VIEWS: 6 PAGES: 20 CATEGORY: Business POSTED ON: 4/13/2011 Public Domain
COMPETITION AND CONTESTABILITY IN NEW ZEALAND’S BANKING SYSTEM Paper presented at the 14th Australasian Finance and Banking Conference, Sydney, December 2001 Robert Smith David Tripe1 Centre for Banking Studies Massey University Abstract: This paper explores use of the Rosse-Panzar statistic as a basis for assessment of competitive conditions in the New Zealand banking market. From pooled regressions for the whole period analysed, 1996 to 1999, it was found that total bank revenues appear to have been earned under conditions of monopolistic competition. Cross-sectional competitive conditions regression analysis suggests that for 1996 total bank revenues were earned under conditions of monopolistic competition, while, for 1997, the analysis implies that monopoly or conjectural variation oligopoly conditions were present. For 1998 and 1999, indications are that the New Zealand banking market behaved like a natural monopoly in a perfectly contestable market, perfect competition, or sales-maximising firms subject to a break-even constraint. 1 Contact author. Address: Centre for Banking Studies, Massey University, Private Bag 11-222, Palmerston North, New Zealand. Phone +646 3595799 ext 2337, fax +646 3505651, E-mail D.W.Tripe@massey.ac.nz The level of competition in the commercial banking sector has always been a subject of strong interest. In both New Zealand and Australia, bank customers are concerned that oligopolistic conditions apply within the banking sector, causing banks to provide inadequate services and extract excessive profits. This was part of the background to the Wallis Inquiry in Australia, and there was debate in both countries as to whether further mergers between banks should be permitted, and under what conditions. Questions were raised as to what might be an acceptable level of competition in the New Zealand banking sector [Tripe (1997), (1999)]. It is also arguable that concerns about potential monopoly have helped to provide a political constituency for the new bank that is being established by New Zealand Post, commonly referred to as the Kiwi Bank.2 This paper explores a possible approach to measurement of competitive conditions in New Zealand banking, using the Rosse-Panzar statistic, figures for which we develop for the New Zealand market. Some theoretical perspectives There are a variety of approaches to measurement of competition such as the number of competitors and market share of the largest participants, but these approaches are relatively crude in that they don’t indicate whether these competitors are able to exercise market power, which is, after all the reason for trying to measure competitive conditions in the first place. As Shaffer (1994) has put it: “Fewer firms in a market (that is, a concentrated structure) will generally lead to less competitive conduct (in terms of higher prices and reduced output levels) and less competitive performance (higher 2 For further discussion of this, refer to Tripe (2001). 2 ratios of price to cost and higher profits at the expense of lower consumer welfare).” A more systematic attempt to quantify the potential for firms to exercise power in a market can be expressed through the Herfindahl-Hirschman Index (HHI), which takes account of the relative sizes of all market participants. Values of this index can be expressed in a range from either 0 to 1 or 0 to 10,000, as follows: HHI = Σ(Si/S)2 (1) where Si is the size of the ith firm and S the size of the total market.3 Si/S thus represents firm i’s market share. An HHI of zero would thus indicate that there were a large number of firms in the market, none of which had a market share greater than 0.7% (based on whole numbers and the 0 to 10,000 scale), while an HHI of 10,000 would indicate that there was only one participant in the market. The HHI is used as the basis for merger approvals in a number of environments, most notably the United States banking market, where a proposed merger will attract the attention of the competition authorities if the HHI in a particular market exceeds 1800.4 The HHI has an advantage in terms of being reasonably straightforward to calculate, with suitable data not particularly difficult to obtain. It is inherently a static measure, describing the situation at one point of time, and tells us nothing about the actuality of any collusive behaviour by market participants.5 3 If market shares are expressed as fractions, the HHI will be in the range 0 to 1, if as percentages, it will be in the range zero to 10,000. 4 For a further, more detailed exposition, see Rhoades (1993). 5 The HHI thus cannot give any insights into the dynamic framework for monopolistic competition discussed by Schumpeter (1943). 3 Orthodox microeconomic analysis would tell us that the real factors influencing a firm’s ability to exercise monopoly power lie in the slope and elasticity of the demand curves for its products. What we really want to know is whether, despite observed measures of competitive conditions, a banking firm is able to raise its prices without suffering a severe decline in its revenues. However, because they are often fearful of the potential losses of market position that might arise from doing so, banks are usually reluctant to engage in such experimentation to establish the exact shapes of the demand curves for their products, especially as they are likely to be influenced by competitor reactions. It is therefore difficult to assess the shape of the relative demand curves, and even more difficult for outside observers trying to ascertain the extent of effective competition. One way in which competitive conditions might be reflected would be in frequency of price changes: there is a hypothesis that market power would lead to less frequent and slower adjustments in favour of the consumer, but more frequent and faster adjustments in the bank’s favour [Shaffer (1994), p 8]. Hannan and Berger (1991) and Neumark and Sharp (1992) have found some evidence for this in consumer deposit pricing. This area of research could be worth pursuing for comparing some aspects of bank behaviour in the New Zealand and Australian markets, but the authors have not as yet found any record of this having been attempted. Another approach to the conundrum of competition in banking markets and its effect is through the theory of contestable markets. This argues that the actual number of competitors and their respective market shares are not important as long as new entrants are able to enter the market freely. The threat of new entrants undercutting price and displacing an incumbent’s market share induces optimal pricing across likely market structures. The free movement of resources in and out of markets thus provides a check on uncompetitive behaviour [Oliver (1996)]. 4 The Rosse-Panzar Statistic In response to previous researchers’ failure to find significant connection between concentration and monopoly power, J N Rosse and J C Panzar developed a non-structural estimation technique to assess the level of monopoly power being exercised in a particular market. They applied Shepherd’s Lemma to a firm’s profit maximising first order conditions to show that the sum of elasticities of gross revenue with respect to each input price will be negative in the case of collusive equilibrium, and less than one for monopolistic competition. A detailed review is provided in Rosse and Panzar (1987), and it is beyond the scope of this paper to provide a detailed account of the analysis leading to the relative propositions. We will, however, note one key assumption. For the positive values of the Rosse-Panzar statistic to be reliable (and capable of being interpreted in accordance with the respective models of monopolistic and perfect competition), the models depend on the firms in question being in long-run equilibrium, and having stable cost and revenue functions. A key limitation of the Rosse-Panzar test is that it can give misleading results under a number of circumstances, such as when the banks in the sample have not completely adjusted to market conditions. The bias in this instance is towards the spurious appearance of market power. “This anticompetitive bias means that, in the absence of a reliable test for market disequilibrium, the Rosse-Panzar test cannot be used to rule out competitive pricing, as some studies have claimed” [Shaffer (1994)]. But generally when the test indicates the market is competitive, we can be relatively sure that monopoly power is not being exercised. 5 Another weakness of the Rosse-Panzar test is that it cannot distinguish between competitive pricing and simple cost-plus pricing. Since cost-plus pricing is not specifically associated with a particular level of market power, the implications for interpreting the Rosse-Panzar statistic are not clear once this limitation has been taken into account. The Rosse-Panzar methodology has been applied to assess competitive conditions and contestability in banking by a number of researchers. Shaffer (1982) used the methodology to examine the competitive conditions for banks in New York. Nathan and Neave (1989) used the same statistic to develop measures of competition and contestability for the Canadian banking system. Lloyd-Williams et al (1991) tested for evidence of contestability on a sample of 72 Japanese commercial banks for 1986 to 1988. Shaffer (1993) supported and extended Nathan and Neave’s findings over a longer period of time using a different method. Vesala (1995) tested for competition in the Finnish banking market and the characterisation of banks’ pricing behaviour using the Rosse- Panzar methodology. Molyneux et al (1994) utilised the Rosse-Panzar statistic to assess competitive conditions in five major European Community banking markets. Hondroyannis et al (1999) applied this technique to assess competitive conditions in the Greek banking system. We have no indication that the technique has ever been used in the New Zealand or Australian banking markets. Methodology The relevant elasticities which are summed to produce the Rosse-Panzar statistic are derived as coefficients in the following multiple regression, which is effectively a log transformation of a production function.6 6 The specification is the same as that used by Nathan & Neave (1989), which is considered reasonable given the comparative similarity in banking market structure between Canada and New Zealand (with both markets dominated by a small number of full service banks). Other researchers have used different 6 lnTR = β0 + β1 (lnPF) + β2 (lnPKB) + β3 (lnPL) + β 4 (lnAST) + β 5 (lnBR) + β6 D5 (2) where: ln = natural logarithm TR = total revenue PF = ratio of interest expense to total deposits (unit price of funds) PKB = ratio of annual expense on premises, furniture, fixtures and equipment to the number of branches (unit price of capital) PL = annual salary and benefits per full-time equivalent employee (unit price of labour) AST = total bank assets BR = number of branches to the total number of branches in the system D5 = 1 for the 5 largest banks, 0 for the others From this we develop a statistic HComp = β1 + β 2 + β3 (3) We also test for equilibrium, using a statistic HEqm , based on the following regression equation: lnROA = α0 + α1 (lnPF) + α2 (lnPKB) + α3 (lnPL) + α 4 (lnAST) + α5 (lnBR) + α 6 D5 (4) Where ROA = return on assets Relative to this HEqm = α1 + α2 + α3 (5) specifications. For example, in their study of the Greek market, Hondroyiannis et al (1999) looked at provisions, but these have not been important (or different between banks) in New Zealand over the period studied. 7 We can interpret the values obtained in accordance with the principles outlined in Table 1, based on a table provided by Hondroyiannis et al (1999). Table 1: Interpreting the Rosse-Panzar H Statistics Competitive environment test, HComp Equilibrium test, HEqm H < or = 0 Monopoly or conjectural variation H<0 Disequilibrium short-run oligopoly 0 < H< 1 Monopolistic competition H=0 Equilibrium H=1 Perfect competition or natural monopoly in a perfectly contestable market or a sales maximising firm subject to a break even constraint Regressions were run first to test for equilibrium and competitive conditions in the New Zealand banking sector using pooled data over the whole period 1996 to 1999. The pooling procedure improved the parameter estimates by increasing the sample size, although pooling assumes that parameter values are constant over time, an assumption which may be unrealistic. A variety of approaches were tested within SAS, using the MIXED procedure, and, for the pooled equilibrium test using lnROA as the dependent variable, the best results were obtained using the UN(1) correlation structure model (where it is assumed that there is no correlation between years and heterogeneous residual variance each year). Regressions were then run for equilibrium and competitive conditions in New Zealand banking using the cross-sectional data for the individual years 1996, 1997, 1998 and 1999. Even though the individual year sample sizes were relatively small, the separate samples were retained as they permitted assessment of differences at different times, which might not be observed with the pooled data. 8 Data In our efforts to calculate the Rosse-Panzar statistics, we have used data for New Zealand banks for four years 1996 to 1999 inclusive. This has allowed us to extract figures for return on assets, total revenue, interest expense, premises expense, staff expense, total deposits and total assets. Data for the number of branches and number of employees were derived from the KPMG Financial Institutions Performance surveys for New Zealand. Banks in New Zealand have a variety of different balance dates, with one bank currently reporting in March, two in June, four in September, and the rest in December. For part of the period reviewed there was also a bank with a February balance date. Because there had been significant variations in general levels of interest rates over the period reviewed, it was thought most appropriate to standardise to a common December financial year (with the relevant annual results adjusted using data from the quarterly disclosure statements required to be provided by all New Zealand banks).7 In some cases, banks have not provided breakdowns of premises-related and staff expenses (and where banks did provide a breakdown, these were only available at half-year and year end). We have therefore had to try and approximate figures based on the breakdown of expenses in other time periods (for those banks with March and September balance dates), or based on the breakdown of expenses of other banks whose business was considered comparable. Some other assumptions were also necessary to make the data useable. The New Zealand banks do not report deposits consistently, and we have 7 It was not considered practical to satisfactorily adjust the data for the bank with the February balance date, and data for this bank were therefore not adjusted. 9 therefore used figures for total interest-bearing liabilities (which are required to be disclosed) in calculating the unit price of funds. Regardless of which variable was in the denominator, banks in a stronger position, and thus better able to exercise monopoly power in the acquisition of deposits, are likely to exhibit significantly lower costs of funds. A problem with both of the regressions is with calculated values, specifically lnROA, being undefined as a result of profit being negative. We therefore follow a common approach used for log transformations in adding a positive constant to ROA for every observation (the same constant for each one), large enough to make the smallest sum positive, before taking the natural logarithm.8 This approach changes the interpretation of the coefficients somewhat, but does not bias them, so that the test for equilibrium for lnROA remains valid. It is not believed that these assumptions will have had significant effects on the results obtained. Results and Discussion The equilibrium test and competitive conditions tests for the pooled data regressions are reported in Tables 2 and 3. The estimation of lnROA (to test for equilibrium conditions) yields an R- squared of 0.237, which suggests that the explanatory cost variables account for little of the variation in lnROA. The regression coefficients for the unit prices of fixed assets, labour and funds are all positive, with the first two significant at the 5% level. A test for co-variance provided no evidence of multicollinearity among the explanatory variables. 8 As used, for example, by Berger & Mester (1997). 10 Table 2: Pooled Equilibrium test for New Zealand Banks, 1996-1999. 1996-1999 t Constant -3.391 -7.990* ln PF 0.116 3.230* ln PKB 0.039 2.290* ln PL 0.061 1.720 ln AST -0.007 -0.340 ln BR 0.032 1.540 D5 -0.026 -0.310 H 0.216 4.390b R2 0.237 Adjusted R2 0.223 Notes: Dependent variable is ln ROA, t is the t-statistic for the parameter estimates, * indicates that coefficient estimates are significant at the 5% level, b is the t-statistic for the hypothesis H = 0. Testing of the hypothesis that HEqm = 0 gave a t-statistic of 4.39, allowing us to reject the null hypothesis at the 1% level. We were therefore unable to confirm that the data are in long-run equilibrium for the pooled regressions over the period 1996-1999, and were therefore obliged to treat the results of the competitive conditions test with caution. In the lnTR regression equation (the competitive conditions test) the R- squared was 0.057, indicating that the model explains only a very small proportion of the variability in lnTR. The regression coefficients for the unit prices of funds, capital and labour had mixed signs, with the first two significant at the 5% level. The coefficient of lnAST (which tests for scale economies) was positive and statistically significant at the 5% level, suggesting that larger banks may have higher revenues per dollar of assets. The coefficient of lnBR was positive and significant at the 5% level, which 11 indicates that banks with greater numbers of branches generate higher revenues. Table 3: Pooled Competitive Conditions test for New Zealand Banks, 1996- 1999. 1996-1999 t Constant 2.158 3.020* ln PF 0.661 9.060* ln PKB 0.173 5.920* ln PL -0.076 -1.260 ln AST 0.841 23.130* ln BR 0.162 5.010* D5 -0.081 -0.770 H 0.758 -2.615a 8.190b R2 0.057 Adjusted R2 0.039 - Notes: Dependent variable is ln TR, t is the t statistic for the parameter estimates, * indicates that coefficient estimates are significant at the 5% level, a is the t-statistic for the hypothesis H = 1, b is the t-statistic for the hypothesis H = 0. The D5 dummy variable which distinguished New Zealand’s largest five banks was not significant at the 5% level, indicating no additional explanatory power. The five largest banks thus do not exhibit any significant revenue effects beyond those explained by assets and branches.9 The t-statistic for testing the hypothesis HComp = 0 for the pooled competitive conditions test is 8.19 (p-value < 0.0001), indicating that we can reject the null 9 Regressions were re-run with the D5 variable omitted, but this had no material effect on the outcome of the analysis (apart from slightly increasing the significance of the majority of the coefficients). Results have not therefore been reported. 12 hypothesis. The t-test for testing the hypothesis HComp = 1 is –2.615 (p-value 0.0132), allowing rejection of that hypothesis. HComp thus lies between 0 and 1, with an observed value of 0.758, which disproves monopoly or conjectural variation short-run oligopoly, natural monopoly in a perfectly contestable market, or a sales maximising firm subject to a break-even constraint. Total bank revenues thus appear to be earned under conditions of monopolistic competition. Results from the cross-sectional equilibrium tests are reported in Tables 4 and 5. For three of the four years studied, the regressions to determine the existence of equilibrium conditions generated R-squared values between 0.21 and 0.33, while for 1999, the R-squared value was 0.743. the coefficients for unit prices of funds, fixed assets and capital were not significant (except for the price of labour in 1999) and had mixed signs. Table 4: Cross-Sectional Equilibrium Test for New Zealand banks 1996-1999 1996 t 1997 t 1998 t 1999 t Constant -2.258 -3.180* -7.247 -2.117* -3.292 -3.367* -2.896 -5.961* ln PF -0.078 -1.053 -0.445 -1.504 -0.061 -0.349 0.085 1.313 ln PKB 0.009 0.303 -0.040 -0.303 0.027 0.591 0.066 1.717 ln PL -0.003 -0.043 0.230 0.582 -0.031 -0.399 0.106 2.425* ln AST -0.038 -1.007 0.061 0.493 0.020 0.373 -0.055 -2.173 ln BR 0.015 0.474 -0.011 -0.072 0.030 0.353 0.110 2.908* D5 0.040 0.423 0.053 0.120 -0.167 -0.853 -0.177 -1.348 H -0.071 -0.766b -0.282 -0.677b -0.065 -0.254b 0.257 3.248b R2 0.210 0.247 0.327 0.743 Adjusted -0.468 -0.204 -0.177 0.550 R2 Notes: Dependent variable is ln ROA, t is the t-statistic for the parameter estimates, * indicates that coefficient estimates are significant at the 5% level, b is the t-statistic for the hypothesis H = 0. 13 For 1996, 1997 and 1998, t-values obtained were not high enough to allow us to reject the null hypothesis that H - Eqm = 0, whereas for 1999, the t value obtained was 3.248 (equivalent to a p-value of 0.006), which forces us to reject the null hypothesis. We can thus conclude that the market was in long run equilibrium for 1996 to 1998, but not for 1999 (although the value of HEqm was positive). There is no obvious reason as to why the market should have been in equilibrium in the three earlier years but not in 1999: if anything, one would have thought that the market was more stable in 1999 than in earlier years, when we had seen mergers and strenuous cost-cutting. The results obtained, however, force us to take care in interpreting the competitive conditions test for 1999. Table 5: Cross-Sectional Competitive Conditions Test for New Zealand Banks 1996-1999. 1996 t 1997 t 1998 t 1999 t Constant 2.126 2.546* -6.612 -1.657 1.498 0.718 4.413 3.062* ln PF 0.667 7.655* -0.218 -0.631 0.965 2.580* 0.893 4.669* ln PKB 0.184 5.031* -0.017 -0.111 0.096 0.993 0.154 1.356 ln PL -0.094 -1.343 0.388 0.953 0.097 0.587 -0.018 -0.142 ln AST 0.843 18.765* 1.042 7.224* 0.876 7.729* -0.756 10.156* ln BR 0.159 4.327* -0.275 1.549 0.187 1.045 0.187 1.673 D5 -0.069 -0.626 -0.914 0.103 -0.248 -0.594 0.124 0.319 H 0.757 -2.229a 0.154 -1.743a 1.157 -0.287a 1.028 0.121a 6.929b 0.317b 2.111b 4.388b R2 0.999 0.980 0.995 0.994 Adjusted 0.998 0.968 0.992 0.990 R2 -statistic for the parameter Notes: Dependent variable is ln TR, t is the t estimates, * indicates that coefficient estimates are significant at the 5% level, a is the t-statistic for the hypothesis H = 1, b is the t-statistic for the hypothesis H = 0. 14 In the cross-sectional competitive conditions test with lnTR as the dependent variable, the regression coefficients for the unit prices of funds, fixed assets and labour are of mixed signs, a result similar to that found by Molyneux et al (1994). The regression coefficient of lnPF was significant at the 5% level in all years except 1997, while the coefficient of lnPKB was significant only in 1996 and the coefficient of lnPL was not significant in any of the years reviewed. The lnAST coefficient was positive and statistically significant at the 5% level in all years, suggesting that larger banks have higher revenue per dollar of assets. The coefficient of the variable recording size effects in terms of branches, lnBR, is statistically significant in only one year, 1996, while the dummy variable, D5, which distinguished New Zealand’s five largest banks is not significant at the 5% level in any year. Table 5 has showed us the calculated values for HComp for the four years reviewed. Table 6 shows us the interpretation of the t-statistics relating to these values, in terms of their statistical significance. We can thus say that the existence of a perfectly competitive banking market has been disproven for years 1996 and 1997, and that monopoly conditions have been shown not to exist in 1996, 1998 and 1999. These findings are not inconsistent with the estimated values obtained for HComp, which has a relatively low value for 1997, raising concerns as to the extent of competition in the banking market in that year, with an inference of monopoly or conjectural variation oligopoly. Table 6: Hypothesis testing for HComp for cross-sectional test. Year 1996 1997 1998 1999 t-statistic for HComp = 1 -2.229 -1.743 0.287 0.121 p-value 0.044 0.1 0.778 0.905 t-statistic for HComp = 0 6.929 0.317 2.111 4.388 p-value .00001 0.755 0.05 0.0006 15 For 1996, the regression analysis yields a value between 0 and 1 for HComp, which indicates that total bank revenues were earned under conditions of monopolistic competition. Values of HComp in this range seem to be relatively frequent in international studies, and in a personal communication, Sherrill Shaffer has indicated that monopolistic competition may not be the only explanation for values in this range. He has suggested that, if the theory were more completely worked out, one might find these values might be accounted for by use of limit pricing to deter entry, expense preference behaviour, or some other explanation. For the final two years of the study period, 1998 and 1999, HComp is significantly higher than zero but not significantly different from unity. This indicates that the New Zealand market appears like a natural monopoly in a perfectly contestable market, or to be characterised by perfect competition or sales-maximising firms subject to break-even constraints. For 1999, we have seen that the sample cannot be shown to be in long-run equilibrium, but as the value of HEqm was significantly positive, we can be satisfied that we do not have monopoly or conjectural oligopoly behaviour. Concluding discussion We have ruled out monopoly power for the New Zealand banking market with the exception of the cross-sectional analysis undertaken for 1997. There are no obvious reasons as to why the year 1997 should have been exceptional. The model would appear to require stable cost and revenue functions, and there must be some doubt as to whether these would have existed during any of the period analysed, having regard to the severe cost-cutting undertaken by the banks throughout the period. 16 It is therefore interesting to compare the results obtained from use of the Rosse-Panzar statistic with the trends in the HHI during the period reviewed, and which are shown for total bank assets in Figure 1. Figure 1: HHI trend for total assets: New Zealand banks. 1750 1700 1650 1600 1550 1500 1450 1400 30/6/96 30/9/96 31/3/97 30/6/97 30/9/97 31/3/98 30/6/98 30/9/98 31/3/99 30/6/99 30/9/99 31/3/00 30/6/00 30/9/00 31/12/96 31/12/97 31/12/98 31/12/99 31/12/00 The sharp increase in the September quarter of 1998 reflects the acquisition of Countrywide Bank by the National Bank of New Zealand, but this change was not reflected in the trend for the Rosse-Panzar statistic. It is interesting that the trend of the HHI statistic provides no support for the apparent finding of uncompetitive conditions in 1997: perhaps this result is a consequence of statistical noise or some other weakness in the methodology. The problems may derive from weaknesses in the Rosse-Panzar methodology itself. The technique has an anti-competitive bias, meaning that when competitiveness is indicated we can be reasonably sure that monopoly power is not present. The test cannot, however distinguish between competitive 17 pricing and simple cost-plus pricing, the implications of which weakness have yet to be ascertained. There is an identifiable need to carry out more work on the Rosse-Panzar methodology, expanding on suggestions that values of HComp between zero and one might be explained by limit pricing to deter entry, expense preference behaviour, or firms having some objective other than profit. Reflecting methodological issues identified by Shaffer (1982), it may be worth omitting figures for newly-established banks in case their figures distort the data for the rest of the banking sector. This research could also be extended using further time periods, and by comparable studies of the Australian market. The approach followed may well have considerable potential as a tool for monitoring competitive conditions, particularly in the face of future mergers in the New Zealand and Australian banking sectors. 18 References: Berger, A. N. & Mester, L. J. (1997). Inside the black box: what explains the differences in the efficiency of financial institutions? Journal of Banking and Finance. 21(7). 897-949. Hannan, T. H. & Berger, A. N. (1991, September). The rigidity of prices: evidence from the banking industry. American Economic Review. 81. 938- 945. Hondroyiannis, G.; Lolos, S. & Papapetrou, E. (1999). Assessing competitive conditions in the Greek banking system. Journal of International Financial Markets, Institutions and Money. 9. 377-391. Lloyd-Williams, D. M.; Molyneux, P. & Thornton, J. (1991). Competition and contestability in the Japanese commercial banking market. Institute of European Finance, Research Papers in Banking and Finance. No 16. IEF: Bangor. Molyneux, P.; Lloyd-Williams, D. M. & Thornton, J. (1994). Competitive conditions in European banking. Journal of Banking and Finance. 18. 445- 459. Nathan, A. & Neave, E. (1989, August). Competition and contestability in Canada’s financial system: empirical results. Canadian Journal of Economics. 22(3). 576-594. Neumark, K. & Sharp, S. A. (1992). Market structure and the nature of price rigidity: evidence from the market for consumer deposits. Quarterly Journal of Economics. 107. 657-680. Oliver, R. (1996, December). The contestability of the New Zealand banking market. New Zealand Banker. 35-36. Panzar, J. C. & Rosse, J. N. (1987). Testing for “Monopoly” Equilibrium. The Journal of Industrial Economics. XXXV (4). 443-456. Rhoades, S. A. (1993, March). The Herfindahl-Hirschman Index. Federal Reserve Bulletin. 188-189. Schumpeter, J. A. (1943). Capitalism, Socialism and Democracy. London: Allen & Unwin. Shaffer, S. (1982). A non-structural test for competition in financial markets. Proceedings of a Conference on Bank Structure and Competition, April 12-14, 1982. Federal Reserve Bank of Chicago. 225-243. 19 Shaffer, S. (1993). Market conduct and excess capacity in banking: a cross- country comparison. Working Paper 93-28 (Federal Reserve Bank of Philadelphia). Shaffer, S, (1994, March/April). Bank competition in concentrated markets. Federal Reserve Bank of Philadelphia Business Review. 3-16. Tripe, D. (1997, February). How concentrated/competitive is the banking sector? New Zealand and Australia. Australian Banker. 18-20. Tripe, D. (1999). Mergers on the cards: should we have more bank mergers in New Zealand? Wellington: The Finance Sector Union. Tripe, D. (2001, April). Do Kiwis need a people’s bank? B+FS. 13-14. Vesala, J. (1995). Testing for competition in banking: behavioural evidence from Finland. Bank of Finland Studies E1. 20