market Structure and Efficiency: The Revised
Model to Intermediary Hypotheses.” International Journal of Economic Research 2 (1): 23–36.
Sathye, M. 2005. “Market Structure and Performance in Australian Banking.” Review of
Accounting & Finance 4 (2). doi:10.1108/eb043425.
SBP. 2000. Pakistan: Financial Sector Assessment 1990–2000. Karachi: State Bank of Pakistan.
SBP. 2003. Pakistan: Financial Sector Assessment 2003. Karachi: State Bank of Pakistan.
SBP. 2004. Banking Statistics of Pakistan 2003–04. Karachi: State Bank of Pakistan.
SBP. 2006. Financial Stability Review 2006. Karachi: State Bank of Pakistan.
Shepherd, W. G. 1972. “The Elements of Market Structure.” Review of Economics and Statistics
54: 25–37. doi:10.2307/1927492.
Stigler, A. 1964. “A Theory of Oligopoly.” Journal of Political Economy 72: 44–61. doi:10.1086/
258853.
Tetsushi, H., T. Yoshiro, and U. Hirofumi. 2012. “Firm Growth and Efficiency in the Banking
Industry: A New Test of the Efficient Structure Hypothesis,” RIETI Discussion Papers Series
12-E–060.
Wooldridge, J. 2002. Econometric Analysis Of Cross Section and Panel Data. Cambridge, MA:
MIT Press.
 INTERNATIONAL REVIEW OF APPLIED ECONOMICS 15

Annex 1. Banks included in the panel used to estimate the profit function

Table A1. Banks in the balanced panel.

Public Sector Commercial Banks
1 National Bank of Pakistan
2 First Women Bank Ltd.
3 The Bank of Punjab
4 The Bank of Khyber
Local Private Banks
5 Allied Bank Ltd.
6 Bank Alfalah Ltd.
7 Askari Bank Ltd
8 Bank AL Habib Ltd.
9 SAMBA Bank Ltd.
10 Faysal Bank Ltd
11 Habib Bank Ltd.
12 MCB Bank Ltd.
13 United Bank Ltd.
14 Habib Metropolitan Bank Ltd
15 Soneri Bank Ltd
16 Silk Bank Ltd
17 NIB Bank Ltd.
18 Standard Chartered Bank (Pakistan) Ltd
19 Summit Bank Ltd
20 JS Bank Ltd.
Foreign Banks
21 Al-Baraka Bank*
22 Citibank N.A
23 Deutsche Bank AG
24 Bank of Tokyo—Mitsubishi UFJ, Ltd.
*: Reclassified into Local Private banks