Testing for Competition in the Nigerian Commercial Banking Sector

Transcription

1 Modern Eonomy, 2013, 4, Published Online July 2013 ( Testing for Competition in the Nigerian Commerial Banking Setor ufus Adebayo Ajisafe *, Anthony Enisan Akinlo Department of Eonomis, Obafemi Awolowo University, Ile-Ife, Nigeria * rajisafe@oauife.edu.ng eeived April 9, 2013; revised May 9, 2013; aepted June 9, 2013 Copyright 2013 ufus Adebayo Ajisafe, Anthony Enisan Akinlo. This is an open aess artile distributed under the Creative Commons Attribution Liense, whih permits unrestrited use, distribution, and reprodution in any medium, provided the original work is properly ited. ABSTACT The study determined the degree of ompetition in the banking setor between 1990 and 2009 using Panzar and osse (P) methodology. The data for the study were obtained from the annual reports and statement of aounts of fifteen ommerial banks in Nigeria whih were purposively seleted for the study. The data olleted were analysed using dynami panel generalised method of moment estimation tehnique with fixed effet. The results of the analysis showed that the Nigerian ommerial banks were haraterised by monopolisti ompetition with H-statisti signifiantly different from zero for all sample periods and sub-sample periods. The value of H-statisti ranged between and The study onluded that the banking industry in Nigeria exhibited monopolisti ompetition whih supports the results obtained from previous studies in the developed eonomies. Keywords: Banking; Competition; P Model; Dynami Panel; H-Statisti 1. Introdution The Nigerian banking system has experiened some fundamental hanges sine independene. At independene, the banking markets were dominated by a relatively small number of foreign banks. After about three and half deades, the number of banks expanded and the ownership struture diversified. Initially, it was dominated by the publi setor banks, and later (1992 to be speifi), the private setor banks beame the dominant partiipants. As at this period, the government intervened in the banking markets to ontrol resoure alloation and promote the indigenization poliy of the eonomy. The poliies of finanial repression pursued by the government direted at the Central Bank of Nigeria (CBN) during this period, was to ontrol interest rates, volume and diretion of redit in the eonomy [1]. Less attention has been paid to the issue of ompetition in the finanial setor but suh ompetition matters for a number of reasons. In the first plae, the degree of ompetition in the finanial setor an affet the effiieny of the prodution of servies, the quality of produts, and the degree of innovation in that setor. Seondly, it has been observed theoretially that the degree * Corresponding author. of ompetition in the finanial setor an affet the aess of firms to external finaning [2]. Also, less ompetitive banking an be more ostly thereby lowering the quality of servies whih invariably redues the effetive demand for external finaning and thus redues growth. These effets may further vary by the degree of ompetition in the ountry over finanial setor [3]. The impat of ompetitive and regulatory hanges on banks an be judged by gross measures of performane (effiieny) suh as profitability and failure rates, but how suh hanges an affet the effiieny with whih banks transform resoures into various finanial servies are of vital importane to the eonomists. The development in tehnology has shaped the way in whih banks arry out their business. That is, a new and improved tehnology in the banking system is expeted to redue bank ost over time. This has failitated development of new, more sophistiated finanial produts as well as the introdution of alternative delivery hannels to the traditional branh network [3]. This in onjuntion with deregulation has intensified the finanial setor ompetition in the industrialized nation. The question that arises from the above is that: what is the degree of ompetition in the Nigerian banking setor? The study intends to know whether the reent legal and institutional reforms in the

2 502. A. AJISAFE, A. E. AKINLO setor were suffiient to transform the market struture into a more ompetitive mode or whether there are still some serious obstales inherited from the earlier system that prevent the realization of ompetition. Thus, the objetive of this study is to determine the degree of ompetition in the Nigerian banking setor. In the remaining part of this paper, Setion 2 fouses on the literature review, Setion 3 disusses the model used in the study; while Setion 4 gives the result of the analysis and Setion 5 onludes the study. 2. Literature eview eplete of literature abound on the degree of ompetition in the banking setor among whih are Prasad and Ghosh (2005), Panzar and osse (P) (1987), Bikker and Haaf (2001) and Claessens and Laeven (2004) among others. [4] analysed whether ompetition has yielded signifiant benefits in terms of greater produt sophistiation and ost redution. This was done by examining the degree of ompetition in the Indian banking system for the period and the two-sub periods, and The study was arried out with the pereption that ompetition in Indian banking setor has inreased sine the introdution of the finanial setor reforms in The estimated model used interest revenue and total revenue as dependent variables and employed P methodology in its estimation due to its advantage of using bank speifi data and hene aptures the unique harateristis of different banks. The results point to monopolisti behaviour of banks aross time periods and aross bank groups, with a more robust H- statisti for the seond sub-period and for private and foreign banks. [5] argued that a higher degree of banking ompetition is a major issue for eonomi development and it is expeted to provide welfare gains by reduing monopoly power of banks and ost ineffiienies, favoring the redution of loan rates and then investment. The study used quarterly data for Czeh banks, in order to provide evidene on the effets of banking ompetition in the Czeh epubli. In the first plae, the study measured the level and the evolution of banking ompetition between 1994 and 2005, using both the traditional IO approah and the new empirial IO approah. The traditional approah proposed strutural test while the new empirial approah proposed the non-strutural test. Aording to the traditional approah, ompetition was measured by onentration indies suh as the market share of the five largest banks or the Herfindahl index alulated for total bank assets and loans. This approah suffers from the fat that they infer the degree of ompetition from indiret proxies suh as market shares. The new empirial IO approah infers banks ondut diretly. The new ap- proah also takes into onsideration the measure of ontestability. The approah therefore uses the osse-panzar model whih is a non-strutural test to measure the degree of ompetition using the H-statisti whih is the sum of the elastiities of total revenues to input pries. The authors find no improvement in banking ompetition during the transition period. [6] examined ompetitive onditions and market struture in the banking industry, and investigate their interrelationship. That is, the study tested empirially the relationship between onentration and ompetition. Also, the study seeks to measure the degree of ompetition in the European banking markets. As a result of the defiienies of the strutural models, the study developed the non-strutural models of ompetitive behaviour and applied the Panzar and osse (P) model to measure the degree of ompetition without using expliit information about the struture of the market. The P model assessed the elastiities of interest revenues with respet to hanges in banks input pries. In order to distinguish ompetitive behaviour on loal, national and international markets, for eah ountry, three sub samples were taken: small or loal banks, medium-sized banks and large or international banks. For all the 23 ountries onsidered in the estimation, the result showed the existene of monopolisti ompetition in the banking industry. [7] examined what drives bank ompetition and was of the opinion that ompetition in the finanial setors matters for a number of reasons. They observed that the degree of ompetition in the finanial setor an matter for the effiieny of the prodution of finanial servies, the quality of finanial produts and the degree of innovation in the setor. They used bank- level data, and applied the P methodology of ompetitiveness to estimate the extent to whih hanges in input pries are refleted in revenues earned by speifi banks in 50 ountries banking systems. The ompetitiveness measure was then related to indiators of ountries banking system strutures and regulatory regimes. The study found out that systems with greater foreign bank presene and fewer ativity restritions in the banking setor were more ompetitive and that entry restrition on ommerial banks ould redue ompetition. The study equally found out that more onentrated banks are more ompetitive. The above findings onfirmed that ontestability determines effetive ompetition espeially by allowing (foreign) bank entry and reduing ativity restritions on banks. While the result suggested that struture is less important for the ompetitive behaviour of the banking setor, it shows that ompetition poliy in the finanial setor is more ompliated than perhaps previously thought. In part, this might be as a result of the finanial servie industries undergoing rapid hanges, triggered by deregulation and tehnologial advanes. These hanges have

3 . A. AJISAFE, A. E. AKINLO 503 made the definition of a finanial market and any partiular finanial servie more omplex, and may have made market struture indiators less valuable measures of the ompetitive nature of finanial systems. Developing proper ompetitiveness test and methodologies will remain an important area of researh and poliy fous. [8] examined the effet of the hanges in the struture of the Jamaian eonomy on the banking industry. This study is relevant as a result of the onsolidation trend in the industry following the finanial risis during this period. It was observed that inrease in market onentration has signifiant impliations for the level of ompetition as well as the welfare of the ustomers in the banking setor. In analysing the hanges in the level of onentration and ompetition in the banking industry over the period, the study used the Herfindahl-Hirshman Index. The results from the analysis revealed that there was a slight inrease in ompetition following finanial liberalisation. It is to be noted that after the finanial risis in the mid 1990s, the industry beame more onentrated whih might suggest that there was a deline in ompetition among the banks. However, as a result of the ambiguity in the result obtained from the strutural model, the study used a more robust P methodology to measure the market power of the industry. The result from this alternative methodology revealed that ompetition fell slightly immediately following the liberalisation period. Furthermore it was revealed that the hypothesis of monopoly and perfet ompetition were both rejeted in favour of monopolisti ompetition for the entire sample period. The interation term used in apturing the hanges in market power over time indiated that there was a steady deline in ompetition throughout the speified sample period. [9] assessed the degree of bank ompetition and disussed effiieny with regard to banks finanial intermediation in Ghana. In the study they applied panel data to variables derived from a theoretial model and find support for the presene of a nonompetitive market struture in the Ghanaian banking system, possibly hampering finanial intermediation. The eonomi osts of the nonompetitive behaviour might have been exaerbated by the persisting domesti finaning needs of the government, making it aptive to the banks behaviour and fostering ineffiieny in the banking system. Also, large defiit finaning through the issuane of treasury bills has not only rowded out the private setor in apturing banks investments, but has also put pressure on interest rates, thereby making aess to bank lending even more diffiult for the private setor thus hampering private setor development. Therefore, further private setor development appears to be very muh dependent upon sound fisal adjustment, and the possible link between fisal poliy and the effiieny of the banking sys- tem should deserve further attention. The result of the study further indiated that onsolidation of the Ghanaian banking setor is expeted due to sale matters. Furthermore, barrier to ompetition on interest revenue is an indiation that ompetition is stifled in the Ghanaian banking system. This ould be as a result of the non transparent fee struture of the banks whih help to shield the bank market struture from ompetition. Following from here, there is the need for further study in the area of ompetition and effiieny in the banking industry. [10] analysed the ompetitive nature of the Tanzanian banking industry from 2004 to Utilizing a rih bank level data set, the study employed the P methodology to ompute the ompetitive index, taking into aount risk, effiieny, regulatory and maroeonomi fators. The result showed that banks in Tanzania earned their inome under onditions of oligopolisti ondut. Moreover, the ompetitive index derived from an interest revenue equation was not signifiantly different from that obtained using an aggregate revenue measure. This suggests that the degree of ontestability from traditional intermediation ativities approximates overall bank behaviour. The overall message is that greater market ontestability an be ahieved by adopting measures aimed at stimulating ompetitiveness in the banking setor, inluding onsolidating gains on the maroeonomi front and allowing more foreign bank entry so as to inrease the spread of banking servies. [11] investigated the degree of bank ompetition in the euro area, the US and UK before and after the reent finanial risis, and revisits the issue whether the introdution of EMU and the euro have had any impat on bank ompetition. The results suggested that the level of bank ompetition onverged aross euro area ountries in the wake of the EMU. The reent global finanial risis led to a fall in ompetition in several ountries and espeially where large redit and housing booms had preeded the risis. The result obtained showed that the degree of ompetition in the US and UK banking setor support monopolisti ompetition. This orroborate the work of [7,12] whih applied this method to a large sample of ountries, finding evidene of monopolisti bank ompetition with varying degree aross.ountries. [13] assessed the degree of ompetition and relative effiieny of the FY Maedonia s banking system a setor whih has undergone a substantial amount of hange sine the mid-1990s. In their analysis, P methodology was adopted to test for the degree of ompetition using quarterly data for the period for 20 ommerial institutions in Maedonia. In general, the results obtained show that ompetition in the banking setor remains relatively weak. From the literature review, it was observed that major-

4 504. A. AJISAFE, A. E. AKINLO ity of the banks in various ountries exhibited monopolisti ompetition. To buttress the above literature, [14] found monopolisti ompetition for New York banks, [15] in his study of European banking using data for found monopolisti ompetition for Spain, UK, Frane, Germany and monopoly for Italy. Also, [16] using data obtained monopolisti ompetition for Frane, Germany, Italy and the US. The degree of ompetition has been tested in the European area, US and to some extent in Afria but to the best of our knowledge, it has not been tested in the Nigerian banking setor. Thus, there is the need for the study. 3. The Empirial Model and Tehnique of Analysis 3.1. Empirial Model To measure the degree of ompetition, the study adopted the methodology of Panzar and osse (1987). The P methodology developed from a general equilibrium market model relied heavily on the premise that ompetition is measured by the extent to whih hanges in fator input pries are refleted in firms equilibrium revenues [17]. Let us onsider the revenue and ost relationship faing a partiular bank: Thus, y, n, z (1) i i i i i C C y, w, (2) i i i i Where: i = Total revenue of bank i; C i = Total osts of banks i; y i = Output of the bank i; n = Number of banks; w i = Vetor of fator input pries of bank i; zi = is the vetor of J exogenous variables affeting the revenue funtion; e zi = is the vetor of L exogenous variables affeting the ost funtion. From Equations (1) and (2), profit is defined as: П y, n, C y, w, (3) i i i i i i i For profit to be maximized, marginal revenue must equal marginal ost [18]. Thus we differentiate Equation (3) with respet to the revenue and ost funtion and equate to zero as shown: di dci i. 0 d y, n, dc y, w, i i i i i 1,, Ci yiw i y n (4) 1 i i i From Equation (4), the profit maximization ondition holds at the market equilibrium level. Given that the profit maximizing output level representing the equilibrium value is defined as: y y, w, (5) i i i i Substituting Equation (5) into Equation (1) and assume that n is endogenously determined in the model, then Equation (1) beomes: y, w, i i i i Equation (6) is the redued form for revenues of the representative bank whih is the produt of the equilibrium output of bank i and the ommon prie level. It is important to note that market power is measured by the extent to whih a hange in fator input pries ( w ) is refleted in the equilibrium revenue ( ) earned by the bank and the measure of ompetition defined as H-statisti formulated by P evaluates the elastiity of total revenue with respet to hanges in fator input pries [4,9,17]. K di W ki H. k 1dWki i where k is the number of fator input pries used, representing the prie of apital, prie of labour and prie of fund respetively. Empirially, if we write Equation (4) in mathematial form, and follow the work of [9] we have, and J 1 i 0 1 yi d j ji Ln Ln Ln J 1 K L 1 Ci 0 1 yi bk Wki Vi li Ln Ln Ln Ln J 0 1Ln yi djln j1 0 1 K1 l1 The natural logarithms of the variables used in Equation (8) are taken, beause it is assumed that, time series data have overall trends of exponential growth. For profit maximizing bank, M = MC, therefore Equation (8) beomes K L yi bk WKi Vl li Ln Ln Ln ji K1 l1 J 11Ln yi djln ji j1 k L 0 0 bkln Wki VlLn li k1 l1 (6) (7) (8) (9) earranging Equation (9) and olleting the like terms gives (10)

5 . A. AJISAFE, A. E. AKINLO 505 Ln yi 0 0 L J VlLn li l1 j1 k k k 1 djln * * i ji b Ln W Ln Ln y w. (12) K K 1 k ki W Ln Y i i (11) Equation (11) is a mathematial representation of Equation (4) written in log linear form. It ould be observed from Equation (6) that the redued form equation for the revenue funtion is the produt of the equilibrium output of the bank and its ommon pries [9,18] written in logarithm form as shown in Equation (12). Operationalising Equation (12) gives: Ln K it kln ki K 1 Q q1 Ln qi (13) where z is a vetor of exogenous and Q bank speifi harateristis without referene to their origin, either from ost or revenue funtions. Y * is a sale variable whih represents the output of the bank i (Bank size), W ki is a vetor of fator input pries, representing the prie of labour, prie of apital and prie of fund respetively and * is the equilibrium revenue saled by the total asset of the bank. From Equation (13), we now evaluate our H- statisti, whih is now defined as: H B (14) where B k represents the oeffiients of the three dimensional fator input pries defined earlier on. The sign and magnitude of H-statisti matters in its interpretation. In a monopoly market struture, an inrease in fator input pries (W i ) will inrease marginal ost thereby reduing equilibrium output (Y * ) and invariably redue the total revenue generated in that market. This implies that the value of H-statisti is less than or equal to ero (H < O). For a perfetly ompetitive market, an inrease in fator input pries will inrease marginal osts as well as average osts by the same proportion. This may not have any effet on the equilibrium output of banks. As a result, ineffiient banks are fored out of the market. This made the remaining firms to fae inreased demand whih eventually leads to an inrease in output pries and revenue in the same proportion as osts. The value of H- statisti is then equal to unity (H = 1). In the ase of monopolisti ompetition, an inrease in input pries will lead to a less than proportional inrease in revenue, as the demand for banking faing individual banks is inelasti. The value of H-statisti lies between 0 and 1 [8,9,17,19]. It has been observed that one of the assumptions of the methodology of Panzar and osse that makes it valid for further analysis is that, the banking setor is assumed to be in equilibrium. In line with this assumption, an equilibrium test will be performed using Equation (15) (see [3,7-9]. it k Wki Yi Ln 1 OA Ln Ln K 1 Q q1 K Ln qi i (15) OA is the unadjusted return on assets. (1 + OA) is used as the dependent variable instead of OA for the sake of onveniene, sine it is believed that OA an take on small or negative values. The result of the equilibrium test obtained from equation 15 is then tested using F-statisti that the sum of E = β 1 + β 2 + β 3 = 0. 1 The main reason for this test is to be sure that, in equilibrium, returns on bank assets should not be statistially orrelated with input pries Tehniques of Analysis In order to estimate our models in Equations (13) and (15), eah of the variables used in the model are tested for stationarity using panel unit root test [20-23]. This is neessary beause of the nature of our time series data. It has been observed that most of our time series data are not stationary at levels and that the series are adequately represented by first differene [24]. [25] developed a proedure utilizing pooled ross setion time series data to test the null hypothesis that eah individual time series ontains a unit root against the alternative hypothesis that eah time series is stationary. As both the ross setion and time series dimensions of the panel grow large, the panel unit root test statisti has a limiting normal distribution. They onluded that the use of panel unit root tests may prove to be partiularly useful in analyzing industry level and ross ountry data. It was also observed that the pooling approah or panel based unit root tests yields higher test power than performing a separate unit root test for eah individual. Various approahes have been used in performing panel unit root tests. This inlude: [23,25,26], Fisher-type tests using ADF and PP test, see [27-29]. In the proess of estimating the models, we first employ the panel pooled least square estimate and then used the panel GMM to test for the robustness of the model. From Equation (13), we determine the degree of ompetition using the value of H-statisti, while Equation (15) is an equilibrium test used for the validity of H-statisti. 1 β 1, β 2, and β 3 are the oeffiients of the fator input pries w 1, w 2, and w 3.

6 506. A. AJISAFE, A. E. AKINLO 3.3. Measurement of Variables In this setion, the variables used in the estimation of our models are given appropriate definition. They are defined as used in Equations (13) and (15): * is the equilibrium total revenue of the bank. It is proxied by the gross earnings saled by the total assets of the bank. It is denoted by Geta in the model whih represents the dependent variable. The total revenue of the banks is onsidered beause the value of non-interest inome has inreased over the years. This view is supported among others by [4,9, 14]. In other to aount for size differenes, the total revenue is divided by total assets. w ki is a vetor of fator input pries, representing the prie of labour, prie of apital and prie of fund respetively. In the estimated model, w 1 represents the prie of labour whih is measured as a ratio of personnel expenses to total asset or personnel expenses as a ratio of number of employee. In this study, the ratio of personnel expenses to total assets is onsidered as our indiator beause there are missing data for the number of staffs for many banks. w 2 represents the prie of apital whih is the ratio of apital expenses to fixed asset, while w 3 represents the prie of fund whih is the ratio of interest expenses to total deposits [2,8,17]. The output of the firm is measured by the amount of loans. The intermediation approah onsiders banks as finanial intermediaries that onvert deposits and borrowed funds into loans and investment. In this study, we onsider only loan as our output variable whih is measured in value terms. Other variables inluded in the study are bank speifi fators whih reflet differenes in osts, size, risk, struture and produt mix. The bank speifi variables inlude: ratio of equity to total assets represented by r 1, ratio of loans to total assets represented by r 2, ratio of non-performing loans to total loans represented by r 3, the ratio of total deposit to total assets represented by TDTA, and total assets whih is the sale variable represented by TA, ontrols for size of the bank and proxy for sale eonomies. 4. Empirial esults 4.1. esult of Panel Unit oot Test The result of the Panel unit root tests are presented in Tables 1 and 2 respetively. Table 1 presents the result of the panel unit root test with individual effets while Table 2 presents the unit root test with individual effets and linear trends. All the variables used in the estimation of our model are in their log form. The two results are reported for omparison purposes. It ould be observed from the tables that all the variables are stationary at levels. That is, they are integrated of order zero I(0). Given the unit root property of the variables, we then proeed to the estimation of our model aross full sample and Table 1. Panel unit root test with individual effets. Variables LLC Im, PS ADF PP Geta w 1 w 2 w 3 r 1 r 2 r 3 TDTA TC TA OA (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.0001) * (0.0000) * (0.0002) * (0.0002) * (0.0000) * (0.0000) * (0.0000) * (0.0000)* (0.0000) * (0.0000) * (0.0001) * (0.0000) * (0.0000) * (0.0000) * (0.0000)* (0.0001) * (0.0060) * (0.0000) * (0.0000) * (0.0000) * (0.0058) * (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.0000) * (0.7607) 2, (0.0000) * (0.9928) (0.0000) * (0.0000) * (0.0000)* (0.2171) (0.0000) * (0.5655) *** (0.2388) ** (0.0000) * * 1%; ** 5%; *** 10%. This indiates rejetion of null hypothesis of unit root. Probabilities for Fisher tests are omputed using an asymptoti Chi-square distribution. All other tests assume asymptoti normality. LLC Levin, Lin and Chu, Im, Ps Im, Pesaran and Shin, ADF Augmented Dikey Fuller, PP Phillip Peron. The probability values are shown in parenthesis. Table 2. Panel unit root test with individual effets and individual linear trends. Variables LLC Im, PS ADF PP Geta (0.0000) * (0.0000) * (0.0000) * (0.0000) * w (0.0000) * (0.0011) * (0.0016) ** (0.0001)* w (0.0000) * (0.0223) ** (0.0326) ** (0.0060) ** w (0.0000) * (0.0000) * (0.0000) * (0.0000) * r (0.0000) * (0.0000) * (0.0000) * (0.0000) * r (0.0000) * (0.0000) * (0.0002) * (0.0699) *** r (0.0101) * (0.0004) * (0.0032) ** (0.0000) * TDTA (0.0000) * (0.0000) * (0.0000) * (0.0000) * TC (0.0008) * (0.0497) ** (0.0740) *** (0.5257) TA (0.0000) * (0.0000) * (0.0001) * (0.0012) * OA (0.0000) * (0.0000) * (0.0001) * (0.0000) * * 1%; ** 5%; *** 10%. This indiates rejetion of null hypothesis of unit root. Probabilities for Fisher tests are omputed using an asymptoti Chi-square distribution. All other tests assume asymptoti normality. LLC Levin, Lin and Chu, Im, Ps Im, Pesaran and Shin, ADF Augmented Dikey Fuller, PP Phillip Peron. The probability values are shown in parenthesis.

7 . A. AJISAFE, A. E. AKINLO 507 sub-sample periods using both the pooled least square estimate and the GMM tehniques esults of Competition in the Banking Industry After the unit root test has been performed, and found out that our variables are stationary at levels, the model in equation 13 whih is linear in its unknown parameters are then subjeted to empirial investigation using panel data with fixed effets to aount for any heterogeneity among the industry (banks) as well as to avoid speifiation problems. The ommon effet speifiation has been hosen for our model beause the firms operating in the industry are ountry speifi and they are likely to share the same harateristis. The model in Equation (13) is estimated using the Pooled Least Square and the result is presented in Tables 3-5. It ould be observed from the result presented that the banking setor in Nigeria exhibited a monopolisti ompetition over the entire sample period based on the value of H-statisti whih lies between zero (0) and one (1). This is suffied to say that the null hypothesis of perfet ompetition and monopoly is rejeted at one perent level of signifiane, in favour of the alternative hypothesis, that the banking setor in the ountry is haraterized by monopolisti ompetitive market. The Wald test performed on the signifiane of H-statisti showed that, it is statistially signifiant and different from zero and one with F-statisti of and at one perent (1%) respetively (see Table 3). The results as presented in Table 3 show that the prie of loanable fund (w 3 ) is positively related to the gross earnings and ontributed the highest portion of the H-statisti for the entire period ( ). This was Table 3. Pooled least square estimate with fixed effets ( ). Variables Geta Coeffiient t-statisti C * Log (w 1 ) * Log (w 2 ) Log (w 3 ) * Log (TA) Log (r 1 ) Log (r 2 ) Log (r 3 ) Log (TDTA) ** ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5 perent level. Summary statistis: 2 = ; F-statistis = (0.000); DW = ; Shwarz riterion = ; H-statistis = ; F-statisti on Wald test for H = 0: , p value = ; F-statisti on Wald test for H = 1: , p value = ; Ho: H = 0 (Monopoly); Ho: H = 1 (Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). Table 4. Pooled least square estimate with fixed effets ( ). Variables Geta Coeffiient t-statisti C * Log (w 1 ) *** Log (w 2 ) Log (w 3 ) * Log (TA) Log (r 1 ) Log (r 2 ) (r 3 ) Log (TDTA) *** ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5perent level. ( *** ) Statistially signifiant at 10 perent level. Summary statistis: 2 = ; F-statistis = (0.0100); DW = ; Shwarz riterion = ; H-statistis = ; F-statisti on Wald test for H = 0: , p value = ; F-statisti on Wald test for H = 1: , p value = ; Ho: H = 0 (Monopoly); Ho: H = 1 (Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). Table 5. Pooled least square estimate with fixed effet ( ). Variables GETA Coeffiient t-statisti C * Log (w 1 ) Log (w 2 ) Log (w 3 ) * Log (TA) Log (r 1 ) Log (r 2 ) (r 3 ) Log (TDTA) * ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5perent level. ( *** ) Statistially signifiant at 10 perent level. Summary statistis: 2 = 0.90; F-statistis = (0.0000); DW = ; Shwarz riterion = 4.516; H-statistis = ; F-statisti on Wald test for H = 0: , p value = ; F-statisti on Wald test for H = 1: , p value = ; Ho: H = 0 (Monopoly); Ho: H = 1(Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). followed by the prie of labour (w 1 ) whih is positively signifiant with a value of This equally shows that labour is an important fator that influene the revenue earned by eah bank. The oeffiient of the prie of apital (w 2 ) was positive but not signifiant and onsistently lower than the oeffiient of the prie of labour and prie of fund. The non-signifiane of the prie of apital ould be as a result of high operating expenses and heavy fixed ost inurred during the sample period. In respet of the other explanatory variables in the model, the ratio of equity to total asset (r 1 ) is positively related to the gross earnings (Geta). This implies that the banking industry in Nigeria supports risk taking in their port-

8 508. A. AJISAFE, A. E. AKINLO folios in order to inrease their earning apaity. It an be dedued that a higher level of risk apital leads to an inrease in the gross earnings of the banking industry. The ratio of loans to total asset (r 2 ) is positively related to the gross earnings (Geta). The positive sign on the r 2 oeffiient is expeted beause aumulation of more loans reflets more potential inome through interest revenue. The ratio of non-performing loan to total loans (r 3 ) revealed a negative sign and is not signifiant. This shows that as the non-performing loan inreases in value, the amount of inome generated by the bank dereased. Total asset (TA) whih is expeted to have a positive relationship does not onform to a priori expetation. The result showed that total asset is negatively related to gross earnings. Total deposit as a ratio of total assets (TDTA) reported the right sign in its oeffiient and statistially signifiant. This shows the importane of deposit in the balane sheet. As more deposits are reeived in the industry, the asset base of the industry expands, whih eventually inrease the inome generated in the industry. For robustness hek, the time dimension of the sample period was divided into two, and The reason for the subdivision is to hek whether there ould be an improvement on the result, due to the reform introdued in the setor in Tables 4 and 5 depit the result of the subsample periods for and The two periods represents the pre and post reform era in the Nigerian banking industry. The period marked the period of bank liquidation while the oinided with the period when Nigerian bank reapitalized whih marked a serious reformation in the setor. In the pre-onsolidation era, the result obtained for the period is not enouraging. The result showed that the prie of loanable fund is statistially signifiant and possessed the appropriate sign with t-statisti of Also, it ontributed the highest oeffiient of the H-statisti with The value of the H-statisti as obtained from the table is whih is statistially different from zero (0) and one (1). It shows that the banking industry in Nigeria for that period exhibits monopolisti ompetition whih supports the result obtained from other studies (see [8,30,31]). To orroborate the signifiane of the H-statisti, the F-statisti obtained from the Wald test is highly signifiant at one perent with a value of and for H = 0 and H = 1 respetively (see Table 4). Total deposit as a ratio of total asset (TDTA) is statistially signifiant at 10% level of signifiane with a value of and onforms to a priori expetation. Other fator pries (pries of apital and prie of labour) also onform to a priori expetation but their ontributions to gross earnings are very low, though not signifiantly different from the entire sample period. The prie of labour is signifiant at 10 perent and inreases the revenue earnings of the bank by about 0.04 perent. The low ontribution of the fator inputs is due to the fat that most of the banks as at this period are weak in terms of apital inadequay, violation of banking law, rules and regulations, and lak of proper management among others. The result of the seond sub-sample period ( ) is presented in Table 5. The result shows that prie of fund is statistially signifiant and onform to a priori expetation. The variable is signifiant beause of the inrease in the level of deposit in the banking industry as a result of the introdution of the reforms in the setor. The reform in the setor possibly led to the effetive mobilization of saving for investment purposes, whih ultimately boosted inreased in the deposits of the banking setor. The prie of apital and prie of labour onform to a priori, but not signifiant. The value of the H-statisti is whih is slightly lower than the pre onsolidation period ( ). It was observed from the result that all the variables follow the same trend as in the first sub sample period and the entire sample period ( ). The ratio of deposit to total asset is highly signifiant in all the period whih shows the importane of deposit in the banking industry Generalized Least Squares Estimation One of the assumptions underlying Ordinary Least Squares Estimation (OLSE) is that the mean and variane are onstant and that errors are unorrelated with one another [23,27]. But when using ross setion data, this assumption may not be true sine it is possible for the varianes of the observations to differ from eah other. When this happens, we say that the random variable and the random error are heteroskedasti. The ross setion heteroskedastiity allows for a different residual variane for different ross setion (Eviews 7, pg. 304). To allow for a different variane for eah bank, we estimate equation 13 using ross setion weights. The result of the estimation is as shown in Table 6. The results in Table 6 showed that all the independent variables are statistially signifiant and positively related to gross earnings (total revenue) of the banks with the exeption of the sale variable (TA) whih is negatively signifiant. The 2 is very high with a value of , meaning that the variation of the explanatory variable with respet to the dependent variable is about 73%. The F-statisti is also highly signifiant at 1% level. The Durbin Watson statisti is also lose to 2, whih show that there is no serial orrelation among the variables. The result showed that bank speifi harateristis or weights is very important in determining the relationship between the fator pries (prie of labour, prie of apital and prie of fund) and the total revenue of the banks. Furthermore, the H-statisti value ranges between zero and one whih supports the

9 . A. AJISAFE, A. E. AKINLO 509 Table 6. Pooled generalised least square estimate with ross setion weight ( ). Variables Coeffiients GETA t-statisti C * Log (w 1 ) * Log (w 2 ) * Log (w 3 ) * Log (TA) * Log (r 1 ) * Log (r 2 ) *** (r 3 ) Log (TDTA) * ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5perent level. ( *** ) Statistially signifiant at 10 perent level. Summary weighted statistis: 2 = ; F-statistis = (0.0000); DW = ; H-statistis = ; F-statisti on Wald test for H = 1: , p value = ; Ho: H = 0 (Monopoly); Ho: H = 1(Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). previous results of monopolisti ompetition with a value of The F-statisti obtained from the Wald test showed that the null hypothesis of H = 0 and H = 1 are rejeted at one perent level of signifiane with a value of and respetively Generalized Method of Moment Estimates (GMM) As a robustness hek, we estimate Equation (13) using the Generalised Method of Moment (GMM) estimation tehnique. A fundamental assumption of regression analysis is that the right hand side of a model is unorrelated with the disturbane term. If this assumption is violated, both ordinary least square and weighted least square are biased and inonsistent. Where the right hand side is orrelated with the residual, instrumental variable regression an be estimated. One of the approahes adopted is the use of Generalized Method of Moments (GMM). GMM generally aount for heteroskedastiity and serial orrelation between exogenous variables and the disturbane term. The GMM speifiation used in this study is based on orthogonally ondition between a funtion and instruments. The study employed the Arellano-Bond estimation with lagged endogenous variable and ross setion fixed effet. In the study, period speifi predetermined instruments were used to indiate that the number of instruments expand dynamially over time. Aside, the tehnique allows us to relax the assumption of strit exogeneity of pooled least square estimation tehnique. The results of the Generalised Method of Moment (GMM) estimate is reported in Table 7 for the full sample period ( ) whih onfirmed the result of the generalized least square estimate in Table 6. The fator pries: prie of labour (w 1 ), prie of apital (w 2 ) and prie Table 7. Panel generalised method of moments with fixed effets. VAIABLES GETA Coeffiients t-statisti GETA ( 1) Log (w 1 ) ** Log (w 2 ) Log (w 3 ) * Log (TA) * Log (r 1 ) Log (r 2 ) (r 3 ) Log (TDTA) * ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5perent level. ( *** ) Statistially signifiant at 10 perent level. J-statisti = ; Instrument rank = 148; H-statistis = ; F-statisti on Wald test for H = 0: , p value = ; F-statisti on Wald test for H = 1: , p value = ; Ho: H = 0 (Monopoly); Ho: H = 1 (Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). of fund (w 3 ) are all signifiant and onform to a priori expetation. The prie of fund is positive and signifiant with a t-value of and oeffiient value of The prie of apital is also signifiant at 10% with a t-value of and oeffiient value of while the prie of labour is also positive and signifiant with a t-value of and oeffiient value of The lagged endogenous variable (Geta) is positively related not signifiant. It shows that the revenue earned by the bank in the previous period is positively related to the revenue earned in the urrent period, but with little impat on the total revenue. All other variables follow the trends observed in Table 3. Moreover, from the result in Table 7, the value of our H-statisti is , whih is the sum of elastiities of banks equilibrium revenue with respet to the fator input pries, that is, the sum of the oeffiients of prie of labour, prie of apital and prie of funds as shown in Equation (14). This value onfirmed the result of the other H-statisti obtained in this study. This means that ommerial banks in Nigeria is haraterised by monopolisti ompetition but the level of ompetition is too low as observed by [30] whih reports a low H-statisti for Austria (0.154) and Denmark (0.050). The prie of fund (w 3 ) ontributed the highest value of the H-statistis whih support the Pooled OLS and generalized least square estimate. From the result of our H-statisti, it ould be observed that the result is onsistent with previous studies whih support monopolisti ompetition. Furthermore, our results also support the fat that, prie of apital ontributes minimally to total revenue when ompared to other input pries. This is the least important omponent of H- statistis see [8,15]. The values of the J-statisti and instrument rank showed that the instrument used is valid sine the value of the instrument rank (148) is greater

10 510. A. AJISAFE, A. E. AKINLO than the number of estimated oeffiient (9). The J-statisti alulated is When ompared with the values of the following information riteria (BIC = 0.982, HQIC = 0.966, NIC = 0.986) in order to determine the validity of the instrument, it was then onluded that the instrument used is valid. The model seletion riteria perform reasonably well for sample sizes above Market Equilibrium Test for H-Statisti An important feature of the H-statisti is that equilibrium tests must be performed on observations that are in longrun. The equilibrium test is suggested based on the fat that, in equilibrium, rates of return aross banks should not be orrelated statistially with input pries [30]. To test whether this assumption holds in the ase of Nigeria, we remodeled the P methodology by using the return on asset as proposed in the literature [3,9,30]. It should be noted that, the measure of OA an take on negative values on some oasion due to banks losses in any year, thus, the variable is adjusted simply for the small negative values and omputed as (1 + OA) for onveniene. Equation (15) is then estimated for the market equilibrium test. The results obtained from the pooled data revealed that the market is in equilibrium based on the value of H-statisti of with F-statisti of and p value of obtained from the Wald test. This is statistially signifiant and different from zero and one at 1% level (see Table 8). The results from the equilibrium test also show that all the fator input pries are positively related to return on asset and statistially signifiant with the exeption of the prie of apital. The prie of labour (w 1 ) ontributes mostly to the value of H in ontrast to our previous results in whih the Table 8. Pooled generalised least square estimate with ross setion weights (Equilibrium Test). Variables Coeffiients 1 + OA t-statisti C * Log (w 1 ) * Log (w 2 ) Log (w 3 ) ** Log (TA) * Log (r 1 ) * Log (r 2 ) (r 3 ) Log (TDTA) ** ( * ) Statistially signifiant at 1perent level. ( ** ) Statistially signifiant at 5perent level. ( *** ) Statistially signifiant at 10 perent level. Summary weighted statistis: 2 = ; F-statistis = (0.0000); DW = ; H-statistis = ; F-statisti on Wald test for H = 0: , p value = ; F-statisti on Wald test for H = 1:15458, p value = ; Ho: H = 0 (Monopoly); Ho: H = 1(Perfet Competition); Ho: 0 < H < 1 (Monopolisti Competition). prie of fund ontributes the highest oeffiient to H-statisti. This implies that as a result of an inrease in the prie of labour, they are enouraged to improve on their apaity, whih invariably inreases their ontribution to the total revenue of the industry (banks net inome). Based on our alulation of H-statisti, the result obtained from our estimation of equilibrium test shows that our observations are in equilibrium. This is premised on the value of H-statisti reported earlier (H = ). This value is loser to zero than one (1). This implies the rejetion of the null hypothesis of H = 0 and H = 1 at 1 perent level of signifiane. Other variables in the model are statistially signifiant with the exeption of non-performing loans (r 3 ). This is beause some of the loans granted by the banks to individuals or firms were diverted to other uses rather than what they were meant for whih made the repayment diffiult. 5. Conlusion Competition in the banking setor has reeived greater attention of the eonomists in reent years but the degree of ompetition has generated a serious debate both in the developed and developing eonomies. It has been observed in the literature that a higher degree of ompetition in banking industry is expeted to provide welfare gains through redution in the pries of fator input and thereby aelerating growth. From the empirial analysis of the study, it was observed that ommerial banks in Nigeria showed evidene of monopolisti ompetition whih orroborate the result obtained in the previous studies. 2 The result obtained is in ontrast to the theory whih emphasizes oligopolisti nature of the industry beause of the dominane of some banks in the industry. EFEENCES [1] M. Brownbridge, The Impat of Publi Poliy on the Banking System in Nigeria, [2] L. Weill, On the elationship between Competition and Effiieny in the EU Banking Setors, Kredit und Kapital, Vol. 37, No. 3, 2003, pp [3] S. Claessens and L. Laeven, Finanial Dependene, Banking Setor Competition, and Eonomi Growth, World Bank Poliy esearh Working Paper, No. 3481, [4] A. Prasad and S. Ghosh, Competition in Indian Banking, IMF Working Paper, [5] L. Weill, A. Pruteanu-Podpiera and F. Shobert, Banking Competition and Effiieny: A Miro-Data Analysis on the Czeh Banking Industry, Comparative Eonomi Studies, Vol. 50, No. 2, pp doi: /palgrave.es Shaffer 1982, Molyneux etal 1994, Prasad and Gosh 2005, Buhs and Mathisen 2005 among others.

11 . A. AJISAFE, A. E. AKINLO 511 [6] J. A. Bikker and K. Haaf, Competition, Conentration and Their elationship: An Empirial Analysis of the Banking Industry, Journal of Banking and Finane, Vol. 26, No. 11, 2002, pp doi: /s (02) [7] S. Claessens and L. Laeven, What Drives Bank Competition? Some International Evidene, Journal of Money Credit and Banking, Vol. 36, No. 3, 2004, pp doi: /mb [8] D. Denvil, Testing for Competition in the Jamaian Banking Setor: Evidene from Bank Level Data, Finanial Stability Unit, Bank of Jamaia, [9] T. Buhs and J. Mathisen, Competition and Effiieny in Banking: Behavioural and Evidene from Ghana, IMF Working Paper, International Monetary Fund, [10] A. M. Simpasa, Competitive Conditions in the Tanzania Commerial Banking Industry, Afrian Development eview, Vol. 23, No. 1, 2011, pp doi: /j x [11] Y. Sun, eent Developments in European Bank Competition, International Monetary Fund Working Paper, [12] J. A. Bikker and L. Spierdijk, How Banking Competition Changed Overtime, Tjalling C. Koopmans esearh Institute, Disussion Paper Series, No , [13] G. Alessandro and. Kevin, Bank Competition and Effiieny in the FY Maedonia, South-Eastern Europe Journal of Eonomis, Vol. 2, 2008, pp [14] S. Shaffer, A Non-Strutural Test for Competition in Finanial Markets, In: Proeedings of a Conferene in Bank Struture and Competition, Federal eserve Bank of Chiago, Chiago, [15] P. Molyneux, Y. Altunbas and E. Gardener, Effiieny in European Banking, Wiley, Chihester, [16] O. De Bandt and Philip E. Davis, Competition, Contestability and Market Struture in European Banking Setors on the Eve of EMU, Journal of Banking and Finane, Vol. 24, No. 6, 2000, pp doi: /s (99)00117-x [17] J. C. Panzar and J. N. osse, Testing for Monopoly Equilibrium, Journal of Industrial Eonomis, Vol. 35, No. 4, 1987, pp doi: / [18] A. Kuotsoyiannis, Modern Miroeonomis, International Edition, Mamillian Press Ltd., London, [19] C. A. Northott, Competition in Banking: A eview of the Literature, Bank of Canada Working Paper, , [20] J. M. Wooldridge, Introdutory Eonometris: A Modern Approah, South-Western College Publishing, Cininnati, [21]. S. Pindyk and D. L. ubinfeld, Eonometri Models and Eonomi Foreasts, MGraw-Hill, Singapore City, [22] K. S. Pesaran and Y. Shin, Testing for Unit oots in Heterogeneous Panels, Journal of Eonometris, Vol. 115, No. 1, 2003, pp doi: /s (03) [23] W. H. Greene, Eonometri Analysis, 6th Edition, Prentie-Hall, Upper Saddle iver, [24]. C. Hill, W. E. Griffiths and G. C. Lim, Priniples of Eonometris, 3rd Edition, John Wiley and Sons, In., New York, [25] A. Levin, C. F. Lin and C. Chu, Unit oot Tests in Panel Data: Asymptoti and Finite Sample Properties, Journal of Eonometris, Vol. 108, No. 1, 2002, pp doi: /s (01) [26] B. Baltagi, Eds., oot Test for Panel Data, In: Advanes in Eonometris, Vol. 15: Nonstationary Panels, Panel Cointegration, and Dynami Panels, JAI Press, Amsterdam, pp [27] G. S. Maddala and S. Wu, A Comparative Study of Unit oot Tests with Panel Data and a New Simple Test, Oxford Bulletin of Eonomis and Statistis, Vol. 61, No. S1, 1999, pp doi: / s1.13 [28] I. Choi, Unit oot Tests for Panel Data, Journal of International Money and Finane, Vol. 20, No. 2, 2001, pp doi: /s (00) [29] K. Hadri, Testing for Stationarity in Heterogeneous Panel Data, Eonometri Journal, Vol. 3, No. 2, 2000, pp doi: / x [30] B. Casu and C. Girardone, Does Competition Leads to Effiieny? The Case of EU Commerial Banks, Disussion Paper, Essex Finane Centre, [31] L. G. de ozas, Testing for Competition in the Spanish Banking Industry: The Panzar-osse Approah evisited, Doumentos de Trabajo, No. 0726, 2007.