Ersa Working Paper Proposal - Mangwe and Ziv

8/6/2019 Ersa Working Paper Proposal - Mangwe and Ziv

1/28

1

Bank concentration and the interest rate pass-through in selected

Sub-Saharan African countries

T. Mangwengwende, Z. Chinzara and H. Nel

ABSTRACT

This study investigates the link between bank concentration and interest rate pass-through(IRPT) in four sub-Saharan countries. It also analyses whether there is asymmetry in IRPT and

whether such asymmetry is related to changes in bank concentration. By applying a number ofeconometric methods including Asymmetric Error Correction Models, Mean Adjustment Lagmodels and Autoregressive Distributed Lag models on monthly data for the period 1994-2007,the study found some evidence of a relationship between bank concentration and IRPT in all fourcountries. However, the results reveal that bank concentration has a stronger influence on themagnitude of its adjustment rather than its speed. Of particular note in this investigation is thefact that the findings support both the structure conduct performance hypothesis and thecompeting efficient structure hypothesis in the banking industries of the four countries. Whilethere is some evidence supporting the view that bank lending and deposit rates adjustasymmetrically to changes in policy rates, there is very limited evidence that these asymmetriesare a result of bank concentration. The key implication of the result for African countries is that

increased bank concentration through bank consolidation programmes designed to strengthenbanking industries should not be viewed with cynicism in so far as monetary policy transmissionis concerned because concentration does not necessarily result in the detriment of the effectiveconduct of monetary policy.

Keywords: Bank Concentration, Monetary Policy, Interest Rate Pass-Through,Asymmetric Adjustment, Sub-Saharan AfricaJEL Classification: E52, E58, G28

The Financial support from ERSA is acknowledged. Views and opinions expressed are of the authors and do notnecessarily represent those of ERSA.All authors from the Department of Economics, Rhodes University, P.O. Box 94 Grahamstown 6140. Emails:[email protected]; [email protected]; [email protected], respectively. Correspondence Author


2/28

2

1. INTRODUCTION

While there has generally been consensus on the importance of effective monetary

policy, debate on the most appropriate means to achieve it has created long-standing issues in theliterature of monetary economics and central banking (Rasche and Williams, 2005). Accordingto Mishkin (1995), monetary policy is only effective if its tools are able to quickly transmitmonetary impulses to interest rates and if the resultant new structure of interest rates affects realexpenditure. If market rates are sluggish in their adjustment to policy rate changes then thedesired goals of the monetary policy change may not be achieved despite the size of the changein the policy rates.

Given that the strength1 of the interest rate pass-through (IRPT) is necessarily dependenton the setting of lending and deposit rates by banks, research has focused on identifying elements

of the banking industry that could provide incentives or disincentives for banks to adjust theirmarket rates in line with policy rate changes. Amongst the most contentious of these is bankconcentration which is particularly interesting because different industrial organisation theories,and the interpretation of those theories, provide ambiguous implications for the relationshipbetween changing bank concentration and the conduct of banks and, ultimately, the IRPT. Bains(1956 in Hannan and Berger, 1991) structure-conduct-performance hypothesis suggests a clear positive relationship between increasing market concentration and a weak IRPT based on theexpected behaviour of market participants in highly concentrated markets. In contrast, theefficient structure hypothesis suggests that the IRPT would strengthen as the market becamemore concentrated as all inefficient banks that failed to timely respond to policy rate changes

were competed out of the market (Allen and Gale, 2003).Empirically, in the work of Hannan and Berger (1991), Neumark and Sharpe (1992) and

Scholnick (1996), banking sector concentration influences bank conduct and consequently thenature of the IRPT. Conversely, in the studies conducted by Cottarelli and Kourelis (1994),Berstein and Fuentes (2005) and Abbasoglu et al. (2005), banking sector concentration does notinfluence the IRPT. As evidenced by this, not only are the theories not in agreement but thedisparities are confounded by conflicts in the evidence from empirical research.

Understanding whether or not bank industry concentration is related to a weak or strongIRPT is important because it will inform banking sector competition regulation. If concentrationimpedes the IRPT, and consequently the effectiveness of monetary policy, then it is ofparamount importance that the banking industry should be regulated to increase the number ofmarket participants and reduce concentration. If concentration improves the IRPT then bankingregulators may need to consider the possibility of consolidating their banking industries toencourage more concentration to enhance the effectiveness of monetary policy2. In addition to

1 A perfect IRPT refers to a pass-through process in which official rates are fully and quickly reflected in banklending and deposit rates. This makes IRPT analysis synonymous with the measurement of its speed and magnitude.2 It is important to note that the results on whether high or low bank concentration is related to a weakening orstrengthening IRPT can only partially inform banking sector competition regulation as other implications of a


3/28

3

informing the regulation of the banking industry, identifying a relationship has implications forthe setting of monetary policy when bank concentration is changing. To illustrate, should bank

concentration be related to a slower IRPT then the timing of official rate changes must change toensure that they are set early enough to have the desired effect at the desired time. If, on the otherhand, bank concentration is related to a weakening of the magnitude of the IRPT, then policychanges must be large enough to cater for the diminished reflection of official rates in banklending and deposit rates. These issues have been made poignant given trends in bankconsolidation in Africa and a lack of empirical research on the matter on the continent.

Thus the current study attempts to fill this void by investigating this potentially importantlink between bank concentration and the strength of the IRPT in Botswana, Nigeria, South Africaand Zambia. Furthermore, the study analyses whether there is asymmetry in pass-through andwhether this asymmetry is a consequence of bank concentration.

The rest of this paper is organised as follows: Section 2 provides a literature review;Section 3 presents an overview of monetary policy in the four countries; Section 4 discusses theanalytical framework and data used in the paper; Section 5 presents the results; and Section 6gives the conclusions and their policy implications.

2. THEORETICAL AND EMPIRICAL LITERATURE

While bank concentration and the IRPT can be described and identified separately, thereis no explicitly stated theoretical representation of the interaction between them (Hoffman andMizen, 2001)3. Consequently, to study their relationship it is necessary to merge anunderstanding of market concentration and the pricing decisions of banks as explained by

industrial organisation theory4. Given this, theoretical questions arise regarding why and how theconcentration of the banking industry would influence the effectiveness of the IRPT.

With regard to the important theoretical debate, a leading authority on bank modeloperation, the Monti-Klein model of banking microstructure shows that there is a cost functionfor running a bank that depends on the aggregate value of the assets (A) being managed by the bank and its factors of production such as its capital (K) and labour (L). Furthermore, underconditions of oligopoly, the rational profit maximising spread in a banks retail rates ( ) isa function of those factors of production, the aggregate value of bank assets, the semi elasticitiesof demand and supply (L, D) and, very importantly, the number of banks in the industry (N).

This is represented in the equation below.

highly/lowly concentrated banking sector must also be considered. For example, questions concerning the impact ofconcentration on financial stability (see Allen and Gale, 2003).3 There are theories (explored here) regarding why it may be made smoother/stronger or more rigid/weak, however,these are not actually IRPT theories but theories surrounding IRPT changes, for example, theories on bank

behaviour.4 In essence central banks, as monetary policy authorities, enact policy measures by influencing the setting of banklending and deposit rates. However, where factors exist to reduce or enhance these rate, such factors are in factchanging the speed and magnitude of the IRPT. Industrial organisation theory offers reasons why bank price settingmay deviate or closely follow policy rate changes.


4/28

4

(1)

where is the retail interest rate set by banks on lending and is the retail interest rate set bybanks on deposits.Equation (1)5 shows how profitability in the banking industry is influenced by the

number of participants in the industry. More importantly (and relevant to this study) is that it presents the setting of bank lending and deposit rates as a function of the banking industrysmarket structure. As the number of participants and the asset holdings of a bank change, itadjusts its lending and deposit rates to maintain its interest rate spread. Put simply, the modelrefutes the idea that rational profit maximising banks make decisions independent of the marketstructure in which they are in. Ultimately, this means that the IRPT should be related to thestructure of the banking industry. This view of bank costs and their operation has been used in

the work of Van Leuvensteijn et al. (2008) and Corvoisier and Gropp (2002).In addition two competing theories exist to explain how bank concentration affects the

setting of bank lending and deposit rates and by extension the IRPT. The first is the Structure-Conduct-Performance (SCP) hypothesis that was pioneered by Bain (1956 in Hannan andBerger, 1991). This hypothesis suggests that the structure of an industry will influence theconduct of its members, which will in turn influence the performance of that industry. Forexample, an uncompetitive banking industry (structure) will most likely result in inefficient practices and a reduced incentive to respond quickly to monetary policy stance changes(conduct) results in a sluggish and incomplete IRPT.

The second theory is the Efficient Structure Hypothesis (ESH) which suggests thatconcentration would increase the overall efficiency of the banking industry resulting in banks pricing their products more competitively. This is based on the precept that increasedconcentration is due to more efficient banks growing more rapidly than less efficient banks, ormore efficient banks taking over less efficient ones. If this were the case, at least up to somepoint, banks would price their services more competitively6, rather than less competitively.

Theoretical literature further suggests that the response in the lending and deposit ratesmay be asymmetric7. Two competing hypotheses have been put forward to explain this possibleasymmetry in IRPT. On the one hand, the collusive behaviour hypothesis posits that banks inhighly concentrated markets may respond more positively to impulses to adjust their lending

rates8

upwards rather than downwards because of their collective market power, thus resulting inasymmetric lending rate adjustments. In such a case simply assessing the adjustment under the

5 This is synthesised from the more elaborate presentation of the Monti-Klein model in Van Leuvensteijn et al.(2008).6 Here the word competitively is taken to refer to pricing strategies that closely reflect the changes in the officialrate. This is the case because the competition would remove the scope for large disparities between the two or forslow adjustments in the bank rates in response to official rate changes.7 Most empirical research has focused on the symmetric response of bank rates noting simply whether bank rateadjustments strengthen or weaken without explicitly considering the potential differences between increases anddecreases.8 Note that lending rates represent returns on assets, namely, loans.


5/28

5

assumption that the rate would adjust similarly up and down would miss the important fact thatbank concentration may be associated with a weakening of the impulse to reduce lending rates

and not the impulse to increase them. For this reason ignoring the possibility of asymmetry inadjustment could lead to erroneous conclusions about the presence and nature of a relationshipbetween bank concentration and bank lending and deposit rates. On the other hand, the adversecustomer reaction hypothesis suggests that in an environment where borrowers have bargainingpower, rigidities may occur in downward adjustment of deposit rates following an expansionarypolicy stance and in upward adjustments of lending rates following a contractionary monetarypolicy stance.

The empirical research on this topic has been primarily concerned with investigatingreasons for retail rate rigidities in the light of official rate changes and so presents a myriad offactors that influence the IRPT, of which bank concentration is only one factor9. The results of

these studies have presented ambiguous conclusions on both the existence and nature of therelationship between bank concentration and the IRPT which has served to compound theambiguity created by the competing theories.

Cottarelli and Kourelis (1994) measured the speed of the adjustment of bank lendingrates in 31 industrial and developing countries by regressing the lending rate on a distributed lagof money market rates. In this way, they estimated the effect on lending rates of shocks in moneymarket rates following these shocks after three months, after six months, and in the long run.Secondly, they explained the cross-country differences in these multipliers by regressing them onseveral variables related to the structure of the financial system, such as the degree ofconcentration in the banking industry, the existence of constraints on capital flows and barriers toentry, and the size and the efficiency of the money market. Using monthly data on lending,discount and money market rates, the impact multiplier (defined as the change in the lending rateobserved during the month in which the money market rate changes) was close to unity in somecountries but as low as zero in others. Significant differences could also still be observed afterthree and six months. In the long run, however, the adjustment was close to unity for mostcountries. However, banking sector concentration did not have a significant effect on the pass-through.

Cottarelli et al. (1995) studied 63 banks in Italy from June 1986 to December 1993.Using an error correction model on monthly data over the period they established that bank

concentration did have a significant impact on the IRPT. The authors argued the five firmconcentration ratio (CR5) used in Cottarelli and Kourelis (1994) was a weak measure of marketconcentration and so they used the Herfindahl Hirschman Index (HHI). Their findings supportedthe existence of a negative relationship between bank concentration and the speed of the pass-through and were in line with Hannan and Bergers (1991) findings for US banks. Consistentwith the structure-conduct performance hypothesis, banks in the most concentrated local markets

9 This is to say that while the question of whether a relationship exists or not is indicated in a considerable amount ofliterature fewer studies have focused on explicitly investigating the relationship between the IRPT and bankconcentration.


6/28

6

were found to pay money market deposit account rates that ranged from 25 to 100 basis pointsless than those paid in the least concentrated markets.

In contrast to this finding, Berstein and Fuentes (2005) found results that were similar toCottarelli and Kourelis (1994). They found that the speed of adjustment was affected by bankingsector expectations and interest rate volatility, but not concentration. Taking monthly data ondeposit interest rates of different maturities in Chile from May 1995 to December 2002 andconcentration ratios of the three largest (CR3) banks, the five largest banks (CR5) and the HHI,their study showed that bank interest rates responded by between 75% and 88% to changes in theinterbank interest rate. However, the authors conceded that their results were not in line withtheir expectations and so sought an explanation. In their research they found that concentrationaffected the coefficient of the lagged variables in their model and thus it had a long-run influencethat could be missed in a short-run analysis. From this they concluded that market concentration

negatively affected the interest rate pass-through, the only issue being one of when this effectwould be felt in the pass-through. Importantly here, was the observation that the influence ofmarket concentration could be missed if the analysis of the relationship was confined to the shortrun. In addition, the study also identified the potential for an asymmetric relationship in whichbank concentration only influences one type of change in the bank lending and deposit rates10.

Abbasoglu et al. (2005) presented another study that refuted the link between interest ratepass-through rigidity and the level of concentration in the banking sector. Taking CR3, CR5 andthe HHI for Turkish banks from 2001 to 2005 they saw no evidence suggesting that there was arelationship between bank concentration and the IRPT. Of importance in their work was theidentification that the IRPT was influenced by the level of banking sector competition. Thisobservation was important because Van Leuvensteijn et al.s (2008) work would suggest thatcontention existed in the study of bank concentration and the IRPT because researchers had beenstudying the effect of competition on the IRPT and not that of bank concentration. As a result,the expected relationship between bank competition and the IRPT was sometimes mistakenlyrepresented as the relationship between bank concentration and the IRPT11. In Pakistan, Centraland Eastern Europe Quayyum et al. (2005) and Egert et al. (2007) found evidence supporting arelationship between bank concentration and the IRPT. While Quayyum et al. (2005) used Boxet al.s Intervention Approach and Egert et al. (2007) used a bivariate error correction system,their results both revealed that developing countries had discernable relationships between the

concentration of their banking sector and their IRPT.

10 To illustrate, we follow Tomasz (2003) who observes that bank concentration need not exert a symmetric effecton the pass-through: it may speed up the pass-through in times of declining interest rates while slowing it down in

periods of increasing interest rates. In such a case, failure to consider the asymmetric response would yieldconfusing results.11 Such a mistake can be understood when concentration can be seen sometimes to be closely related to competition,for example, where the performance of an industry (competitiveness) is greatly influenced by the structure of thefirms in the industry (concentration). Where concentration is high, for example, competitive pressures may beabated by tacit collusion on the parts of the participating firm. In such a situation concentration and competition areclosely and negatively related.


7/28

7

From the foregoing, it is evident that the theoretical literature is ambiguous on therelationship between bank concentration and IRPT. While the Monti-Klein model illustrates that

a profit-maximising bank takes cognisance of the number of players in the market whenadjusting its profit margin, whether this profit-maximising behaviour is detrimental to effectivemonetary policy implementation when the banking environment is concentrated is unclear. Thenet effect on monetary policy is likely to depend on whether concentration promotes collusiveand uncompetitive behaviour by banks as the SCP suggests or whether it enhances efficiency byflushing out uncompetitive banks as the ESH suggests. Furthermore, it has been seen thatconcentration is likely to result in asymmetry in the response of retail rates to changes in officialrates. The empirical literature still suffers from the same ambiguity as the theory. Moreover,relevant IRPT studies in African countries tend to shy away from the issue of bankconcentration. In this regard, the current study is particularly important as it is the first attempt to

address this issue for African countries.

3. OVERVIEW OF MONETARY POLICY AND BANKING INDUSTRY IN THE

COUNTRIES

3.1 Monetary Policy

The principal objectives of monetary policy in all four countries were the maintenance ofstable price levels, the preservation of an environment conducive for economic growth and themaintenance of the balance of payments. However, despite similar objectives the targets ofmonetary policy differed with South Africa and Botswana using inflation targets while Zambia

and Nigeria used monetary aggregates (South African Reserve Bank, 2009; Bank of Botswana,2009, Central Bank of Nigeria, 2009; Bank of Zambia, 2009).

In each of the four countries interest rates are a critical tool for the conduction ofmonetary policy. Consequently, the behaviour of private banks in setting lending and depositrates could potentially have a profound effect on the effectiveness of monetary policy. In Figure1 plots showing how well the movements in policy rates are mirrored by movement in lendingand deposit rates are given for each country.

Figure 1: Trends in official and retail rates


8/28

8

There is a clear positive trend for the policy, lending and deposit rates in South Africasuggesting a close relationship between these interest rates. Zambia also has similar trends in the

movement of its interest rates. For Botswana, the trend is clearest between the policy rate and thelending rate, while the trend with deposit rates is weak. Of the four countries, Nigeria presentsthe weakest trend amongst the interest rates. While the general pattern of rates shows a positiverelationship, like in the other countries, the movement is not as closely related.

To complement this preliminary graphical analysis, correlation coefficients for eachcountrys deposit and lending rate with the central bank rate were computed and the results arereported in Table 1.

Table 1: Correlation between Central Bank Rate with Deposit Rate and Lending Rate

Botswana(BOTS) Nigeria (NIG)

South Africa(SA) Zambia (ZAM)

Central BankRate

Central BankRate

Central BankRate Central Bank Rate

Central Bank Rate 1.0000 1.0000 1.0000 1.0000

Deposit Rate 0.1930 0.7847 0.9780 0.9769

Lending Rate 0.8717 0.8623 0.9931 0.9928

In line with the graphical plots, SA has the highest correlation coefficients, followed byZambia whose correlation coefficients are marginally below those of SA. Botswana and Nigeriahave nearly similar correlation coefficients for both the official rate and the lending rate, whileBotswana has by far the lowest correlation coefficients between official and deposit rates.

While it should noted that the graphical plots and the correlation coefficients are notmeant to draw conclusions about the speed and magnitude of the IRPT, they do give a picture onhow well the retail rates seem to respond to the official rate and thus help stimulate questionsabout the effectiveness of monetary policy in the four countries. For instance South Africa andZambia exhibit the closest trends and strongest correlation in the policy, lending and depositrates while Botswana exhibits weakest correlation between official rate and deposit rate. Doesthis imply that pass-through is strongest in South Africa and Zambia and pass-through to depositrate is quite rigid in Botswana? However, more analysis is needed to answer these questions.

3.2 Banking concentration and interest rate margin

The Monte-Klein model suggests that banks adjust their profit margins to changes in thenumber of competitors in the banking industry. To visually examine this claim, time series plotsof bank concentration ratios are compared to time series plots of interest rate margin (i.e. lendingminus deposit rate) in Figure 2.


9/28

Figure 2: Bank Concentration and Net Interest Rate Margin

Adapted from a New Database on Financial Development Structure 2007 (World Bank, 2009)


10/28

10

Generally, with the exception of South Africa, bank concentration declined over thesample period. The decreases seem to be more apparent in Nigeria and Zambia. The major

influence in the declining concentration was the adoption of more liberal financial regulation, theencouragement of banking sector growth to cater to growing markets, and the desire to improvefinancial service delivery and capacity. In Botswana, the threat of market power abusenecessitated the encouragement of increased industry participation (Kayawe and Amusa, 2003:5). In an effort to develop the financial services sector the requirements for starting up a Nigerianbank were relaxed with drastic reductions in statutory reserve requirements. The resultant rapidincrease in the number of banks led to the existence of many banks without sound financialfoundations and so many of them failed. This failure of banks necessitated the introduction of a bank consolidation exercise in 2004 to restore financial stability which was threatened by thefragile banking industry (Soludo, 2004: 2). In Zambia financial reforms in 1991 led to a more

liberalised financial sector which drew more banking sector participants. However, thisliberalisation also resulted in increased bank failures and the stunted growth of existing banks(Musonda, 2008). In South Africa, it is surprising that despite post-democratisation policiesmeant to ensure that a large number of previously unbanked black population got access tofinancial intermediaries (SARB, 2010), concentration has not significantly decreased. This wasmainly due to bank consolidation that began in the late 1980s when profit margins were fallingdue to non-performing loans and the high cost of modern technology (Gidlow, 2003). Gidlow(2003) further argue that despite the inception of the Banks Act of 1995 that encouraged entry offoreign banks to enhance competition, concentration has continued to rise as foreign bankslimited their activities to niche markets due to their failure to compete with the big four SouthAfrican banks in the core areas of retail banking12.

Turning to the relationship between profit margin and the level of concentration, it isevident from Figure 2 that the relationship is not clear-cut, except for Zambia, where thedownward movements in bank concentration seem to be mirrored by the downward movement inthe profit margin. In the case of Nigeria, while both bank concentration and interest rate margindecrease, the year-to-year trends in the two variables do not show any co-movement. ForBotswana, no relationship seems to exist at all between the two variables. While bankconcentration gradually decreased during the period, the interest rate margin remained relativelystable despite the unexplained spike between 1998 and 2000. In South Africa, while the

relationship between the two is quite clear in some periods, for instance 1995-1997 and 2001-2004, the trends in the two variables are largely inconsistent. Most surprisingly, the interest ratemargin sharply increases despite the decrease in bank concentration between 2004 and 2007.Coefficients of correlation between bank concentration and interest rate margin were alsocomputed and are reported in Table 2.

12 For a comprehensive discussion of the South African banking industry see Aziakpono and Wilson (2010).


11/28


12/28

12

countries and sample periods were selected primarily according to data availability. The data islargely sourced from the IMF International Financial Statistics (IFS) CD ROM 2009, Central

Bank Reports and the New Database on Financial Development and Structure 2007 (Becket al.,2007 and World Bank, 2009).

Methodology, Modelling Framework and Econometric Procedure

In order to address the question of whether bank concentration is relevant in IRPT thisstudy use two approaches. The first approach uses the overall average bank concentration for theentire period in each of the four countries. The average bank concentration is then compared withthe IRPTs corresponding to the same time period to evaluate whether countries with the highest(lowest) bank concentration have the weakest (strongest) IRPT. The second approach involvestracing the dynamics of IRPT and bank concentration over time using an 8-year rolling windowtechnique15. The idea here is to examine whether the time-to-time dynamics in bankconcentration are related to time-to-time dynamics in IRPT in each of the countries.

The starting point in analysing IRPT is the specification of a primary model which showshow rational commercial banks make pricing decisions on their loans and deposits cognisant ofthe cost of obtaining accommodation from the central bank (Scholnick, 1996 and Aziakpono etal., 2007). This model is a simple specification the long run relationship between policy rates andbank lending/deposit rates as follows:

(2)

where is the endogenously determined retail rate (in this study deposit and lending interest

rates), is the policy dictated rate, is an intercept which denotes a mark up/mark down onthe retail rates to reflect market conditions e.g. credit risk premium on the lending rate (seeMarotta, 2009 ), is the slope parameter that shows the fraction of the change in official ratesthat is reflected in changes in the retail rates in the long run. A close to 0 would imply that thelong run IRPT is slow and weak. While a of 1 would be desirable as it implies that the policyrate is fully reflected in bank retail rates, it would be unlikely because of a number of factorssuch as asymmetric information, market imperfections, menu cost, switching costs and adversecustomer reactions (Aziakpono and Wilson, 2010: 19). Consequently, is likely to be between0 and 1, with a value close to 1 implying high and near complete pass-through, and vice versa.

In rare circumstances, overpass-through might occur when commercial banks respond to higherrisks posed by asymmetric information by charging very high interest rates charges resulting in>1 (De Bondt, 2005; Aziakpono and Wilson, 2010). Variations in the magnitude of IRPT arereflected in the variation in the size of, while variations in the speed of IRPT adjustment arereflected by the variation in the time taken for to be fully reflected in the bank lending anddeposit rates. Finally is a stochastic error term.

15 The 8-year period was chosen to ensure that each period had approximately 100 observations to prevent smallsample bias.


13/28

13

The simplest way to estimate the long run parameters 0 and 1 would be by using theOrdinary Least Squares (OLS) technique. However it is well documented that time series usually

possess some properties that violate some of the core assumptions of the OLS technique, notablyconstant mean, homoscedasticity, and no serial correlation. Series that violate any one or all ofthese three assumptions are said to be non-stationary and using OLS in estimating long runparameters in regressions containing such series may lead to spurious results whose parametervariances and standard errors cannot be relied upon (Guajarati, 2002). Thus a natural startingpoint in time series analysis is testing whether the series are stationary. In this study we employthe Augmented Dickey Fuller (ADF) approach. This technique improves the original DickeyFuller (DF) approach by adding lags of the dependent variables to automatically correct forautocorrelation in the DF equation. A potential issue becomes how the appropriate lag length thatguarantees that autocorrelation is corrected is selected. The current study uses the Schwartz

information criterion. Estimating the ADF equation requires that a maximum lag length isselected. With the wisdom that monthly data is being used, the maximum lag length is set attwelve months16. While the ADF test performs well when the serial correlation in the error termsare well approximated by a low order AR(p) process without any large negative roots, the test isbiased towards rejection of the null hypothesis in cases where the error terms follow a MA orARMA process (Davidson and MacKinnon, 2004: 622). Thus, the Kwiatkowski, Phillips,Schmidt, and Shin (KPSS) tests, which directly test for stationarity, unlike the unit root testwhich tests for stationarity in an indirect manner, is also employed as a robust test.17

For series that are stationary, OLS will be used to estimate the long-run parameters inequation (2). Otherwise the series will have to be differenced to make them stationary beforeestimating (provided that the series are stationary at first difference), in which case the long-runproperties of the series is lost and the resulting parameter will be interpreted as representing theshort-run relationship between the official rate and the retail rates. Based on equation (2) themodel for estimating the short-run parameter is the following autoregressive distributed lag(ADL) model:

t

m

j

jtj

n

i

ititt RRCBRCBRRR ++++= =

=

1110

(3)

where is a first difference operator, n and m denote the maximum number of lags chosen based

on the Akaike Information Criterion (AIC), t is a white noise error term, and 0 and 1 are short-

run parameters which in this context can be interpreted as the short run/immediate pass-throughand due to various impediments as mentioned earlier it is likely to be lower than the long-runpass-through (1). Following Kwapil and Scharler (2009), the symmetric LR impact of a policyrate change can then be computed from equation (2) as follows:

16 An alternative for choosing the lag length would be using the quarterly monetary policy review meeting. Howeverwe believe that they will enhance their previous stance rather than abruptly take a new policy stance as this willtrigger undesirable money market volatility. Thus a 12-month maximum lag length is quite reasonable.17 Since the two techniques for testing for stationarity are widely explored in several empirical studies, thetheoretical underpinning behind them will not be discussed here. See Brooks (2008: 318-322).


14/28


15/28

15

tests using the cointegration regression Durbin-Watson and error correlation based tests are used

to ensure robustness of the EG results.

The EG technique involves estimating the long-run equation (2) then testing whether theestimated residuals obtained are stationary. Where the residuals are stationary the variables arecointegrated. The CRDW technique compares the Durbin-Watson (d) values computed fromestimating the potentially cointegrating regression equation (2) to some critical values under thehypotheses that d= 0 ord= 2. If the computed dstatistic is greater than the critical values thenthe series in the potentially cointegrating regression are cointegrated. The critical values are: 1%(0.511), 5% (0.386), 10% (0.322) (Gujarati, 1995: 726). Lastly, error-correction-basedcointegration tests whether potentially cointegrating series have a statistically significant errorcorrection term (see Kremers, 1989; Hendry and Ericson, 1991; Artis and Zhang, 1998). Sincethe LR pass-through and the asymmetry in pass-through will be determined based on the errorcorrelation model, it is important to give more attention to the error-correction-based test.

Assuming that cointegration exists between the central bank rate and the retail rates, adynamic model that ties any short run disequilibria to the long run equilibrium can berepresented as follows:

(8)

where is the error correction term [i.e. residual from OLS estimation

equation (2)], whose coefficient, is a measure of degree of monthly adjustment towards longrun equilibrium. The error correction cointegration test involves testing whether is statisticallydifferent from zero, in which case cointegration is said to exist between the official rate and theretail rate. As Aziakpono et al. (2007) note, a negative and statistically significant would implythat any short run deviation of the official and retail rate from their long-run equilibrium willcorrected through the operation of market forces.

Once an error correction is calculated it is possible to ascertain the speed at which thebank rates adjust back to equilibrium after a change in official rates. This is done by computingmean adjustment lags which, in the context of the study, can be interpreted as indicating the

speed of the IRPT. Whereas the error correction term shows the adjustment towards equilibriumwithin a month, the MALs show the total time taken (speed) to return to equilibrium. FollowingDoornik and Hendry (1994), the MAL is calculated from equation (8) as follows:

1 / (9)

ttt

m

j

jtj

n

i

itittCBRRRRRCBRCBRRR +++++=

=

=

)( 1111

10


16/28

16

In the case of monthly data the MAL presents how many months it takes for the changein Central Bank rates to be fully18 reflected in bank lending and deposit rates. If the mean

adjustment lag is high, then there is a high rigidity/slow adjustment in the response of bank ratesto policy rate changes. The opposite would be true with a low mean adjustment lag suggestinglow rigidity/fast adjustment of bank rates to policy rate changes.

These MALs are more correctly described as symmetric MALs as they reflect theresponse of bank rates whether they are above or below their equilibrium level. There is a way tocompute the asymmetric response of bank rates to Central Bank rate changes. These asymmetriclags would show how long it takes bank rates to adjust up to equilibrium and down toequilibrium. These asymmetric MALs would effectively show how fast bank rates adjustupwards and downwards.

To determine the asymmetric effects Scholnik (1996) suggests the separation of the

residuals (here marked as ) from the cointegrating equation into two series and ,where:

, 0,

and (10)

, 0,

where is the mean of the error correction which is equal to zero since it is the residual series ofthe cointegrating equation. When a residual is above its mean it can be interpreted as the bank

lending/deposit rates being above their equilibrium level with the policy rates and consequentlyexpected to move down to equilibrium.Conversely, when the residual is below its mean it can be interpreted as the bank

lending/deposit rates being below their equilibrium level with policy rates and consequentlyexpected to move backup to equilibrium. By splitting the residuals it is now possible to observethe speed of adjustment up or down for lending and deposit rates after Central Bank ratechanges.

Once the residuals are split into two series an asymmetric error correction system iscalculated from which the asymmetric MALs can be calculated. The asymmetric error correctionequation is presented as follows:

tk

m

j

jt

n

i

ittt ECECRRCBRCBRRR ++++++=+

=

=

2111

10(11)

The relevant asymmetric mean adjustment lags become:

18 The word fully refers to the complete LR impact rather than a full reflection of the change in the Central Bankrate, i.e. if LR adjustment is only 80% then the MAL shows how long after the initial response in the bank rate ittakes for the full 80% response to be reflected.


17/28

17

1 / (12)

and

1 / (13)

The mean adjustment lags in equations 12 and 13 show the asymmetric adjustment in banklending and deposit rates when they are above and below equilibrium respectively. If the meanlags are different then the adjustments of the bank rates can be seen to be different.

However, whether or not true asymmetry exists requires the use of the Wald test with a 1 distribution on the restriction that equations 7 and 8 are in fact equal19. If the Wald testreveals asymmetric responses when residuals are above equilibrium (and bank rates are pushedto move down) and when residuals are below equilibrium (and bank rates are pushed to moveup) then it can be concluded that bank rates will adjust differently during periods of

expansionary monetary policy, when bank rates are expected to rise, and contractionarymonetary policy when bank rates are expected to fall, as indicated in the theory discussed earlier.

5. RESULTS

5.1 Stationarity and Cointegration Test Results

The initial step involved testing the series for stationarity. The results are reported inTable A1 (in Appendix). Most the series were I(1). Although in rare circumstances some serieswereI(0), this was not simultaneous across both policy and retail rate, thus, the second stage wasto test all the series for cointegration. Cointegration was done both for the entire period and in 8-year rolling windows. Results for both the entire period and the rolling windows, along withcorresponding average concentration ratios (and their rankings), are reported in Table A2.

Given that three tests of cointegration are used, the procedure in deciding if series arecointegrated follows Kremers et al. (1992), who suggests that cointegration exists when one ofthe tests confirm cointegration at least at a 5% level of significance or at least two of the testsconfirm cointegration at 10% level of significance.

Starting with the entire period, the cointegration results show strong evidence ofcointegration in South Africa and Zambia where all the tests identify cointegration at least at 5%for both lending and deposit rates. In the case of Nigeria deposit and lending rate cointegration is

confirmed by only two of the tests. Botswana has the weakest evidence of cointegration for bothlending and deposit rates and the cointegration is only identified by the ECM coefficient test.Particularly noticeable in the results is that the coefficient of the ECM strongly confirms thepresence of cointegration in all the countries for both lending and deposit rates at a 1% level ofsignificance. This is important given Artis and Zhangs (1998: 5) assertion that the ECM statisticfor testing cointegration can generate more powerful tests than those based on the ADF and

19 The Wald test works on the null that the coefficients of the asymmetric error correction terms ( and inequation 6) are not statistically different from zero. Rejecting this null suggests that the responses of bank rates toimpulses to move down to equilibrium and up to equilibrium are in fact asymmetric.


18/28

18

CRDW statistics. For this reason the results still demonstrate strong evidence of cointegrationeven where the ECM test is the only one to confirm its presence. Noteworthy is that the pattern

of concentration does not seem to be related to the existence of cointegration in the series. Forinstance, SA, whose bank concentration level is marginally lower than that of Botswana (withthe highest bank concentration), has the strongest and Botswana has the weakest cointegrationresults, while Nigeria, with the lowest bank concentration level, also has stronger cointegrationresults than Botswana.

In the case of rolling window cointegration results, there is cointegration for both SouthAfrican deposit and lending rates in all sample periods except 2000-2007(3)20. In all the periodswhere cointegration is found the finding is robust as at least two tests confirm its presence at the5% level. This is with the exception of the deposit rate 1999-2006 (1) and 2000-2007(3) whereonly the CRDW test confirms cointegration at 5% while the ADF test confirms it only at 10% in

the former case and no test confirms cointegration in the latter case. For Botswana, all lendingrates are cointegrated with Central Bank rates. However, cointegration does not exist betweendeposit rate and Central Bank rates for four rolling windows: 1996-2003(3), 1997-2004(4), 1999-2006(6), and 2000-2007(7). In the cases where cointegration is found it is confirmed by at leasttwo of the three tests. In the case of Nigeria, in contrast to Botswana, all deposit rates arecointegrated with Central Bank rates but not all lending rates are cointegrated with policy rates.Lending rates are not cointegrated with Central Bank rates in the following rolling-windows:1995-2002(4), 1996-2003(5), 1997-2004(6), 1998-2005(7) and 1999-2006(8). As withBotswana, where cointegration is found the results are robust and are confirmed by at least twoof the tests. Zambia is the only country in which the lending and deposit rates show evidence ofstable long-run relationships with the Central Bank rates in all the rolling windows. The resultsare also robust here and are confirmed by at least two of the tests for each period.

Changes in average bank concentration across the rolling periods do not seem to besystematically related to the cointegration results in any of the countries or across countries. Forexample, while the increase in average bank concentration from rolling window 1998-2006 torolling window 1999-2007 seem to be associated with weakening cointegration between theofficial rate and both lending and deposit rates, the decrease in average bank concentration fromrolling window 1999-2007 to rolling window 2000-2008 is associated with even moreweakening cointegration between the official rate and both lending and deposit rates. The same

is evident in Botswana, where in some cases increasing bank concentration is associated withstrengthening cointegration results and in other periods weakening bank concentration isassociated with strengthening cointegration results. Across countries, for SA, while the mostconcentrated rolling periods seem to be associated with weak/no cointegration, it is the leastconcentrated rolling periods that show weak/no evidence of cointegration in Botswana andNigeria.

20 The number in brackets denotes cointegration rank with 1 being the most concentrated and 7 being the leastconcentrated.


19/28

19

One explanation for these results is that IRPT hinges on more than just the level of bankconcentration: other factors, such as the ownership of the banks, legislation and bank

supervision, may determine whether there is a long-run relationship between policy rates andbank lending and deposit rates (cf. Allen and Gale, 2003).What the analysis on the two levels suggests is that bank concentration may not be an

influence in the relationship between bank and central bank rates. Put differently, whether or notbank lending/deposit rates and central bank rates move together appears to be independent of thelevel of banking sector concentration. However, more scrutiny is necessary before acomprehensive conclusion can be drawn.

5.2 Symmetric and Asymmetric Adjustments

For the periods where cointegration was found, symmetric error correction models were

estimated. Corresponding asymmetric error correction models were also estimated to separate theerror correction mechanism between periods when bank rates are above and below equilibrium.The results for both the entire period and the rolling windows together with the correspondingbank concentration ratios for each period are reported in Table A3.

In all four countries symmetric error correction for both lending and deposit rates issignificant, as shown by the fact that all EC t-1 values are significant at the 1% level. In additionlending rates adjust faster than deposit rates, suggesting a greater reluctance for banks to adjustdeposit rates than lending rates. However, since this is true in all cases, regardless of the level ofbank concentration, we can see that whether or not the lending rate adjustments are faster thandeposit rate adjustments is not related to the level of concentration. This result is in line with the

intuition that since lending rates represent returns on assets (loans) and deposit rates representexpenditure on liabilities (deposits), banks would adjust lending rates faster than they woulddeposit rates regardless of the level of concentration. In addition, it can be seen that the speed ofadjustment is not related to the level of concentration. For example, Botswana with the greatestconcentration has the slowest adjustment of lending rates while Nigeria has the leastconcentrated banking industry and the third slowest adjustment of lending rates.

In the case of rolling windows, as expected, the results in all periods in which nocointegration is identified, and the error correction is computed to perform the ECM coefficienttest, show that there is no statistically significant error correction both symmetrically andasymmetrically. For example, in Botswana for the periods 1999-2006 and 2000-2007 there is no

evidence of cointegration and the corresponding ECM values are not significant. However, in sofar as the symmetric adjustment is concerned there are some intuitively unappealing results inwhich cointegration is identified and yet the corresponding error correction is not significant. Inthe symmetric error correction this is the case for South Africa 1999-2006(1), Botswanasdeposit rate 1998-2005(7) and lending rates 1995-2002(4) and 1997-2004(6), and Zambiasdeposit rate 1998-2005(7). According to Gonzalo and Lee (2000), it is possible to observe such aconflicting outcome in cases where the underlying series are fractionally integrated. In suchcases the series are not actuallyI(1) as identified by the stationarity/unit root tests. Consequently,the cointegration results on such series are in fact spurious. In all cases, however, there is no


20/28

20

pattern between the changes in the bank concentration and the size or significance of thesymmetric error correction.

In terms of the asymmetric error correction, there are fewer periods in which both theasymmetric error correction terms are significant than periods in which at least one of them is notsignificant. Neither in the size nor significance of the positive (EC+t-1) and negative errorcorrection (EC-t-1) is there a clear relationship between bank concentration and the errorcorrection mechanism of the cointegrated series.

5.3 Speed of Adjustment: Mean Adjustment Lags

While the error correction coefficients show by how much bank rates adjust back toequilibrium in each month, they do not show how long the complete adjustment will take. Todetermine the speed of the adjustment of the bank rates to Central Bank rate changes the study

computes the mean adjustment lags as described in the analytical framework. Both thesymmetric and asymmetric mean adjustments are computed. The average levels of concentrationfor each of the periods in each of the four countries are then compared to the mean adjustment toevaluate if there is any systematic relationship between the two. The results for both the entiresample period and the rolling windows period are reported in Table A4.

Starting with the results for the entire period, Botswana has the highest level ofconcentration and the largest symmetric MAL/slowest adjustment for both deposit and lendingrates. After a shock to the equilibrium relationship it takes on average 2.6 months (approximately78 days) and 1.7 months (approximately 50 days) for deposit rates and lending rates respectivelyto return to equilibrium. It follows that for a discernable relationship to be seen between bank

concentration and the speed of adjustment, the country with the lowest concentration must havethe fastest symmetric adjustment/lowest MAL. However, despite having the lowest level ofconcentration Nigeria does not have the smallest MALs. The fastest adjustment for both depositand lending rates occurs in South Africa where, despite having the second highest level ofconcentration, it takes 0.7 months (approximately 19 days) and 0.5 months (approximately 13days) for deposit rates and lending rates respectively to return to equilibrium. This analysis doesnot show a relationship between the level of concentration and the symmetric adjustment of banklending/deposit rates to Central Bank rate changes.

In the case of the asymmetric adjustment, high bank concentration seems to be associatedwith slow adjustment down to equilibrium in lending rates. Botswana, with the mostconcentrated banking sector, has the slowest adjustment of lending rates followed by Zambia andthen Nigeria, which has the least concentrated banking sector. This suggests that the moreconcentrated the banking industry the slower the speed in the reduction of lending rates back toequilibrium. This is evidence in support of the Structure Conduct Hypothesis which suggests thatconcentrated markets are susceptible to collusion which would slow down profit-reducingreductions in lending rates. This evidence supports the findings of Corvoisier and Gropp (2002)which also identify evidence in support of the SCP in the banking industries of EU countries.However, some caution is necessary in drawing conclusions on these results as this relationshipexists in the absence of a significant MAL+ for South Africa, which had the fastest symmetric


21/28


22/28

22

In terms of the lending rates, as with the other countries in the study a clear relationship couldnot be identified.

The analysis of results identified traces of a relationship between bank concentration andthe symmetric and negative adjustment of deposit rates and falling lending rates, and moves onto an investigation of the relationship between bank concentration and the magnitude of theadjustment of bank rates in the short and long run, symmetrically and asymmetrically, inresponse to official rate changes.

5.4 Magnitude of Adjustment

As presented in Appendix A5, the short-run symmetric adjustment in deposit ratesfollowing a change in official rates is smallest in Nigeria, where concentration is smallest, and isat its largest in South Africa, where concentration is highest22. Based on this, the size of the

short-run adjustment of deposit rates is positively related to bank concentration. This is to saythat the greater the concentration, the larger the size of the adjustment. However, the symmetricadjustment of lending rates does not show a similar pattern as the inclusion of Botswana (withthe smallest adjustment) means that no clear relationship can be seen between concentration andthe size of lending rate adjustments.

Appendix A5 also presents the long-run symmetric and asymmetric magnitudes ofadjustment. As can be seen, there is no clear pattern between concentration and the symmetricmagnitude of adjustment of lending and deposit rates. In the case of lending rates, while Nigeriahas the smallest long-run adjustment and the lowest level of concentration, the largest adjustmentis not associated with the most concentrated country (Botswana).

The symmetric long-run adjustment results are in line with the findings of the short-runresults where no clear pattern can be seen between bank concentration and either a rising orfalling size of adjustment following a positive or negative change in official rates. However,unlike the case of the short-run adjustments, the Wald test shows that the response is onlyasymmetric for lending rates in Zambia and Nigeria and deposit rates in Zambia (countries withthe lowest level of concentration). This indicates that asymmetry in the long-run adjustment ofbank rates may be related to the level of banking sector concentration.

In South Africa the symmetric adjustment of long-run lending and deposit rates isnegatively related to the level of banking sector concentration in the rolling windows from 1994-

2001 to 1999-2006 (lending rates) and 1995-2002 to 1999-2006 (deposit rates). Increasingconcentration is associated with lower long-run adjustments in bank rates following changes inofficial rates. This is in keeping with the SCP that suggests collusive behaviour will hinder thefull reflection of changes in official rates in bank rates.

In Botswana the only sign of a relationship between the symmetric long-run adjustmentand bank rate adjustment can be seen in lending rates in the rolling windows 1994-2001 to 1996-2003 where falling concentration is associated with larger long-run adjustments. This is a similar

22 Botswana is eliminated from the analysis as it does not have a statistically significant value for the short-runadjustment. This leaves South Africa as the most concentrated country in so far as the symmetric short-runadjustment is concerned.


23/28

23

relationship to the one found in South Africa and it provides additional support for the influenceof the SCP in so far as symmetric lending rate adjustments are concerned (see Appendix). The

results for Nigeria do not give any evidence of a relationship between concentration and thesymmetric adjustment of both lending and deposit rates. In contrast the evidence from Zambia issimilar to the findings for Botswana and South Africa where a relationship can be seen betweenfalling concentration and larger symmetric magnitude of adjustments for lending rates.

In so far as the relationship between the level of concentration and the asymmetricresponse of the long-run magnitude of adjustment is concerned, there is some evidence of arelationship with the positive long-run adjustment and none with the negative long-runadjustment. Falling concentration is associated with larger long-run adjustments in deposit ratesfollowing both increases (LR+) and decreases (LR-) in the official rate. The results for lendingrates do not depict any relationship.

In Botswana the only evidence of a relationship between the asymmetric adjustments andthe level of concentration is in the lending rates. Over the sample period a falling level of bankconcentration is associated with a larger long-run positive adjustment in lending rates followinga positive change in the official rate. In Nigeria there is evidence that both lending and depositrates positive long-run adjustments are related to the level of banking sector concentration. Inthe rolling windows 1996-2003 to 1999-2006 decreases in the level of concentration wereassociated with increases in the size of the positive long-run adjustment of deposit rates, while inthe rolling windows 1994-2001 and 1995-2002 falling concentration is associated with adecreasing size in the positive long-run adjustment of lending rates. In Zambia the positive long-run adjustment of lending rates is positively related to the level of concentration. As theconcentration level falls the size of the positive long-run adjustment also falls.

Of particular importance is that in all four countries, whether the size of the negativeadjustment is smaller or bigger than the corresponding positive adjustment for deposit andlending rates, it is not related to the level of banking sector concentration.

6. CONCLUSION AND POLICY IMPLICATIONS

The paper analysed the relationship between bank concentration and interest rate pass-through infour SSA countries. Also analysed is whether bank retail rates respond asymmetrically tochanges in the official rate and the whether this asymmetry is associated with levels or changes

in bank concentration. The analysis was done both for the entire period (to see whether a countrywith the highest concentration has the most effective monetary policy or otherwise) and eightyear rolling windows (to trace whether the dynamic responses in pass-through are linked tochanges in average bank concentration). The findings of the study are as follows:

Regarding the speed of adjustment, there is evidence that in some cases, bankconcentration is related to the speed of the IRPT. In the analysis of the entire period there isevidence of a negative relationship between bank concentration and speed of lending rateadjustments down to equilibrium suggesting that the greater the concentration, the slower thereduction in lending rates by banks following a change in official rates. Put differently, bankconcentration can be seen to negatively impede expansionary monetary policy when lending


24/28

24

rates are expected to fall. However, once the analysis is moved to the rolling windows, the trendover time suggests a relationship only in the adjustment of deposit rates. In addition, where the

trend over time shows a relationship between bank concentration and the symmetric and negativeadjustment of deposit rates, the nature of the relationship is not consistent. In Botswana andZambia the relationship is a negative one, suggesting that greater bank concentration results inslower adjustments of deposit rates, while in Nigeria it is a positive one, suggesting that greaterbank concentration is associated with faster adjustments in deposit rates. In addition the evidencefrom the Wald tests shows that bank concentration is not related to the presence of asymmetry inthe speed of adjustment of both lending and deposit rates. This result is the same for the analysisof the entire period and the analysis through the rolling windows.

In so far as the magnitude of the adjustment is concerned; there is more evidence tosuggest a relationship with bank concentration than there is to suggest a relationship between

bank concentration and the speed of the IRPT. There is a positive relationship between thesymmetric adjustment of deposit rates and the level of bank concentration in the analysis of theentire period. This analysis also shows a relationship between falling concentration and largerlong-run negative adjustments in lending rates which suggests that the responsiveness of banksto monetary policy impulses to reduce lending rates is negatively affected by bank concentration.In other words, bank concentration stifles the magnitude of the IRPT and by implication theeffectiveness of monetary policy. In the rolling window analysis the evidence suggests arelationship with deposit rates as well. However, in some cases the relationships are positivewhile in others they are negative, suggesting that there is no consistent relationship between bankconcentration and the magnitude of the IRPT in both the short run and the long run,symmetrically or asymmetrically. In some cases there is evidence that supports the StructureConduct Performance hypothesis while in others there is evidence supporting the EfficientStructure Hypothesis.

It is important to note that the relationships were not consistent between the countries andacross the two levels of analysis. This suggests that the relationships were either not the naturaloutcome of bank concentration changes and IRPT (supporting researchers such as VanLeuvensteijn et al., 2008) or that there are some mitigating factors that are preventing theobservation of the relationship. While the true explanation may still be debated, what is clear inthe evidence is that bank concentration is not consistently related to the speed and magnitude of

the IRPT. As a result, while the results show that bank concentration can be negatively andpositively related to the speed and magnitude of the IRPT it should not be expected that greaterconcentration will either automatically retard or accelerate the IRPT and make monetary policytransmission any more or any less effective. Additionally, the fact that the relationships observedbetween bank concentration and the IRPT cannot be captured neatly by one hypothesis suggeststhat the influence of bank concentration will not be universal but will capture the trend within acountry at any one time. Consequently, trying to identify a single relationship will continue to provide inconclusive results where the reality is that bank concentration can be related to theIRPT in different ways in different countries at different times.


25/28

25

The most important implication of these results is that the African trend towards bankconsolidation (in line with the global trend noted earlier) is not to be immediately regarded with

alarm in so far as effective monetary policy is concerned. Simply because concentration is likelyto increase is not to say that the IRPT will weaken or monetary policy will be less effective.However, this is not a vindication of proponents of increased bank consolidation andconcentration. Such a conclusion can only be informed in part by this study because there arestill other concerns around the operation of highly concentrated markets, for example, thepotential abuse of market power access to financial capital.

As competition commissions consider legislation relating to policy on acceptablemergers, acquisitions and consolidation exercises in the banking industry they may now do soinformed of the fact that the evidence in the selected African countries in this study is that thelevel of banking concentration can be, but is not always, related to the either the strength or

weakness of the IRPT of policy changes to bank deposit and lending rates.


26/28


27/28


28/28

28

KWIATKOWSKI, D., PHILLIPS, P. C. B., SCHMID, T. P., and SHIN, Y., 1992.Testing thenull hypothesis of stationary against the alternative of a unit root. Journal ofEconometrics, 54, 159178.

LAGO-GONZLEZ, R., and SALAS-FUMS, V., 2005. Market power and bank interest rateadjustments.Documentos de Trabajo Banco de Espaa, No. 0539.

LI, X., 2003. The Long-run and Short-run Multipliers of Fiscal Policy in the Chinese Economy.Massey University, Commerce Working Paper No. 03.02.

MACKINNON, J. G., 1996. Numerical distribution functions for unit root and cointegrationtests.Journal of Applied Econometrics, 11, 601618.

MISHKIN, F. S., 1995. The symposium of the monetary transmission mechanism. Journal ofEconomic Perspectives,9(4), 310.

MAROTTA, G., 2009. Structural Breaks in the Lending Interest Rate Pass-Through and theEuro.Economic Modelling, 26: 191-205.

MUSONDA, A., 2008. Deregulation, Market Power and Competition: An EmpiricalInvestigation of the Zambian Banking Industry. Centre for Study of African Economies.

NEUMARK, D. and SHARPE, S. A. 1992. Market Structure and the Nature of Price Rigidity:Evidence from the Market for Consumer Deposits. The Quarterly Journal of Economics,107(2), May, 657680.

QUAYYUM, A., KHAN, S., and KHAWAJA, I., 2005. Interest Rate Pass-through in Pakistan:Evidence from Transfer Function Approach. The Pakistan Development Review, 44 (4),9751001.

RASCHE, R., and WILLIAMS, M. M., 2005. The Effectiveness of Monetary Policy. FederalReserve Bank of St. Louis Working Paper No. 2005-048B.

SCHOLNICK, B., 1996. Asymmetric adjustment of commercial bank interest rates: evidencefrom Malaysia and Singapore.Journal of International Money and Finance, 15(3), 485

496.SEDDIGHI, H. R., LAWLER, K. A., and KATOS, A. V., 2000. Econometrics: A Practical

approach. London: Routledge.SOLUDO, C. C., 2004. Consolidating the Nigerian banking industry to meet the development

challenges of the 21st century. Address by Mr Charles Chukwuma Soludo, Governor ofthe Central Bank of Nigeria, at the Special Meeting of the Bankers Committee, Abuja, 6July 2004,BIS Review, 43/2004.

TOMASZ, C., 2003. Interest rate pass-through in the Polish banking sector and bank-specificfinancial disturbances. Munich Personal RePEc Archive (MPRA) Paper No. 5133.

VAN LEUVENSTEIJN., SORENSEN, C. K., BIKKER, J. A., and VAN RIXEN, A. A. R. J. M.,2008. Impact of Bank Competition on the Interest Rate Pass-Through in the Euro Area.

Working paper series, No. 885, European Central Bank.

Ersa Working Paper Proposal - Mangwe and Ziv

Documents

Transcript of Ersa Working Paper Proposal - Mangwe and Ziv