THE ECONOMICS OF RETAIL BANKING - An empirical analysis of

THE ECONOMICS OF RETAIL BANKING - An empirical analysis of the UK market for personal current accounts Céline Gondat-Larralde and Erlend Nier* 10 December 2003 (Preliminary. Please do not circulate or cite without permission) Abstract: This paper provides an analysis of the competitive process in the market for personal current


Texto en PDF


 THE ECONOMICS OF RETAIL BANKING rket for personal current accounts Céline Gondat-Larrald10 December 2003 (Preliminary. Please do not circulate or cite without permission) This paper provides an analysis of the competitive process in the market for personal current describe some stylised developments in this market over the past few years. We find a gradual ch market shares over time. This contrasts with a marked dispersion in price, evolution of market shares, we address two key price differentials? (ii) If not, which type of imperfect competition best fits the data? Our conclusions point to the existence of customer switching costs as a key determinant of the nature of competition in the market for personal current accounts. Keywords: microeconomics; retail banking; competition; switching; price elasticity The views expressed in this paper are those of the authors and do not necessarily reflect those of the Bank of England. We would like to thank Charles Goodhart, Patricia Jackson, Kevin James, Glenn Hoggarth, Darren Pain, William Perraudin and Geoffrey Wood for useful discussions on an earlier draft of the paper and Jon Chu and Nyeong Lee for research assistance. _________________________________________ * Both authors are at the Bank of England. Céline Gondat-Larralde can be contacted at celine.gondat-
[email protected]
, Tel:++44 20 7601 3328. Erlend Nier can be contacted at [email protected]
Tel: ++44 207 6013239. Correspondence address: Bank of England, HO-3, Threadneedle Street, London EC2R 8AH.
Motivation This paper analyses the competitive process in the UK market for personal current accounts. Bank current accounts play a pivotal role in the relationship between a bank and its customers: current services), money transmission through cheques and deoverdrafts. As such, they are important for building relationships between a bank and its customers and may serve as a gateway through which suppliers canWe first document some stylised facts as regards the developments in the market for personal current accounts over the past few years. We find that the distribution of markover time. Against this, there is a marked dispersi(i) Are bank market shares res(ii) If not, which type of imperfect competition best fits the data? rst analyse the speed of adjustment of market shares in response to price differentials, taking into account the fact that price differentials may well reflect differences in ribution of the levels of market ices to distinguish empirically between a number of competing hypotheses as to why this adjustment may be slow. A large part of the empirical literature that attempts to analyse competition in banking is based on the Structure-Conduct-Performance (SCP) paradigm, whicstructure, the firms’ conduct and ultimately their performance. As regards the level of competition in

banking markets, overall these studies have not led to firm conclusions. While some studies have found bility and measures of market structure, many other studies have failed to . One major problem with this literature is that it is not built on a firm theoretical footing. In particular, a positive relationship has been subject to different interpretations and some economists have argued that the causality between structure and performance is reverse – i.e., firms with higher management skills and/or technology, or producing at more efficient scale will have lower costs and therefore higher profits, and as a result will gain large market share that may result in a higher market concentration level, Demsetz (1973). More recently, the contestable market theory has questioned the link between market concentration and performance by emphasising the importance of entry conditions, instead of the market concentration itself, to explain the degree of competition of a marketAnother strand of empirical studies, which is sometimes referred to Organization approach, attempts to estimate a parameter of a structural model that directly measures the degree of imperfect competition. For instance, a number of studies estimate the Panzar-Rosse statistic, which measures the extent to which changes in a firm’s input prices are Typically, in these studies, the parameter estimates fl variations over time that are hard to interpret as changes in the degree of competition. In addition, while in most cases the evidence has been in favour of imperfect competition, as opposed to perfect competition, the test employed is not sharp enough to distinguish between various types and sources of imperfect competition. Finally, the Panzar-Rosse statistic is estimated at the industry level. This assumes that the
This has also been argued by Gilbert (1984), among others. For a review of studies on bank market structure and competition, see Gilbert (1984) or Berger (1995). Berger (1995) proposed a way to distinguish between the different interpretations of the positive link between market structure and performance in banking. See Baumol, Panzar & Willig (1982), Contestable markets and the theory of industry structure, Saunders College Publishing/Harcourt Brace. Panzar & Rosse (1987), “Testing for monopoly equilibrium”, Journal of Industrial Economics, 35, 443-456. De Bandt & Davis (2000) measure the Panzar-Rosse statistic for several European banking markets.
degree of competition is the same in each product market in which the banking firms are active. Arguably, however, competitive conditions may vary significantly from one market to another. Our study directly builds on recent work by Heffernan (2002), who analyses the pricing behaviour of British banks for retail products and provides one of the few exceptions to the empirical literature in that types of imperfect competition. Likemodel(s) of imperfect competition best describe the UK current account market. However, we devise a the level of a firm’s market share and the price it sets. This test allows for a broader set of compettype of friction that may be affecting the competitive process in this market. Frictions in the market for personal current accounts A number of potential frictions may be present in the market for personal current accounts. On the demand side these may be related to switching costs and search costs borne by bank customers. On the supply side they may relate to fixed costs of entry borne by banks (economies of scale). Switching costsSwitching costs may be defined as those costs that a customer incurs when switching accounts from one Switching costs may have several originsers may involve transaction costs. Such costs are likely to arise from the ninflowing payments. Since switchistomer leaving his established ng current accounts may potentialy also result in an increase in asymmetric information between the bank and the customer. Moreover, in some cases, firms may find it in their interest to create artificially or to increase the switching costs their customers face through contractual demption penalties are introduced do not seem to exist in the UK market for current
For a taxonomy of switching costs, see Klemperer (1995).
. Some banks now offer to smooth the switching process by offering a “ready-made” kit to customers wishing to switch to them. Moreover, the BACS members recently introduced an automated system for exchanging information on switching customers' direct debits. Customers are more likely to switch providers if the net benefit from switching is high. The net benefit research calculates the possible moneconsumers could expect to derive from switching providers . According to this research the current account market shows one of the highest dispersion in prices, in terms of the loss consumer faces if she chooses to However, the monetary loss from not choosing the cheapest current account provider, when average balances are taken into account, is relatively small (£26 p.a.) compared to most other products (e.g. £230 p.a. for a variable rate mortgage). On the otheaccess savings accounts – products that may be linked to current accounts – are characterised by both a ively) and a significant monetary loss (£142 p.a. om switching current accounts may therefore appear limited for the average customer, when focusing exclusively on the current account rate dispersion, this may not
As a response to Cruickshank (2000), the government asked a group chaired by D. Julius to make recommendations as regards changes in the Banking Code. One of the recommendations of the Group was to make account switching easier. See Julius (2001) for more detail. Financial Services Authority (2002). The figures quoted in this paragraph are extracted from Table 2, page 15 in FSA (2002). One caveat to these calculations may be that the expected gross benefits calculated in FSA (2000) only relate to price differentials and may thus only partially reflect the ‘true’ gross benefits from switching current account providers. In particular, they do not account for potential (non-monetary) improvement in the “quality” of service (e.g. new product features, better management by/relationship with the new provider, better access to other products) associated with switching. If quality and price are positively related, then the monetary benefit from switching may be reduced when a potential deterioration in quality as a result of switching accounts is taken into account.
However, the customer will need to weigh the gross benefit of switching against the cost of switching, sts are difficult to quantify and may well differ across customers. For instance, switching costs may well be higher for customers who expect to use the ction costs could be more relevant selection switchers may not immediately be granted suggest that switching costs may current account market when compared to the gross benefit customers can achieve from switching. In particular, data on current account switching behaviour from the lifetime. Switching costs make it costly for a customer to leavcosts will thus only be relevant to those customers account. There might also be frictions that do not have this feature. These may result in costs that will be incurred by customers when they open a current account for the first time as well as by existing current account holders. are a key example of this typecurred when a consumer starts e market that best fit her preferences. Search costs may be incurred either when the customer is cucustomer. Search costs can be substaFor current accounts, search costs may have gone up recently as the complexityffect of this increased product complexity and differentiation is a priori ambiguous: on the one hand, search costs may have increased because complexity makes it more difficult for consumers toproducts may increase consumers’ welfare. (iii) Economies of scale d search costs, which are primarily demand-related, there might be supply-related factors that would result in the market for personal current accounts to be less than highly competitive. An example is the presence of economies of scale that results from exogenous or e costs of setting up and maintainlikely to be substantial, even though recent developments in technology may well have reduced the
minimum efficient scale for some firms (e.g. Internnd branding. This could mean that in equilibrium banking markets are more concentrated than under the assumptions ofcompetition, resulting in Outline of the paper The remainder of this paper is organised as follows. We first present some stylised facts on the UK current account market: after describing our data sources (2.1), we show how individual banks’ market shares, current account characteristics (prices as well as quality char firm-level demand with respect to prices that are associated with the current account. In section 4 we provide evidence related to the type of imperfect competition in this market. Finally, section 5 concludes the paper. DESCRIPTION OF THE DATA Data on the number of current accounts per bank The data on the number of current account customers per bank are obtained from the NOP (National l Research Survey (FRS), among 5,000 individuals selected randomly each month. Polled households are asked detailed questions about mortgages, loans, etc. as well as their demographic characteristics (age, gender, income, working status, living, etc.).
From this source we were able to obtain the numomers per bank on a half-market share, in terms of number of customers as a time series Data on prices (i.e. interest rates) tly, or indirectly, associated with a current account. First, we look at the interest rate offered on positive balances in the customers’ current accounts. Second, since most customers would have the option to arrange for an overdraft facility associated with their current account, we analyse the rate a bank charges on authorwe take account of the possibility that banks may attempt to cross-sell savings products to their current account customers. The distinction between a current account and a savings account is that the latter service does not include money transmission services (through cheques or diremay offer a better interest rate. When a customer account to a saving account, this transfer may be facilitated if both accounts are held at the same . We thus also include the rate offered on instant access saving accounts in our analysis. Each month, the Moneyfacts review publishes the rates quoted by mo) (£1,000 minimum balance with overdraft facility) by each bank in our sample; The interest rates each bank receives on authorised overdrafts (r) (£500 minimum balance) by each
The NOP FRS data were accessed through the X-Press system. This interface allows access to aggregated data (i.e. at a bank level), rather than the “raw” data - i.e. the data at a (polled) individual level. From the point of view of the customer, it may be convenient to hold both types of accounts with the same bank, especially if an automatic sweep facility exists between the two accounts.
time series on fees applicable to current account services for our sample. Data on non-price current account characteristics In our analysis we also attempt to take account ofe number of branches and the number of ATMs and over time were obtained from the and the Association for Payment Clearing Services (APAinformation on the range of transactions (e.g. management of standing orderspayment, transfers) a current account holder can perform over the phoneThis information was available for most banks from the “Which? Magazine” website. However, unlike the information on branches and ATMs it relates to a particular point in time (December 2002). Stylised facts Changes in market concentration records a very gradual decrease from 1,425 to 1,217 over our sample period and suggests that the current account market is moderately concentratedThe decline in this index suggests that the current account market is very gradually becoming less concentrated.
The Herfindahl-Hirschmann index (HHI) summarises the degree of concentration in the market for current accounts, by summing the squared market shares of all banks in our sample. By convention, each market share is multiplied by 100, e.g. if the market share is 1, it enters as 100. As a result, the HHI ranges from 0 to 10,000. The US Department of Justice considers a market with a Herfindahl-Hirschmann Index (HHI) below 1,000 as unconcentrated; one with an HHI between 1,000 and 1,800 as moderately concentrated and one with an HHI above 1,800 as concentrated.
10
1,0001,2001,4001,60096H197H198H199H100H101H1
Indexn the aggregate development by fo15,16the ‘building societies’: this peer group includedemutualised (Abbey National, Alliance & Leicesmprises those banks that essentiaInternet (Cahoot, Citibank, First Direct, First-e, Intelligent Finance Smile and Virgin Direct);e Bank, Girobank, Royal Bank of
See Table A in Appendix 1. Some of the banks in our sample are linked by ownership. In principle, we have kept separate entities in our sample if parents and subsidiaries have retained separate retail franchises. For example, we have included Nat West rather than the post-merger RBS Group in the group of the Big Four, since NatWest is considered to have a separate retail franchise from RBS. Adding RBS back onto the figures for NatWest changes the level but not the time profile of the market shares. Similarly, during most of the sample period, Halifax and Bank of Scotland were separate entities. As regards the direct banks, some of these are subsidiaries of other banks in our sample. For instance, Cahoot is owned by Abbey National, First Direct by HSBC, Intelligent Finance by Halifax and Smile by the Cooperative Bank. Again, in these cases we treat subsidiary and parent as separate retail entities, as ‘direct’ banks and parent companies have retained separate retail franchises.
Chart 2a shows that while the combined market share of the “big four” banks in the market for current llen by some 7 percentage posocieties – including those that demutualised - have made significant inroads into the market for current some 9 percentage points over this development has been helped by strong consumer recognition of their brands. At the same time, or other electronic means, have been able to increase their market share quite steeply, albeit from a low base, Chart 2c. But the absolute increase, at 1 % points, is smaller than that for the former buildistill only accounted for some 2% Bank of Scotland, Bank of Scotland and small and medium banks such as Clydesdale or Yorkshire Bank) has decreased from 16% to 14% over the sample
Among the Big Four, HSBC/Midland is the only bank not to lose market share. Much of the increase is due to the successful expansion of First Direct Bank, whose market share increased from 1.2% in 1996 H1 to 1.6% in 2001H2.
12
96H197H198H199H100H101H1
Per Cent
96H197H198H199H100H101H1
Per Cent
96H197H198H199H100H101H1
Per Cent
96H197H198H199H100H101H1
Per CentCharts 3a to 3c show how the average interest rates quoted by banks within each of the four peer groups defined in Table A (see the Appendix 1) have evolved over time. Price behaviour varies markedly by . This is mainly due to new direct banks offering higher rates. Current or ex building societies, market
1996 –2001
)
Chart 2c: ‘Direct’ banks, market share (1996 – 2001)Chart 2d: All others, market share (1996 – 2001)
13
0.01.02.03.04.05.06.07.08.096H197H198H199H100H101H1
Big 4
Building societies or ex building societies
Direct banks (a)
All others (b)
Base ratePer cent
(c)(d)
0.01.02.03.04.05.06.07.08.096H197H198H199H100H101H1
Big 4
Building societies or ex building societies
Direct Banks (a)
All Others
Base ratePer cent
(c)(d)
(b)
96H197H198H199H100H101H1
Big 4
Building societies or ex building societies
Direct Banks (a)
All Others (b)
Base rate
Per cent
(c)(d)The Big Four banks and the “Other” category show a similar pattern over timdespite significant movements in the Bank of England’ions turn out to offer the lowest rates on current accounts and charge th
Average current account rates within bank
Average instant access saving rates within
Excluding Virgin Direct. Excluding Co-operative Bank and Safeway. Citibank and Smile enter ‘direct’ banks. Cahoot, First-e and Intelligent Finance enter ‘direct’ banks.
Excluding Cahoot. Virgin ‘direct’ enters ‘direct’ banks. Citibank and Smile enter ‘direct’ banks. First-e and Intelligent Finance enter ‘direct’ banks.
Average overdraft rates within bank
Excluding Virgin Direct and First-e. Excluding Safeway. Citibank and Smile enter ‘direct’ banks. Cahoot and Intelligent Finance enter ‘direct’ banks.
y, these institutions offer among the lowest rates on instant access Price differentials do not appear to vary much through time. The apparent lack of variation through time is confirmed when the standard deviation of each variable of interest is decomposed into a ‘between group’ – i.e. cross-sectional – comp time series – component: for most price dispersion is proportionally more severe in r price dispersion across banks is differences in current account characteristics. The ‘direct’ banks the same range of services as traditional “bricks and mortar” banks. Measuring the characteristics attached to a product is a difficult task, first because one needs to pick the characteristics that matter for customers and second because of the scarcity of data on those different qualitative characteristics. We focus on four current account characteristics: First, the extent of a bank’s branch network may the bank. Graph 4a shows the distribution of the number of branches by peer group over time. For those banks that do not offer “bricks and mortar” facilities - i.e. the ‘direct’ banks (except Citibank which has a few branches in the UK) and Safeway Bank, a numberthis category, the relatively stable number of HSthe significant decline (-
This latter effect is mostly due to Woolwich’s behaviour in the first part of the period, and to the change in Halifax’s current account pricing strategy in the last year of the period. The coefficient of variation adjusts for differences in the mean of the series.
25%) seen for Natwest and Lloyds TSB experienced the number of branches at a bank level (averaged over time). In our econometric tests in sections 3 and 4, we scale the number of branches by the number of a bank’s customers and use this number as a proxy stomer has access to his or her bank manager).
199596979899200001
Big 4
Building societies or ex building societies
Direct banks
All others (a)
N
o. of branches
(a) Excluding Safeway
Average number of branches by peer group (1995-2001)
16
CitibankNorthern RockCo-opHFCYorkshire BankClydesdaleAlliance &LeicesterGirobankBank of ScotlandWoolwichNationwideAbbey NationalHalifaxHSBC/MidlandNatwestBarclaysLloyds TSBAv NBRANCHM network and hence the average proximity of an ATM for the customer indicates the convenience of cash management associated with the accountdifferently from the previous case as it is possible for the customers customers) at no cost. Hence we allocate the company to that ‘direct’ bank. GraATMs by peer group over time and Graph 5b displays the same information at a bank level (averaged over time). Almost all banks in our sample maintaperiod. In sections 3 and 4, we use the (logarithm of the) number for the density of its ATM network (i.e. how easy cash management is for a bank customer).
In July 2000, members of the LINK network abolished the so-called ‘disloyalty charges’ that were imposed by card issuers as a penalty on customers for using another member’s ATM. Even though few operators have since made use of this possibility, ATM operators remained entitled to impose a surcharge on customers for use of their ATMs. Overall, this means that for most customers, the ATMs of their own bank may have been the preferred means of access to ATM services during most of the sample period.
Chart 4b: Average number of branches by bank (1995-2001)
17
Chart 5a:Average number of ATMs by peer group (1995 - 2001)
199596979899200001
Big 4
Building societies or ex building societies
Direct banks (a)(b)
All others (c)
N
o. of ATMs
(a) Direct banks' NTAMs are parents' NATMs(b) Excluding First-e and Virgin Direct(c) Excluding Safeway
100015002000250030003500400045005000Northern RockCo-opSmileClydesdaleYorkshire BankWoolwichAlliance & LeicesterGirobankBank of ScotlandNationwideRoyal Bank of ScotlandHalifaxIntelligent FinanceAbbey NationalFirst Direct BankHSBC/MidlandBarclaysNatWestLloydsTSBAverage no. of ATMsactions that customers can perform over the phone and over the Internet respectively is used to proxy for ease of remoteand 7 show the distribution of the two indices in December 2002. Sitime series for these indices, we need to assume thatbution of the indices has not changed significantly over time, i.e. that level effects are common
Chart 5b: Average number of ATMs by bank (1995-2001)
number of operations related to standing orders, direct debits, bill payments, transfers and ordering a customer can perform remotely. The higher the number of operations a bank customer can perform remotely, the higher th
Phone Index1CahootWoolwichNorthern RockAbbey NationalNationwideNatWestIntelligent FinanceAlliance & LeicesterBank of ScotlandBarclays BankCitibankClydesdaleCo-opFirst Direct BankHalifaxHSBC/MidlandLloyds TSBSmileYorkshire Bank
Net Index1Northern Rock WoolwichAlliance & LeicesterBank of ScotlandClydesdaleNatWestAbbey NationalNationwide Yorkshire BankBarclays BankCahootCitibankFirst Direct BankHalifaxHSBC/MidlandIntelligent FinanceLloyds TSB Smile From graphs 4a and 5a, as well as 6 and 7 it a is becoming less of a ducing their branch network, commersome important reduction in their operating costs. At the same time, new information technology
Chart 6: Phone Index (end 2002)
Chart 7: Internet Index (end 2002)
developments have allowed banks to create new, cheaper ways to attract customers (i.e. ATMs, phone or Internet banking). Nowadays, most banks choosemultiple channels, including the internet and telephone. This might explain why we observe the entry of DETERMINANTS OF CHANGES IN MARKET SHARE The analysis carried out in the previous section shows that there is a gradual adjustment in bank market shares over time. It also shows that over the sames to persist. In this the gradual adjustment in bank market shares. To measure of how fast market shares vary in response to price differentials, we estimate the elasticity of bank-level demand with respect to The value of these elasticities will be indicative of the level of competition in the market for current accounts. In a highly competitive market, the (firm-level) elasticity of demand with respect to price is very high – in theory, infinite. Therefore, any price differential should trigger dramatic changes in market shares almost instantaneously. However, if competitive pressures are less acute because of the presence of frictions in the market, then the price elasticity of demand could be low. non-homogenous products, and characteristics may significantly differ from one current account to is case, the measured effect of price differentials on changes in MS could be small even if the market is close to perfect competition (zero at the limit if price differentials simply reflect differences in quality). Moreover, new product characteristics (more likely to be ents) may have appeared – e.g. phone or internet banking - that may imply a different sort of relations customers. Thus, some differences in prices, they need to be controlled for to avoid omitted variable biases.
In order to measure the dependence of bank-level demand on prices, we estimate the following [1] in bank i’s market share on the current account market measured on a half-year basiis the absolute difference (i.e. in percentage points) between bank i’s rate and the average rate quoted by the rest of the market, averaged over half-year t. Wethe rate on positive balances on current accounts (j=CA); the pre-authorised overdraft rate (j=OD); are the four non-price characteristics measuredcustomer in half-year t; the (logarithm of the) number of Automaticssumed to be constant over the stomer can perform over the phoneer can perform over the phonemi-elasticity of demand - how market shares ice dispersion in the market. In the case where a market is highly competitive, any price differential that is unrelated to quality differentials should trigger a significant
See Table B in Appendix 1 for variable definitions. Given the high correlation between these two indices, only one of them is used in multivariate regressions. The coefficient in equation [1] can be interpreted as a semi-elasticity because the independent variable RD is the absolute difference rather the relative difference between prices.
change in market share – i.e., the coefficient nt from zero and its value, in absolute terms, should be high. However, if a market is less competitive, then the (absolute) value of Unfortunately, there is no “accepted” threshold below which competition in a market would be situations between those two extremes. The test performed when estimating equation [1] will therefore competitive. However, we examine the responsiveness of banks’ market shares on the current account markdifferentials. This allows us to compare elasticities across rates: the current – both directly linked to current accounts – as well as the instant access savings markets. ges in market shares and price s are paid by the bank to the customer, whereas overdraft rates are paid by the customer, the expectrates is the reverse of the expectoffering a high rate on its current account (or on its instant access savings account) should see its market share increase. Such a positive relationship should be stronger, the more elastic the bank-level demand is. Similarly, a bank charging a lower overdraft rate than the market should see its market share incrshould be stronger, the more elastic the bank-level demand is. This translates into: these coefficients. the market would lead to all eme case, market shares would not respond to rate differentials (i.e. each coefficient would be zero). Several estimation techniques are avtween changes in market share l dataset: we can either use a ‘within’ or fixed-effect estimator, or focus on the cross-sectional dimension of the data (i.e. use a ‘between’-effect estimator), or exploit both (i.e. time and cross-sectional) dimensions by pooling th
do not appear to vary much through time and the changes in market share aapparent lack of variation through time is confirmed when the standard deviation of each variable of interest is decomposed into a ‘between group’ – i.e. cross-sectional – component, and a ‘within group’ – i.e. time series – component: for most variables, thchange in market share appears to be a gradual, a time series that is ranging from 1996 to 2001 is ssfully the dynamics ooled estimations. Our benchmark regressions (shown period and by then regressing time averages of the change in market share on time averages of the price tween’ regressions). We also estimaccount of both the time dimension and the cross-sectional dimension of the dataset in a symmetric way. the number of observations. But such a procedure also has important drawbacks. In sts performed on pooled regressions indicate poor properties for those estimators, including non-normal residuals and heteroskedasticity. Since diagnostic test for the ‘between’ estimations yield better results in these respects, we comment mainly on the ‘between’ regressions. That said, most of the results described in this seone are confirmed by pooled OLS regressions (shown in Appendix 2). Our analysis suggests that on average over the 1996-2001 period, changes in market share are sensitive to current account rates, but less sensitive to the other two rates. This is a plausible result given that the current account rate would be the rate most people focus on when choosing their current account provider. Importantly, these results appear to be robust to the inclusion of current account The three charts below plot each of the three rate differentials against changes in bank market share in the current account market, implicitly assuming that price differe
Of course, this result assumes that the non-price variables we use in our regressions are good proxies for the current account characteristics that bank customers value.
If the elasticity of demand wthe two variables – i.e. a small price differential would trigger a large change in market share. On the contrary, in the case where the price elasticity of bank-level demand is very low, one would observe a very steep, almost vertical line.
20020406080Percentage pointFirst-eIntelligent Finance
_
+
_
20020406080Price differential - Percentage point
_
+
_
Intelligent FinanceVirgin DirectSafewayFirst-e
20020406080Price differential - Percentage point
_
+
_
Intelligent Financechanges in market share and rate differentials for
Chart 8a: Average change in CA market share & CA rate differentials
Chart 8b: Avera
e chan
g
e in CA market share & IA rate differentials
Chart 8c: Average change in CA market share & OD rate differentials
line). This is consistent with a low elasticity of demand with respect to thsuggests a flatter relationship. This would imply a larger elasticity with respect to the CA rate. Indeed, te case is somewhere between a veryThe charts also allow us to check whether the data are consistent with our assumptions on the sign of the relationship between the changes in bank mark ) Indeed, there appears to e case of the overdraft rate, as expected. Finally, the charts reveal that some of the ‘direct’ banks in our sample may be potential outliersmay exercise a significant impact on the measured s in market share and rates linked to current accounts. Hence, we performe‘direct’ banks to check for the robustness of our re(in terms of normality of residuals, homoskedasticity, no pattern in the residuals) for the latter Table 1 below summarises univariate and multivariate regressions performed before introducing product characteristic variables in equation [1]. The percentage change in market sharregressed on a constant and on one rate differential at a time in columns (1) to (3) and on multiple rates simultaneously in the last two columns ((4a) and (4b)). We perform each regression on two different samples, by first including the seven ‘direct’ bankTo the extent that relative price and current account quality are not related, differences in relative price appear to explain a large part of the changes in market share, especially when the ‘direct’ banks are ssions that display the ll ‘direct’ banks are excluded) the current account rate differential coefficient is positive and significant (1b), as measured by the adjusted R-squared, is high, at 82%. In t access savings rate differential coefficient () is also significant and positive, but its value is much smaller, compared with the CA rate coefficient in regression (1b). Finally, the coefficient of the overdraft rate differential (cted but significant
This is confirmed by statistical tests.
ls are not included simultaneously inthat the overdraft rate differential may partially uding all three rate differentials simultaneously (see regressions ngs and the overdraft rate differentials together with the current e regression as compared to the case when the CA rate alone is included (see (1b)). Banks’ market shares still significantly react to current account differentials (=7.1), but much less to instant access savings rate differentials or from zero. Note that the last two cases are potentially different: in the savings rate case, we are examining the demand-price relationship across two , we are examining the demand-price relationship for gh not every current account holder may use such a facility. acility. a time in columns 5 to 8 and simultaneously in the last two columns (9a) and (9b) in Table 2. Given the small size of our sample, we face a trade-off: if we wish to account ith a relatively small number of degrees of freedom. The results suggest that some chtly influence changes in market share, particularly the extent of a bank’s ATM network and the extent to which some operations can be performed remotely. Nonetheless, the coefficient is positive, significant and relatively high when
As a first step we include a dummy variable that takes the value 1 if the bank is a ‘direct’ bank, and 0 otherwise. This dummy turns out not to be significant, suggesting that when price differentials are accounted for, the differences in product characteristics between the two groups do not of themselves lead to changes in market share. The results of these regressions are not shown in the paper but are available on demand. In principle, the number of ATMs could be endogenous – banks facing an increase in their customer base may need to increase their ATM network. This would pose a problem in particular in a regression of the change in market share on the
the ‘direct’ banks are excluded (banks are excluded from the samplection of product characteristics in the regression when ‘direct’ banks are anges in market share, just as in Table 1. ss their economic impact. Based on the results of regression (9b) and the value of the coefficient than its rivals would increase its market shar points over six monthsmeans that a bank maintaining such a differential (1996-2001) would increase its market share by 24.5 perting this and assuming that the relationship between changes in market share and price differentials does not evolve over time, it would take 18 years for such a bank to double its market share if it were to maintain such a price The regression (9b) also suggests that a bank that offers an instant access savings rate higher or an market share increase relative to the average. As a robustness check on these benchmark results, we also estimate regressionstime and the cross-section dimensionsessions). The advantage is that
changes in the number of ATM. However, our benchmark results are based on a regression of the change in market share on the level of ATMs, so that endogeneity is less likely to be a major issue. To give an order of comparison between the two samples (with and without the ‘direct’ banks), the change in market share averaged across all banks is 6.4 percent (respectively 1.2 percent when ‘direct’ banks are excluded from the sample). The average change in market share is a positive figure because it is the large banks that, on average, are losing market shares. (1+0.02)=1.24 or (1+0.02)
tests for these pooled regressions, in particular as regards normality, are not as good as in the case of sults (shown in Table 6 in the Appendix) are very similar to the ones shown in Tables 1 and 2: is always positive and significant, though slightly smaller (around 5) than what we fiin Table 2 that sometimes becomes siTo summarise our findings: mple that includes only traditional “bricks and mortar” banks (i.e. excluding ‘direct’ banks). is positive but much smaller than ity is taken into account, the IA saving rate differential does not influence banks’ market share changes. fferent from zero for most specifications. Overall, there is a moderate sensitivity of changes in market share with respect tocurrent account rate). The results appear consistent with the hypothesis that there may be some frictions in the market for personal current accounts.
Table 1: Changes in the Current Account Market Share as a function of Rate Differentials (when Product Characteristics are not (1) (2) (3) (4)
Dependent variable All banks (1a) ‘direct’ banks (1b) All banks (2a) ‘direct’ banks (2b) All banks (3a) ‘direct’ banks (3b) All banks (4a) ‘direct’ banks (4b)
CA rate differential RD 5.65 8.74*** 9.92* 7.12***
(P-Value) (0.141) (0.000) (0.054) (0.003)
IA rate differential RD 7.32** 1.09** 2.37 0.61
(P-Value) (0.031) (0.014) (0.226) (0.186)
OD rate differential RD -1.69** -0.61*** -0.05 -0.21
(P-Value) (0.045) (0.002) (0.902) (0.195)

Adjusted R-squared 26.5% 82% 37.3% 9.3% 14.1% 40.6% 78.1% 81.1%
R 0.55 0.91 0.63 0.39 -0.43 -0.67
Obs 21 15 23 17 21 16 19 15
F-test Not rejected Rejected RejectedRejected RejectedRejected RejectedRejected
Shapiro-Wilk Test (normality of residuals) Not rejected Not rejected Not rejected Not rejected
*** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. Robust estimates of standard errors. P-values in parentheses.
Table 2: Changes in the Current Account Market Share as a function of all Rate Differentials and Product Characteristic variablDependent variable (5) (6) (7) (8) (9)
All banks (5a) Excl ‘direct’ banks (5b) All banks (6a) Excl ‘direct’ banks (6b) All banks (7a) Excl ‘direct’ banks (7b) All banks (8a) Excl ‘direct’ banks (8b) All banks (9a) Excl ‘direct’ banks (9b)
CA rate differential 9.93* (0.067) 7.88 ***(0.008) 9.92** (0.048) 7.82*** (0.001) 9.23 (0.104) 7.87*** (0.002) 9.53 * (0.087) 7.22 *** (0.005) 8.73 (0.143) 6.68** (0.025)
IA rate differential 2.24 (0.336) 0.41 (0.385) 2.37 (0.244) -0.07 (0.873) 3.32 (0.364) -0.11 (0.849) 2.96 (0.369) 0.07 (0.899) 3.52 (0.391) -0.67 (0.352)
OD rate differential -0.06 (0.890) -0.21 (0.249) -0.05 (0.902) -0.16 (0.394) 0.02 (0.962) -0.24* (0.059) -0.02 (0.959) -0.27 (0.130) 0.03 (0.958) -0.15 (0.118)
Number branches/customer -0.81 (0.828) -1.36 (0.441) -2.94 (0.637) 5.11** (0.021)
Log (Number ATMs) -0.01 (0.996) 1.73** (0.019) -0.51 (0.839) 2.71** (0.012)
Phone index -0.24 (0.652) 0.23* (0.059) -0.46 (0.524) 0.34*** (0.000)
Net Index -0.11 (0.818) 0.22* (0.065)
Adjusted R-squared 76.6% 80.5% 76.6% 85.5% 76.4% 83.6% 76.2% 82% 72.5% 89.7%
Number Observations 19 15 19 15 18 14 18 14 18 14
Rejected Rejected RejectedRejected RejectedRejectedRejectedRejected Rejected (limit case) Rejected
Shapiro Swilk test (normality of residuals) Not Not Rejected Rejected Rejected Not Not
*** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. Robust estimates of standard errors. P-values in parentheses.
DETERMINANTS OF PRICES In this section we attempt to distinguish empirically between a number of diffethe nature of competition in the market for personal current accounts. The starting point is the through time. Price dispersion may be an indication of some form of imperfect competition. For example, price dispersion can be sustained in a dynamic model of competition where customers face rnatively, price dispersion is a could also be consistent with perfect competition and/ or static Cournot competition without switching or search costs, when there are different market segments that are differentiated by different levels of The key to distinguishing empirically between different models of imperfect competition is to draw out the implications of these models fo. In particular, it turns out that these different models have differethe empirical relationship between individual bank market shares and prices. Under the model of dynamic competition with switching costs (Kim, Kliger and Vale (2003)), it can be shown that there is a positive relationship between a firm’s market share and the price it charges its customers. That is, a larger bank would tend to chargebalances, this means that a larger bank would offer overdrafts. The reason is that when customers face switching costs, banks face a trade-offhigh price ( i.e. offering a low rate on positive balances and charging a high rate on overdraft) increases the profit a firm makes on its existing customer base. On the other of attracting new customers and may also result in the bank losing customers. The firm’s current market share determines how this trade-off is resolved. Firms with low initial market shares charge a low price
Heffernan (2002) argues that price dispersion in UK retail banking markets is related to this type of imperfect competition. This trade-off would be eliminated if the bank could price discriminate between existing and new customers.
customers. Firms that start with a high market share charge high prices (offer low overdraft) in order to increase the profit on existing customers. Notice that this is worthwhile for a bank with high market share even though it means that the firm loses some of its existing customers. market share should be stronger, the lower the elasticity of demand with respect to price, that is the less sensitive consumers are with respect to price. The Salop and Stiglitz (1977) model of implies a negative relationship between market share and period the firm that offers the better deal attracts the most customers. In this model there are two groups of consumers, those with high search costs and those with low search cotiglitz show that a two-price equilibrium may exist where one of two firms charges a high price and the other firm offers a low price. The firm charging the high price is able to attract half of the customers with high search costs but none of the informed customers. The firm charging the low price attracts all other consumers, that is, half of the uninformed and all of the informed customers. For any distribution of informed and uninformed customers this implies that a high market share is associated with a low price in equilibrium. Finally, decreasing unit costs ensure that both types of firms earn the same profit in equilibrium. In the context of the current account market the implication is that banks with larates and/or low overdraft rates. Finally, under standard assumptions of een market share and price. In a perfectly competitive market, it is assumed that there are numerous firms, each being so small that it cannot influence other providers’ homogenous, firms are price-takers and all charge the same price, set such an environment, there sconsequently no link between price and market share.
The low price is equal to the competitive price – ie the price that would prevail in the absence of search costs. It is assumed that entry occurs as long as profits are positive. Thus, in equilibrium, every firm earns zero profit.
In an oligopolistic environment, a firm’s action may some strategic interdependence between the firms in the market. In a Cournot setting, firms may choose their rivals, but the price set by each firm is read off the aggregate, industry demand schedule. If ersion may emerge across different quality levels. But in such a framework, there is no reason to beliand market shareIn order to distinguish between the various models of competition we estimate the following regression: ssion: where: - Rjit is the rate j quoted by bank i, averaged over half-year t. We analyse three different rates: the rate on positive balances on current accounts (j=CA); the pre-authorised overdraft rate (j=OD); and the rate on instant access savings accounts (j=IA). - MSit is the level of bank i’s market share on the current account market measured in half-year t. acteristics measured at a bank level over time: the number of branches per customer in half-year t; the (logarithm of) number of ATMs in half-year t; and two indices (assumed to be constant over time) reflecting the range of transactions a current account customer can perform over the phone asThe implication of each model of competition for the relationship between prices and market shares gives rise to the following set of competing hypotheses as regards the nature of competition in the market for current account.
Various complications may arise from particular assumption on the distribution of income and willingness to pay.
for overdraft rates. ) Perfect competition/ Cournot: b=0 ent account and overdraft rates follow directly from regards the coefficient on market shares in the current account masufficiently closely related. A close relationship might arise either because banks bundle savings or because customers prefer a josuch a link between the two markets one would not epaid on instant access savings accounts and market shares in the current account market, i.e. in this case b=0 for the instant access rate. As in section 3, we comment on regressions that are performed on averages across time. Again, the that the “within group” –i.e. time series - variation for the main variabOLS regressions were also performed and can be fatively similar. Our ma� fo;&#xr CA;&#x rat;s, ;뀀in findings are that bInterestingly, the size of effect is larger for overdraft rates than for CA rates.
34
0510152025Average Level of Market Share - Per CentCA Rate - Per CentCahootFirst-eIntelligent FinanceSmile
HSBC/ Midland
atwestBarclaysLloyds TSB
0510152025Average Level of Market Share - Per CentIA Rate - Per CentHSBC/ Midland
atwestBarclaysLloyds TSBIntelligent FianceSafewayVirgin DirectSmileFirst-e

0510152025Average Level of Market Share - Per CentOD Rate - Per CentHSBC/ Midland
atwestBarclaysLloyds TSBYorkshire BankBOSCo-operativeClydesdale
Charts 9a, 9b and 9c show plots of the market sharesle against rates on current accounts, savings accounts and overdrafts. Chart 9a suggests a negative relationship between current account rates and market shares. This relationship appears somewhat weaker if some ofclear relationship between savings rates and market shares for current accounts. That said, it is striking that among the Big Four banks, the larger the market share the lower the instant access savings rate. Also, in line with the pattern for current account rates, the highest strong negative relationship between market share in the current account market and the instant savings rate offered by banks. By contrast, Chart 9c shows a pronounced positive relationship between market
Chart 9b: Average level of CA market share & average instant access rate
Chart 9c: Average level of CA market share & average overdraft rate
average current account rate
shares and rates charged on overdrafts, with the largest banks chardoes not appear from Chart 9c that ‘direct’ banks exert a strong Table 3 shows the results of regressions of each of the three rates on market shares in the current account market. To account for potential outliers, regressions were estimated both on the whole sample of banks and on a restricted sample, excluding the ‘direct’ banks. Columns 10a and 10b confirm that the larger a bank’s market share, the lower the rate it offers on current account balances. ‘Direct’ banks, which exhibit the lowest market sharntribute to this relationship. But ze of the coefficient, the result stays significant at are excluded. By contrast, insignificant coefficients in Columns 11a market shares in the current account market. Finally, Columns 12arelationship between market shares for current accounts and rates charged on overdrafts. This relationship is significant at the 1% level when e whole sample and stays significant at the 5% level when ‘direct’ banks are excluded. rate and the overdraft rate on the bank’s market share and a number of control variables. As argued above, current accounts may be vertically differentiated products: different banks may offer different levels of quality and service. And banks that offer a superior level of quality may be able to charge their customers a higher price. If quality and market share are correlated this may result in an omitted number of ATMs, it is possible that the market share variable proxies, at least partially, for the address these potential biases we re-estimate the effect of the bank’s market share on the rate it offers on current account balances and overdrafts, including the same proxies for quality that were used in section
The linear relationship appears almost exact when no account is taken of the cluster of banks that includes RBS, BOS, Yorkshire Bank, the Co-operative Bank and Clydesdale Bank. Most of these banks are regional, with much of their branch network concentrated in particular areas – e.g. Scotland for RBS and BOS. The share of these banks in their respective regional market may well be larger than their national market share. Adjusting for this, the linear relationship between market shares and overdraft rates may actually be stronger than it appears from Chart 11c.
The results in Table 4 show a number of regressions quality in addition to the market share as explanatory variables. In most of these regressions the coefficient on market share stays significantly negative, typically at the 5% level. In line with the results in Table 3 we find that the coefficient on the bank’s market share is smaller in absolute value when the whole sample and on the restricted sample the size of the coefficient is comparable with the size of sample and close to –0.03 for the The result for the whole sample suggests that a bank whose market share is larger than that of a comparabpoints (i.e. one standard deviation) would offer a current account rate some 65 basis points lower than the comparable bank. Based on the regressions when market share larger by one standard deviation) the estimate is 21 basis points. rdraft rate on market share in ain most of the regressions the coefficient retains a significantly positive sign, typically at the 1 % level, suggesting that any omitted variable bias arising from disregarding the quality dimension would have been mild. In sum, the regressions of both the current account rate and the overdraft rate on market shares lend support to the hypothesis of imperfect competition related to switching costs. As for the savings rate, the evidence is more mixed. Our preferred interpretation is that the market for instant access savings rates is unrelated to the market for currcross-price elasticity of the bank-level demand for curate on market shares we again may well indicate that consumers are able to unbundle their choice of savings account provider and their choice of current account provider. The switching cost model stipulates that the less reactive to price differentials bank customers are, the e to raise their price, given that the erosion of their customer base due to a price
The results for the instant access savings rate are not shown.
increase will be limited. We can therefore test for an’) The lower the price elasticity of demand (i.e. the absolute value of coefficient e of coefficient section 3), the stronger the relationship between level of market share and price should be. In regressions of the overdraft rate the absolute size of the coefficient on market share is larger than the (22a) is three times bigger than the (absolute) value of the coefficient (0.1) in Column (17a). This n market share and the overdraft rabetween market share and the current account rate. ere it was found that the firm-level elasticity of demand with respect to the overdraft rate was smaller than that ofincrease the price in order to increase the profit they achieve on their existing customer base. For a profit-maximising bank this e elasticity of demand with respect in the customer base resulting from an increase in price. Since demand was found to be less elastic with be more pronounced than thcurrent account rate. These results thus provide fuwith the hypothesis of
Table 3: Rates as a function of the Level of Current Account Market Shares (when Product Characteristics are not accounted for) (10) (11) IA Rate (R(12)
Dependent variables All banks (10a) ‘direct’ banks (10b) All banks (11a) ‘direct’ banks (11b) All banks (12a) ‘direct’ banks (12b)
Level of CA MS -0.105** -0.018**
P-Value (0.023) (0.027)
Level of CA MS -0.04 0.007
P-Value (0.248) (0.840)
Level of CA MS 0.296*** 0.187**
P-Value (0.000) (0.028)
Adjusted R-squared 11.1% 7.25% 0% 0% 19.6% 7.5%
R -0.39 -0.37 -0.17 0.04 0.49 0.37
Obs 21 15 23 17 21 16
Rejected RejectedNot rejected Not rejected RejectedRejected
Shapiro-Wilk Test (normality of residuals) Rejected Rejected Not rejected Rejected Rejected Not rejected
*** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. Robust estimates of standard errors. P-values in parentheses.
Table 4: Current Account Rate (R(13) Current Account Rate(14) Current Account Rate(15) Current Account Rate(16) Current Account Rate(17) Current Account Rate
Dependent variable All banks (13a) ‘direct’ banks (13b) All banks (14a) ‘direct’ banks (14b) All banks (15a) ‘direct’ banks (15b) All banks (16a) ‘direct’ banks (16b) All banks (17a) ‘direct’ banks (17b)
Level of CA Market Share (MS) -0.11*** (0.004) -0.002 (0.717) -0.15** (0.029) -0.03** (0.037) -0.08* (0.090) -0.02* (0.071) -0.12** (0.030) -0.02 (0.151) -0.11** (0.018) -0.02 (0.129)
Number branches/customer -2.25** (0.013) 0.54* (0.072) -2.47*** (0.002) 0.89 (0.124)
Log (Number ATMs) 1.05* (0.062) 0.27 (0.129) 0.66* (0.092) 0.53** (0.011)
Phone index -0.08 (0.501) -0.02 (0.475) -0.18*** (0.000) 0.003 (0.900)
Net Index 0.19* (0.073) 0.002 (0.958)
Adjusted R-squared 41.4% 21.6% 19% 9.7% 9.1% 6.5% 20.7% 1.4% 60.1% 43.2%
Number Observations 21 15 20 15 19 14 19 14 19 14
Rejected Rejected(limit case) Rejected(limit case) Rejected(limit case) rejected Rejected(limit case) Rejected(limit case) Rejected(limit case) RejectedRejected
Shapiro Swilk test (normality of residuals) Not Rejected Rejected RejectedRejected RejectedNot Not Not
*** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. Robust estimates of standard errors. P-values in parentheses.
Table 5: Overdraft Rate (R(18) Overdraft Rate(19) Overdraft Rate(20) Overdraft Rate(21) Overdraft Rate(22) Overdraft Rate
Dependent variable All banks (18a) ‘direct’ banks (18b) All banks (19a) ‘direct’ banks (19b) All banks (20a) ‘direct’ banks (20b) All banks (21a) ‘direct’ banks (21b) All banks (22a) ‘direct’ banks (22b)
Level of CA Market Share (MS) 0.30*** (0.000) 0.04 (0.625) 0.41*** (0.000) 0.27* (0.069) 0.23*** (0.010) 0.10 (0.298) 0.29*** (0.003) 0.061 (0.568) 0.32*** (0.004) 0.20 (0.126)
Number branches/customers 1.02 (0.591) -5.90*** (0.001) 2.17 (0.268) -5.76 (0.194)
Log (Number ATMs) -2.01* (0.072) -1.52 (0.522) -1.74 (0.140) -4.00* (0.084)
Phone index 0.31** (0.048) 0.37** (0.025) 0.40** (0.013) 0.21 (0.619)
Net Index -0.02 (0.921) 0.41* (0.074)
Adjusted R-squared 16.5% 31.6% 24.5% 3.3 % 27.3% 19.7% 14.1% 14.9% 34.8% 44.1%
Number Observations 21 16 21 16 20 15 20 15 20 15
Rejected RejectedRejectedRejected(limit case) RejectedRejectedRejectedRejectedRejectedRejected
Shapiro Swilk test (normality of residuals) Not Not Rejected Not Not Not Not Not
*** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. Robust estimates of standard errors. P-values in parentheses.
CONCLUSION This study provides an analysis of the competitive process in the market for personal current accounts in the UK. Analysing the evolution of banks’ market shares and prices, we adcompetition best fits the data? Using the National Opinion Poll (NOP) survey data, we first describe some stylised facts on market shares and prices in the UK market for personal current accounts. While the level of concentration has remained high in this market, the market appears to become gradually more competitive, with building societies and direct banks making some significant inroads duriTo assess the level of competition in the current account market more formally, we derive the elasticity of bank-level demand with respect to the interest rate offered on positive balances and the rate charged on overdraft. This analysis controls for isolate the effect of price differentials on changes in market share. We find a moderate sensitivity of changes in market share to differences in the currelevel demand with respect to the overdraft rate is cwith a moderate degree of imperfection competition in the market for rsistence of price dispersion, we consider three candidate models of imperfect competition: the dynamic model of switching costs by Kim, Kliger and Vale (2003), the standard oligopoly model with they have different implications as regards the re market shares and oligopoly model there should be no relationship between market share and price. For the UK market for personal current accounts we find a positive relationship between market share and price, which points to the importance of switching costs in this market and is consistent with the
model of competition described in Kim, Kliger and Vale (2003). The basic intuition is that each firm ce increases the profit the firm achieves on its existing customer base, but also implies that the firm is losing more customers. The firm’s current market share determines how this trade-off is resolved. A firm’s incentive to raise its price is more pronounced, the larger is the firm’s current market share. The model also predicts that the relationship between market share and price hip between market share and prirate, for which the elasticity of demand is lowest. facilitate switching. In (2000), the government asked a groupBanking Code. One set of recommendations in the reimplemented specifically focuses on ways to facilitate switching account. The banks have implemented improvements to the logistics of the switching process – i.e. the exchange of information between the switchers’ old and the new banks – to improve the speed and the accuracy of the account transfer. Steps have also been taken to increase consumer awarene002)). Even though it may be too early to assess the impact of these that the study points to the importance of switching costs in the UK market for personal current
banking: what do we know?”, Journal of Financial Intermediationt evolution of the UK banking industry and some implications for financial stability”, Competition Commission (2002), De Bandt O & Davis E P (2000), “Competition, contestability and market structure in European banking Demsetz H (1973), “Industry structure, market rivalry, and public policy”, Department of Trade and Industry (2001), “Consumer knowledge”, Report series No. 1, Consumer much consumers save by shopping around for financial products?”, Gilbert, A R (1984), “Bank Market Structure and Competition- A Survey”, r banking products?”, Heffernan S (1992), “A computati, Banking Services Consumer Codes Kim M, Kliger D and Vale B (2003) ‘Estimating institutions: evidence from survey data”, , Finance and Economics Discussion Series,
Klemperer P (1995): "Competition when consumers have switching costs: an overview with macroeconomics and international trade", Llewellyn D and Drake L (1993): “The economics of bank charges for personal customers”, argains and ripoffs: a model of monopolistically competitive price Review of Economic StudiesSharpe S (1990): “Asymmetric information, bank lending, and implicit contract: a stylised model of customer relationships”, Journal of FinanceThe theory of price, Mac Millan Publishing Company. Stigler G (1961), “The economics of information”, Stiglitz J (1989): “Imperfect Information in the Product Market”, R. Schmalensee and R. Willig Eds., , Vol. 1, Amsterdam: North-Holland.
APPENDIX 1 Category Bank Name Entry date
Barclays
Lloyds TSB 96H1
Abbey National 96H1
Alliance & Leicester
Halifax
Nationwide
Northern Rock
Building societies or ex building societies
First Direct
Intelligent Finance
Smile
‘direct’ Banks
Bank of Scotland 96H1
98H1
All Others Yorkshire Bank
Variable Name Definition
Level of bank i’s market share in the current account market measured in half-year t
Relative change in bank i’s market share in the current account market between end of half-year t-1 and end of half-year t
R:
Level of the interest paid by bank i on current accounts (CA) averaged over half-year t
Level of the interest paid by bank i on instant access savings accounts (IA) averaged over half-year t
Level of the interest paid by bank i on overdrafts (OD) in half-year t
RD:
Absolute difference (in percentage points) between the interest paid by bank i on current accounts and the average of the CA rates paid by its competitors in half-year t
Absolute difference between the interest paid by bank i on instant access savings accounts and the average of the IA rates paid by its competitors in half-year t
Absolute difference between the interest charged by bank i on overdrafts d by its competitors in half-year t
Q:
- Nber branches/Customer Number of bank i’s branches divided by the number of its customers, in half-year t
- Log (Nber ATMs) Logarithm of the number of bank i’s ATMs in half-year t
- Phone index Index that can take any integer value between 0 and 13 to reflect the number of operations relating to standing orders, direct debits, bill payments, transfers and ordering a customer of bank i can perform over the phone (calculated for December 2002)
Index that can take any integer value between 0 and 11 to reflect the number of operations relating to standing orders, direct debits, bill payments, transfers and ordering a customer of bank I can perform over the Internet (calculated for December 2002)

Table C: Descriptive Statistic of the panel data set Variable Mean Min Max Overall ‘Between’
2.22 -29.5067.7712.4318.29 10.005.60
6.11 0.0423.856.746.70 0.521.10
0.52 0.104.750.801.24 0.451.54
2.24 0.205.021.201.20 0.870.54
15.22 8.9021.553.663.71 1.020.24
-0.11 -1.354.420.851.15 0.56-7.73
-0.18 -2.913.581.021.40 0.47-5.67
-0.06 -6.956.433.883.86 1.07-64.67
Nber branches per Customer0.68 02.110.410.42 0.16
3.04 1.083.650.540.68 0.070.18
10.95 0134.163.98 00.38
7.98 0113.593.57 00.45
The standard deviation of each variable is decomposed into a ‘between group’ – i.e. cross-sectional – component, and a ‘within group’ – i.e. time series – component. The coefficient of variation is defined by the ratio of the standard deviation to the mean.
APPENDIX 2 – Pooled OLS estimations Table 6: Pooled OLS estimations (with quality variables) on the changes in market share and the 3 rate differentials Dependent variable
All obs. All obs. All obs. All obs.
5.33*** (0.003) 4.54***(0.002) 5.25**(0.011) 4.05**(0.022) 4.82**(0.019) 3.90**(0.028) 5.72*** (0.001) 4.39*** (0.001)
0.65 (0.485) 0.45 (0.596) 0.74 (0.369) 0.28 (0.733) 1.14 (0.385) -0.14 (0.852) 0.22 (0.848) -0.15 (0.901)
-0.50*(0.056) -0.63**(0.035) -0.46**(0.012) -0.48**(0.023) -0.50***(0.007) -0.58***(0.002) -0.40 (0.171) -0.66** (0.031)
branch/cust -4.94*(0.051) -4.82**(0.015) -6.70 (0.113) -5.86 (0.236)
2.74**(0.025) 2.01*(0.056) 2.05 (0.350) 0.25 (0.917)
0.11 (0.596) 0.27 (0.128) -0.42 (0.218) -0.08 (0.783)
Obs. 175 161 175 161 167 153 167 153
Adj. R-squared 20.6% 11.1% 19.7% 9.7% 19.6% 10.4% 22% 11.3%
RejectedRejectedRejectedRejectedRejectedRejectedRejected Rejected
Residuals Normality Rejected Rejected
All results are corrected with the CLUSTER option. The CLUSTER option in the ecA relaxes the assumption that observations are independent within groups, but still maintains that the observations are independent across groups. Robust estimates of standard errors are obtained. *** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. P-values in parentheses. The observations relating to ‘direct’ banks are excluded from the sample.
Table 7: Pooled OLS estimations (with quality variables) on the CA rate and the level of market share Dependent variable
All obs. banks All obs. banks All obs. banks All obs. banks
-0.05**(0.031) 0.001 (0.878) -0.06**(0.019) -0.03**(0.033) -0.03*(0.057) -0.01*(0.056) -0.06** (0.023) -0.02(0.134)
-0.06 (0.577) 0.08 (0.264) -0.6 (0.637) 0.13 (0.141) -0.07 (0.611) 0.14 (0.154) 0.05 (0.630) 0.09 (0.170)
-0.53 (0.330) 0.58* (0.066) -0.82 (0.225) 0.77** (0.041)
0.44*(0.053) 0.15(0.282) 0.47*(0.064) 0.43*** (0.007)
Phone_index -0.03 (0.478) -0.02 (0.325) -0.10*(0.053) -0.003 (0.873)
Obs. 203 177 201 177 192 168 192 168
Adj. R-squared 10.7% 17.3% 10% 7% 7.3% 9.9% 20.8% 23.3%
rejected Rejected(limit case) Rejected(limit case)rejected Rejected(limit case) rejected RejectedRejected
Residuals Normality Rejected Rejected Rejected Rejected Rejected Rejected
Base rate is the bank of England base rate. All results are corrected with the CLUSTER option. The CLUSTER option in STATA relaxes the assumption that observations are independent within groups, but still maintains that the observations are independent across grestimates of standard errors are obtained. *** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. . P-values in parentheses. Table 8: Pooled OLS estimations (with quality variables) on the IA rate and the level of market share Dependent variable
All obs. banks All obs. banks All obs. banks All obs. banks
-0.04 (0.110) -0.02 (0.639) -0.05** (0.043) -0.02 (0.556) -0.01 (0.483) 0.004 (0.812) -0.05** (0.022) -0.01 (0.755)
0.88*** (0.000)0.99*** (0.000) 0.88*** (0.000) 0.97***(0.000) 0.88*** (0.000) 0.99*** (0.000) 0.97***(0.000) 0.98***(0.000)
-1.63*** (0.003) -1.20 (0.112) -0.47 (0.297) 0.38 (0.297)
1.14***(0.001) 0.84**(0.056) 0.78**(0.024) 0.39 (0.331)
Phone_index 0.12***(0.006) 0.10*** (0.014) 0.05 (0.121) 0.11*** (0.007)
Obs. 222 195 207 187 198 178 198 178
Adj. R-squared 36.6% 37.9% 39.8% 47% 44% 59.6% 51.6% 61.2%
RejectedRejected Rejected Rejected RejectedRejectedRejectedRejected
Residuals Normality Not Not Rejected Rejected
Base rate is the bank of England base rate. All results are corrected with the CLUSTER option. The CLUSTER option in STATA relaxes the assumption that observations are independent within groups, but still maintains that the observations are independent across grestimates of standard errors are obtained. *** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. . P-values in parentheses.
Table 9: Pooled OLS estimations (with quality variables) on the OD rate and the level of market share Dependent variable
All obs. banks All obs. banks All obs. banks All obs. banks
0.21**(0.015) 0.06 (0.442) 0.33**(0.001) 0.24**(0.047) 0.17**(0.045) 0.10 (0.265) 0.32** (0.002) 0.22**(0.013)
0.38 (0.184) 0.44**(0.048) 0.21 (0.305) 0.05 (0.815) 0.33 (0.111) 0.15 (0.379) 0.13 (0.619) 0.24 (0.274)
-0.89 (0.638) -4.52***(0.003) 0.60 (0.783) -3.65 (0.134)
-1.75 (0.145) -1.06 (0.531) -2.81*(0.064) -3.83**(0.017)
Phone_index 0.29*(0.054) 0.36**(0.023) 0.43**(0.037) 0.31 (0.260)
Obs. 213 189 213 189 204 180 204 180
Adj. R-squared 16.2% 30% 20.2% 11.7% 23.8% 26% 35.4% 46%
RejectedRejectedRejectedrejected RejectedRejectedRejectedRejected
Residuals Normality Not Rejected RejectedRejected RejectedRejected Rejected
Base rate is the bank of England base rate. All results are corrected with the CLUSTER option. The CLUSTER option in STATA relaxes the assumption that observations are independent within groups, but still maintains that the observations are independent across grestimates of standard errors are obtained. *** denotes statistical significance at one percent level, ** at five percent level, * at ten percent level. . P-values in parentheses.

Documentos PDF asociados:

THE ECONOMICS OF RETAIL BANKING - An empirical analysis of ...
Practical Meta-Analysis Analysis Exercise using SPSS
Daniel Stern: Microanalysis and the Empirical Infant ...
Central Bank | For All of Your Banking Needs
RETAIL INSTALLATION AND MAINTENANCE - usg.com
POLYTECH AIR COMPUTERS PC ARTS PC RETAIL ... - gabaktech.com
Libro Credit Union - Banking & Financial Services ...
Economics in One Lesson - HACER
Economics of Education Review - users.nber.org
to accompany Fundamental Methods of Mathematical Economics
WAGE EQUATIONS AND THE REGIONAL ECONOMICS IN GUATEMALA ...
11th World Congress in Health Economics De Gustibus ...
PAUL SAMUELSON AND FINANCIAL ECONOMICS Robert C Merton ...
STERN REVIEW: The Economics of Climate Change Executive ...
Limited Supermarket Analysis
Motion Analysis and Biomechanics
for Numerical Analysis - Cengage
CONTEMPORARY INSTRUMENTAL ANALYSIS BY RUBINSON AND PDF
An Introduction to Instrumental Methods of Analysis
Dyadic Data Analysis - Purdue University
Data Analysis With Mplus Methodology In The Social ...
Principal component analysis with linear algebra
A Cost‐Benefit Analysis of Waste Incineration with ...
education policy analysis archives - ERIC
“Mercosul, Three Steps from a Customs Union. Analysis of ...
The conjunto piano in 1940s Cuba: An analysis of the ...
Digital Forensic Analysis using HELIX and Autopsy
Internet-Scale analysis of AWS Cognito Security
Intellectual property and access to medicines: an analysis ...
Matrix Analysis, CAAM 335, Spring 2012
XPS Basic Data Analysis - Stanford University
CSi Analysis Reference Manual - Computers and Structures
Principles of Instrumental Analysis, 4th Edition (Skoog, D ...
SOLUTION FOR PROBLEMS IN MATHEMATICAL ANALYSIS DEMIDOVICH PDF
Contemporary Strategy Analysis: Concepts, Techniques ...
Microeconomic Analysis, 1992, Hal R. Varian, 0393960269 ...
Introductory Mathematical Analysis Haeussler Solutions
Microelectronic Circuits: Analysis & Design by Muhammad H ...
Trait Analysis the Tarahumara Indians and Mestizos the ...
He Andado Muchos Caminos Analysis - bxscience.edu