- Email: [email protected]
Market concentration and nonlinear pricing in European banking
Journal of Economics and Business 85 (2016) 1–12
Contents lists available at ScienceDirect
Journal of Economics and Business
Market concentration and nonlinear pricing in European banking Antonis A. Michis ∗ Central Bank of Cyprus, P.O Box 25529, CY-1395, Nicosia, Cyprus
a r t i c l e
i n f o
Article history: Received 26 January 2015 Received in revised form 18 November 2015 Accepted 9 February 2016 Available online 23 February 2016 JEL classification: G21 L11
a b s t r a c t An econometric test is proposed to show the existence of nonlinear pricing in the European market for loans. The test incorporates a measure of industry concentration to examine the impact of market structure on the use of nonlinear pricing tactics by banks. Econometric results using a panel dataset consisting of seven European countries suggest that nonlinear pricing is associated with increasing monopoly power in European banking. © 2016 Elsevier Inc. All rights reserved.
Keywords: Lending interest rates Market concentration Nonlinear pricing
1. Introduction Price discrimination, the practice of setting different prices for the same product or service, is an important area of research in the industrial organization literature and a pricing tactic that is extensively used by companies. This practice can be implemented based on observable consumer characteristics or through the self-selection of consumers who choose different versions of the same product. Price discrimination enables firms with monopoly power to increase their profits by extracting additional consumer surplus. This extraction is achieved by servicing more consumers based on their willingness to pay for a product or service.
∗ Corresponding author. Tel.: +357 22714278; fax: +357 22378164. E-mail address: [email protected] http://dx.doi.org/10.1016/j.jeconbus.2016.02.003 0148-6195/© 2016 Elsevier Inc. All rights reserved.
2
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Numerous empirical studies have investigated the existence of various forms of price discrimination in banking. However, the existence and causes of nonlinear pricing (a form of second-degree price discrimination) in credit markets have not been previously explored in the literature. This study uses data on the European market for loans to examine two hypotheses concerning the interest rate differences between large and small loans: (i) The difference (spread) between interest rates on small loans (“Up to or equal to D 1 million”) and interest rates on large loans (“Over D 1 million”) tends to be positive, in line with nonlinear pricing schedules. (ii) Increasing market concentration (monopoly power) tends to be associated with higher interest rate spreads. Price discrimination can influence profits, consumer welfare, tax revenues and economic efficiency (see Verboven, 2008; Courty & Pagliero, 2012). Nonlinear pricing in particular is a form of price discrimination that can be associated with positive welfare effects (Pepall, Richards, & Norman, 2008). In addition, the examination of the market concentration–nonlinear pricing relationship in the market for loans can provide valuable information for explaining the level of interest rates in Europe. It is therefore of interest to examine whether nonlinear pricing tends to be associated with higher levels of market concentration. The use of nonlinear pricing tactics by banks is an important consideration for markets dominated by a few large banking organizations and characterized by distribution limitations between banks and consumers. In consolidated markets, with limited competition from smaller banks, such distribution issues might arise in the form of reduced lending to small businesses, or difficulties in financing small scale projects. When such limitations exist, total welfare can decline with nonlinear pricing. The econometric test proposed in this study is based on a panel dataset consisting of seven European countries and ten annual time periods. The differences in lending interest rates, with and without nonlinear pricing, are modelled as a function of market concentration and other country- and marketspecific characteristics that likely influence interest rates in Europe. The results, based on a fixedeffects model specification, suggest that nonlinear pricing in the European market for loans tends to be associated with increasing market concentration. The rest of the article is organized as follows: Section 2 reviews the relevant literature, and Section 3 presents the dataset that will be used in the econometric analysis. Section 4 presents the econometric model and the empirical results of the study, and Section 5 concludes. 2. Nonlinear pricing 2.1. Theory There are three main types of price discrimination. First-degree price discrimination is based on setting different prices for each consumer using knowledge of their individual characteristics. It is rarely encountered in practice since it assumes perfect information about consumers’ preferences. With second-degree price discrimination, the price per unit charged for a product or service depends on the number of units purchased. It is a frequently employed pricing tactic by utilities associated with water and electricity supply. Third-degree price discrimination is another frequently used pricing tactic. It uses market segmentation based on observable consumer characteristics in order to set different prices per segment (e.g. providing students with discounts). Nonlinear pricing is the most common form of second-degree price discrimination and describes any pricing schedule in which the prices paid by the consumers for a product or service are not proportional to the quantities they purchase. Most of the theoretical research on nonlinear pricing (as in, for example, Schmalensee, 1981a; Armstrong & Vickers, 2001) has concentrated on two-part tariffs that consist of a fixed fee, which must be paid regardless of the consumed quantity, and a variable fee, which is proportional to the consumed quantity. Two conditions are necessary for the exercise of price discrimination: (i) the existence of market power and (ii) the existence of different market segments (sub-markets), each with a different price
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
3
elasticity of demand (Lipczynski, Wilson, & Goddard, 2013, p. 374). When price discrimination exists, as in nonlinear pricing schedules, the price difference between two products that differ, for example in terms size, can be decomposed into two sources: differences in cost and differences in margins. Price discrimination exists to the extent that the observed price differences are the result of differences in margins rather than differences in cost (Verboven, 2008). However, in many markets the monopolist does not have enough information on demand to assess the ability of consumers to pay for a product or service. In such cases the monopolist can implement a menu of prices and quantities (a nonlinear pricing schedule) to induce consumers to self-select into different segments and reveal information about their demand (Pepall et al., 2008, p. 128). By offering a menu of prices and quantities, the monopolist is able to sell to more consumers that, based on their demand functions, were not willing to buy the product (or service) at the uniform price associated with the monopoly. As a result, the monopolist’s total surplus with second-degree price discrimination is higher than with uniform monopoly pricing. However, since second-degree price discrimination is based on imperfect information about consumers’ preferences, it is less profitable than first-degree price discrimination (Lipczynski et al., 2013, p. 379). Nonlinear pricing can be associated with monopolistic or oligopolistic market structures. In a monopoly context, Oi (1971) demonstrated the optimality of the two-part tariff-pricing scheme, which achieves both allocation efficiency and profit maximization. Setting the price of a product close to marginal cost ensures efficiency, while the fixed-fee component achieves profit maximization by extracting the consumer surplus. Schmalensee (1981b) and Varian (1985) further demonstrated that in a price-discriminating monopoly, total welfare (the sum of the consumer and producer surpluses) can increase, provided that the total quantity produced increases too. In an oligopoly context, nonlinear pricing has been associated with increasing competition among firms, in the form of quantity dscounts. A related effect, demonstrated by Stole (1995), is that the quality distortion induced by monopolists to increase profits decreases with increasing competition in oligopolistic markets. Furthermore, Armstrong and Vickers (2001) and Rochet and Stole (2002) showed that, under specific conditions, two-part tariffs can be nearly optimal in oligopolistic markets.
2.2. Empirical studies The empirical literature on price discrimination and non-linear pricing can be divided into three main categories: (i) methods for identifying and measuring price discrimination (e.g. Clerides, 2004), (ii) examination of the sources of price discrimination (e.g. Borenstein & Rose, 1994) and (iii) examination of the economic consequences of price discrimination (e.g. Leslie, 2004). This study is more closely related to the second group of empirical studies, since it examines the impact of competition on the use of non-linear pricing tactics by banks in Europe. Numerous empirical studies have investigated the relationship between competition and price discrimination with mixed results. For example, Borenstein (1991), Stavins (2001) and Busse and Rysman (2005) found a positive relationship (more competition tends to be associated with increases in price discrimination) in the US gasoline, airline and yellow pages advertising markets, respectively. In contrast, Verboven (1999), Gerardi and Shapiro (2009) and Gaggero and Piga (2011) found a negative relationship in the European car, US airline and UK airline markets, respectively. In some markets a non-monotonic relationship is also possible as demonstrated by Dai, Liu, and Serfes (2014) and Clerides and Michis (2006). These findings suggest that the sources of price discrimination differ by market and each case should be examined carefully. The existence of various forms of price discrimination has been the subject of many empirical studies in the banking literature. Most of these studies have concentrated on topics like racial discrimination (Berkovec, Canner, Stuard, & Hannan, 1998; Cavalluzzo & Cavalluzzo, 1998), social discrimination (Gary-Bobo & Larribeau, 2004), credit card price discrimination mechanisms (Murphy & Ott, 1977), ATM surcharging (Hannan, Kiser, Prager, & McAndrews, 2003) and spatial price discrimination (Degryse & Ongena, 2005). Nearly all of these studies suggest the existence of some form of monopolistic price-discriminatory behaviour. A gap in the existing empirical literature concerns the
4
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
lack of any empirical support regarding the use of nonlinear pricing tactics by banks and how this is associated with changes in the market structure. 3. Data Three different sources of information were combined to construct the panel dataset used in this study. The cross-sectional units in the panel consist of Austria, Finland, France, Germany, Italy, the Netherlands and Spain, and the annual time periods cover the years from 2003 to 2012. The first source of information was the European Central Bank (ECB), from which the following variables were obtained for each country and year: the Herfindahl–Hirschman index of market concentration for loans, the national income and the average annual interest rates for “loans other than revolving loans and overdrafts, convenience and extended credit card debt”, as reported by the monetary and financial institutions sector in each country. The national income variable was preferred since it includes the economic activities of the residents of a country that occur abroad. Frequently, these activities are financed through the domestic banking system. Furthermore, the national income definition does not include the consumption of fixed capital (e.g. depreciation), which is not relevant for modelling interest rate spreads. Unlike GDP per capita, the ECB publishes only absolute values for the national income variable using a common methodology for all countries in Europe. With regard to interest rates, separate variables were obtained for the following loan categories (using the ECB definitions): • • • • • •
“Up to and including D 1 million” and maturity of “Up to 1 year”. “Up to and including D 1 million” and maturity of “Over 1 and up to 5 years”. “Up to and including D 1 million” and maturity of “Over 5 years”. “Over D 1 million” and maturity of “Up to 1 year”. “Over D 1 million” and maturity of “Over 1 and up to 5 years”. “Over D 1 million” and maturity of “Over 5 years”.
Additional sources of information included the Eurostat statistical database, from which the harmonized index of consumer prices for each country and year were obtained, and the European Banking Federation, from which the annual average Euribor rate (euro interbank offered rate) was obtained for each year in the sample. Inflation in the euro area is measured by the harmonized index of consumer prices that is also used in this study. All countries in the European Union construct this index based on the same methodology, which facilitates comparisons. The index is supported by a set of legal acts and standards that minimize the possibility of conceptual discrepancies between countries. The issue of whether the euro changeover (first introduced as a legal tender in 12 countries on 1 January, 2002) had an effect on prices and inflation was investigated by Dziuda and Mastrobuoni (2009). The authors did not find any significant increase in inflation as a result of the euro changeover, although there were incidents of a distortionary effect on prices. These incidents concerned countries in which consumers reported lower transparency and more problems during the changeover process. In the econometric model of Section 4, the interest rate spread will be used as a dependent variable, calculated as the difference between loans of “Up to and including D 1 million” and those of “Over D 1 million” for the three maturity segments: “Up to 1 year”, “Over 1 and up to 5 years” and “Over 5 years”. This variable will be used to test for the existence of nonlinear pricing in the provision of loans and how this is influenced by changes in the market structure. Table 1 presents summary statistics for all of the variables that will be used in the econometric analysis. It can be observed that the average value of the interest rate spread, 0.835, is positive, which suggests that interest rates on loans of up to and including D 1 million are on average higher than interest rates on loans that exceed the D 1 million benchmark. Data limitations do not permit a disaggregated analysis by loan type, since the ECB publishes only aggregated level data. Nevertheless, the market concentration–nonlinear pricing relationship considered in this study can still be investigated with aggregated data. In Appendix A, it is demonstrated
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
5
Table 1 Summary statistics. Variable
Sample
Mean
St dev
Min
Max
Interest rate spread (%) Euribor (%) Inflation (%) Herfindahl–Hirschman index National income (EUR millions)
210 210 210 210 210
0.835 2.309 2.074 0.101 1111,897
0.581 1.267 0.939 0.097 798,947
−0.210 0.685 −0.243 0.017 144,428
3.007 4.571 4.131 0.370 2705,513
0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 2003 Austria
2004
2005 Germany
2006 Spain
2007 Finland
2008
2009 France
2010 Italy
2011
2012 Netherlands
Fig. 1. Herfindahl–Hirschman indices.
with a simple example that the average interest rate spread variable can still be observed correctly with aggregated data and used in econometric models. The time development of the Herfindahl–Hirschman index by country is presented in Fig. 1. The concentration levels vary considerably across countries, most notably for Finland and the Netherlands, and there are also differences in trends (e.g. concentration is increasing in Italy but decreasing in Austria). Herfindahl–Hirschman indices at the national level are frequently used in studies concerned with bank competition in Europe, as in for example: Alegria and Schaeck (2008), Bikker and Haaf (2002), Carbo, Humphrey, Maudos, and Molyneux (2009), Claessens and Laeven (2004) and Coccorese (2005). To further investigate the variability inherent in the interest rate spread variable, Table 2 presents analytical values of this variable per loan duration segment, country and year. From the 210 spread values included in the table, the average interest rate on loans exceeding the D 1 million benchmark is higher than the average interest rate on loans lower than or equal to the D 1 million benchmark in only seven cases. All of these cases (indicated by the shaded areas in Table 2) concern loans with durations of more than five years and were from Austria (years 2004, 2008, 2009, 2011 and 2012), Germany (year 2008) and France (year 2007). As explained in Section 2, price discrimination exists to the extent that the observed price difference between two products that differ, for example in terms of size, is the result of differences in margins rather than differences in cost (Verboven, 2008). In empirical studies, price discrimination is identified either by controlling for differences in cost in econometric models or by assuming that differences in cost are zero or even negative. For this reason, the market concentration-interest rate spread
6
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Table 2 Interest rate spreads. Year Spreads for loans with duration up to 1 year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Spreads for loans with duration between 1–5 years 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Spreads for loans with duration more than 5 years 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
ITA
NED
0.307 0.610 0.780 0.539 0.705 0.706 0.966 0.628 0.326 0.446
1.106 1.143 0.935 0.756 0.928 0.851 1.125 1.104 0.965 1.593
0.716 0.688 0.480 0.413 0.352 0.597 1.512 1.085 1.185 1.319
0.934 0.898 1.008 1.038 0.972 0.797 1.518 1.569 0.710 1.667
2.017 1.815 1.581 1.050 0.742 0.874 2.033 1.745 1.557 1.534
1.633 1.531 1.331 1.186 1.057 1.005 2.269 2.455 2.249 2.501
0.364 0.609 0.188 0.260 0.308 0.532 1.302 0.828 0.483 0.196
0.487 0.271 0.502 0.213 0.517 0.734 1.213 0.971 0.856 0.933
0.447 0.388 0.351 0.103 −0.148 0.023 0.815 0.593 0.340 0.433
0.828 0.776 0.932 0.911 0.576 0.329 1.093 1.148 1.190 1.483
0.633 0.643 0.378 0.433 0.541 0.784 0.531 0.991 1.057 0.900
AUS
GER
SPA
FIN
0.888 0.767 0.589 0.483 0.432 0.443 0.528 0.476 0.371 0.480
1.190 1.265 1.164 1.111 1.035 0.946 0.709 0.994 0.922 0.886
1.124 0.962 0.802 0.614 0.545 0.803 1.628 1.472 1.524 2.392
0.603 0.583 0.455 0.561 0.553 0.498 0.668 0.810 0.595 0.732
0.819 0.351 0.421 0.574 0.536 0.319 0.733 0.093 0.187 0.232
1.010 1.133 0.852 0.566 0.376 0.202 0.597 0.873 0.857 1.041
1.227 1.713 1.194 1.088 0.794 0.991 2.709 2.179 2.021 2.185
0.383 −0.159 0.152 0.257 0.158 −0.082 −0.074 0.069 −0.078 −0.210
0.463 0.341 0.359 0.200 0.034 −0.116 0.087 0.179 0.103 0.075
0.474 0.551 0.616 0.688 0.438 0.890 1.304 1.806 3.007 2.515
FRA
relationship in Section 4 is estimated after controlling for changes in the main cost determinants of loans (Euribor and inflation), changes in market demand (national income) as well as differences between countries and loan durations (fixed effects). With regard to the degree of substitutability between large and small loans (traditional analysis of price discrimination assumes a single product in the market), it should be noted that multiple bank lending relationships exist and are well documented in the empirical banking literature (see Bennardo, Pagano, & Piccolo, 2015). Several explanations have been provided for this phenomenon. For example, a firm restricted to an exclusive relationship faces additional costs in the form of ex post rents (Sharpe, 1990) and it will be required to pay a lemon’s premium if it seeks cooperation with more lenders. This is because the information and data gathered by one bank are not easily accessible by other banks, which limits their ability for accurate evaluation and screening. Other explanations proposed in the literature include the use of multiple lending relationships as a signal of credit quality and the need to diversify and share the risks associated with risky firms with lower credit ratings by many banks (see, Farinha & Santos, 2002). The existence of these multiple bank relationships suggest that frequently firms prefer to substitute large loans from one bank with multiple smaller loans from many banks. In this context, the nonlinear pricing literature can provide valuable insights into the observed average interest rate differentials between large and small loans.
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
7
The positive interest rate differential between small (“Up to and including D 1 million”) and large (“Over D 1 million”) loans should be expected to reflect primarily differences in margins. This is because large loans involve higher monitoring costs for banks that compensate for any differences in underwriting costs per euro loaned. The latter should be expected to be higher for small loans. For example, Minnis (2011) notes that banks should be expected to demand more monitoring from large firms, mainly because they are associated with large loan sizes and additional opportunities for asset substitution. This extra monitoring effort suggests that banks do not consider large loans as carrying intrinsically less risk. In a recent study, Minnis and Sutherland (2015) investigated the factors that influence the use of financial statements and tax returns as monitoring devices for commercial loans. The authors found a positive and statistically significant relationship between loan sizes and the collection of financial statements and tax returns for monitoring purposes. Therefore, higher loan amounts increase the probability of collection, which suggests higher monitoring costs. Apart from higher monitoring costs, large loans are also associated with lower opportunities for diversification and therefore higher risk. This is an additional cost associated with large loans that can increase the provisions for bad loans in a bank’s balance sheet. In an empirical study using data from the Austrian central bank, Rossi, Schwaiger, and Winkler (2009) examined how loan portfolio diversification across different industries and loan sizes affected the realized risk and the profit efficiency of banks in Austria during the period 1997–2003. In the econometric analysis, higher diversification across different industries and loan sizes was found to reduce the risk (provision for bad loans) and increase the profits (risk-adjusted returns) of loan portfolios. Consequently, large loans entail higher monitoring costs and lower opportunities for diversification, which should be expected to offset any underwriting costs differentials between small and large loans1 . In the context of this study, large loans that exceed the D 1 million benchmark are on average associated with lower interest rates, which suggests the existence of nonlinear pricing. It is therefore of interest to examine the factors influencing the level of these spreads, particularly focusing on changes in the market structure of each country, as suggested in the literature. As already noted in Section 2, a necessary condition for price discrimination is the existence of market segmentation, each with a different price elasticity of demand. Differences in interest-rate elasticties between different classes of borrowers (e.g. executives vs. blue collars) have been reported in the literature by Gary-Bobo and Larribeau (2004). Using a model that simultaneously explains the size of a loan and its associated interest rate, the authors found evidence of (first-degree) monopolistic price discrimination in French mortgage data. Blue-collar workers were found to pay higher interest rates than executives and the estimated price mark-ups were too high to reflect default risks only. The econometric results in this study suggest that the borrowers’ price-elasticity of demand for housing in France varies with occupational status, and is inversely related with the lender’s interest rate mark-ups. Furthermore, the loans provided to executives were, on average, 40% higher than the loans provided to blue collar workers. Differences in (quantity-dependent) loan interest rates between large and small borrowers were also discussed in a study by Schreft and Villamil (1992). The authors examined data from the “Survey of Terms of Bank Lending” in the US and found evidence of an inverse relationship between average interest rates and commercial loan sizes. According to the model presented by the authors, a profit maximizing lender with market power and imperfect information about borrowers’ characteristics can use second degree price discrimination as a form of credit rationing that limits borrowing. However, credit rationing is not applied to the largest borrowers. More importantly, the lender has an incentive to offer quantity-dependent loan interest rates even if credit risk and the associated fixed underwriting costs are the same for all classes of borrowers.
1 The degree of offset between the fixed underwriting costs of loans versus the size-related monitoring and diversification costs is a worthwhile topic for future research, but requires the availability of more detailed datasets on loans contracts and their restrictive covenants. Research in this area can provide valuable insights into the reasons for the observed interest rate differentials between small and large loans.
8
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
4. Econometric analysis 4.1. Model In this section, a panel data econometric model is used to examine the determinants of interest rate spreads, with particular emphasis on market concentration. To measure the difference between interest rates on loans below or equal to the D 1 million (pS ) benchmark and interest rates on loans above the D 1 million (pL ) benchmark, the following simple metrics were defined for the interest rate spread: S a = ps − pL
(1)
S b = (ps − pL )2 .
(2)
Both metrics are used as dependent variables in econometric models. The results for the model with the squared metric as the dependend variable are presented in Appendix B. The econometric model used in the analysis has the following form: Sita,b = ˇ0 +
6
ıCi Ci +
2
i=1
ıM Mj + ˇ1 log EBFt + ˇ2 INFit + ˇ3 log HHIit + ˇ4 log Yit j
j=1
+ ˇ5 log T + εit . where Ci represents six country indicator variables, except for the Netherlands, which is represented by the intercept ˇ0 ; Mj represents 2 indicator variables for the different loan maturities in the dataset, “Up to 1 year” and “Over 1 and up to 5 years”. Loans with maturity of “Over 5 years” are captured by the intercept. The remaining variables are defined as follows: EBFt is the Euribor rate based on which banks in the Eurozone exchange funds; INFit is the annual inflation rate of country i in year t, calculated from the harmonized index of consumer prices; HHIit represents the Herfindahl–Hirschman index of market concentration; and Yit is the national income of country i in year t. The model also includes a time trend variable (T ) and the error term εit . An alternative model specification that included time-fixed effects in place of the trend variable was also examined. The two model specifications generated similar results; however, the specification with the trend variable was preferred on the basis that it involved a smaller number of coefficients to be estimated and therefore a higher number of degrees of freedom. 4.2. Results The model was estimated with the GMM method, and the coefficient estimates along with their robust standard errors are included in Table 3. The standard errors were estimated with the heteroskedasticity and autocorrelation robust covariance matrix estimator proposed by Andrews (1991). The model provided a relatively good fit to the data and all of the estimated coefficients were statistically significant at the 1%, 5% or 10% level except in the cases of income, trend, Austria and Finland. The negative coefficient for the Euribor variable suggests that increases in the euro interbank offered rate tend to be associated with reductions in the interest rate spreads. The Euribor rate constitutes a basic input cost for banks and increases in this cost should be expected to reduce the size of interest rate reductions provided on large loans. To see this, consider a market with two segments, one for small loans and one for large loans. The interest rate spread is calculated as the difference in interest rates between small and large loans, due to the existence of a nonlinear pricing schedule. Therefore, the marginal price for small loans is ps (q, .), the marginal price for large loans is pL (q, .) = ps (q, .) − d (q, .) and the interest rate spread d(q, .) is a function of the loan amount. In turn, the loan amount demanded q(ie , .) is a function of the Euribor rate (ie ). By straightforward application of the chain-rule, the partial derivative of the interest rate spread with respect to the Euribor rate is ∂d/∂ie = ∂d/∂q · ∂q/∂ie . In the presence of a nonlinear pricing
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
9
Table 3 GMM estimation results. Variable
Coefficient
St error
Intercept log Euribor Inflation log HHI log Income log Trend Up to 1 year 1 to 5 years Austria Germany Spain Finland France Italy R-squared
2.589 −0.280 −0.072 0.553 −0.067 −0.021 0.254 0.508 0.467 1.181 1.497 −0.100 0.755 1.615 0.591
0.845* 0.060* 0.034** 0.253** 0.047 0.059 0.110** 0.112* 0.376 0.577** 0.400* 0.182 0.318** 0.521*
Significance level: * 1%; ** 5%; *** 10%.
schedule, the partial derivative ∂d/∂q should be expected to be positive. This is because an increase in the loan amount demanded, will tend to increase the reduction (spread) in the interest rate associated with large loans. In contrast, the partial derivative ∂q/∂ie should be expected to be negative. This is because increases in the Euribor rate will, tend to increase the interest rate on large loans and therefore reduce the loan amounts demanded. Consequently, the partial derivative ∂d/∂ie should be expected to be negative, which is consistent with the sign of the Euribor coefficient in Table 3. The coefficient for the inflation rate is also negative and statistically significant at the 5% level. Therefore, an increase in the inflation rate tends to be associated with a reduction in the size of the spreads. This trend results because increased inflation expectations tend to reduce the value of all future loan payments by households and enterprises and therefore raise the cost of lending for banks. As a result, banks working with lower profit margins are less willing to provide interest rate reductions on large loans. In the case of the Herfindahl–Hirschman Index variable, the coefficient is positive and statistically significant at the 5% level, which suggests that higher market concentration (increasing monopoly power) tends to be associated with higher interest rate spreads. This result is consistent with the literature presented in Section 2, according to which nonlinear pricing can be associated with a monopolistic market structure and the extraction of consumer surplus. With regard to the country indicator variables, it is worth noting that the average difference between interest rates on small loans and interest rates on large loans is higher in Italy, Spain and Germany. This is after controlling for changes in the other explanatory variables. In contrast, it is lower in Finland and the Netherlands. This finding suggests that increases in the actual spreads of these two countries are mainly motivated by changes in the other explanatory variables, like for example increases in market concentration and the exercise of nonlinear pricing tactics. As demonstrated in Fig. 1, Finland and the Netherlands are associated with higher concentration levels than the other countries.
5. Conclusions The econometric results of the previous section are consistent with the use of nonlinear pricing in monopolistic markets, which enables firms with market power to extract additional consumer surplus and increase their revenues. Nonlinear pricing is not necessarily harmful for welfare, even though it can be associated with a monopolistic market structure. Schmalensee (1981b) and Varian (1985) showed that total welfare can increase with nonlinear pricing, provided that the total quantity produced increases too.
10
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
However, an important distinction holds. Despite the increase in total welfare, consumer welfare in particular decreases. On the other hand, more consumers are served under nonlinear pricing, which (from the scope of financial intermediation) is important for entrepreneurship and project financing in an economy. According to Cabral (2000, p. 181), price discrimination should not be of concern, as long as there are no distribution issues between firms and consumers. It is therefore important for the monetary authorities to ensure that liquidity and project financing in the market for loans is equally accessible to the whole distribution of consumers and firms in an economy. European banking has gone through considerable changes during the last twenty years, following rapid development in deregulation, financial innovation and the process of globalization. These changes have intensified competition in financial markets and have led to a shrinkage of profit margins for banks. Mishkin, Matthews, and Giuliodori (2013, p. 256–257) note that in response to these changes the banking industry entered a phase of consolidation with strategic mergers and acquisitions. As a result of this consolidation process the total number of banking institutions in the Eurozone dropped from 11877 in 1988 to just 6842 in 2008. The formation of large banking organizations through consolidation is frequently associated with economies of scale and scope, which can increase efficiency in the industry. An additional benefit refers to the opportunity for improving diversification in loan portfolios that can contribute towards a less vulnerable banking system. However, bank consolidation is also associated with two important considerations. First, it can drive many small banks out of the market and therefore reduce lending to small businesses. Second, consolidation can increase market concentration and reduce competition in the industry with a few very large banking organizations dominating the industry. According to Mishkin et al. (2013, p. 266), such banking markets exist in Europe within national boundaries. In a concentrated market where nonlinear pricing is used by banks, distribution issues between banks and consumers might arise in the form of reduced lending towards small businesses or difficulties in financing small scale projects. In such an environment, the gains for the banks might not compensate for the reduction in consumer welfare. This possibility suggests that a worthwhile goal for public policy should be to ensure that small businesses will not face any reduction in lending and liquidity and sufficient financing for small-scale projects will continue to exist. Appendix A. Consider a block-declining tariff (a standard form of nonlinear pricing), which consists of two pricing segments separated by a quantity threshold X* . Following DeSalvo and Huq (2002), this tariff can be represented as follows:
C(X) =
P1 X,
0 ≤ X < X∗
(P1 − P2 ) X ∗ + P2 X,
X ≥ X∗
For loan amounts up to X* units (small loans) the borrower has to pay an amount equal to C(X) = P1 X, while for loan amounts higher than X* (large loans) the expenses are equal to C(X) = (P1 − P2 ) X ∗ + P2 X. Over the first tariff segment the average interest rate for small loans is constant and equal to p¯ s = P1 . In contrast, it becomes variable over the second segment and equal to p¯ L = P2 + (P1 − P2 ) X ∗ /X. The interest rate spread (d) is calculated as the difference in interest rates between small and large loans, due to the existence of a nonlinear pricing schedule in the market. Average values of the interest rate spread variable can be obtained by considering the difference between the two average interest rates as follows:
X∗ d¯ = P1 − P2 + (P1 − P2 ) X
= (P1 − P2 ) 1 −
X∗ X
If P1 > P2 and in the second pricing segment X > X* , then in a price discriminating monopoly the average interest rate spread takes positive values. Consequently, regressing this average interest rate spread variable on market concentration should be expected to provide a positive coefficient
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
11
(everything else being equal), since increasing monopoly power (higher market concentration) should be expected to be associated with increases in the average interest rate spread variable. Appendix B. The econometric model presented in Section 4 was also estimated using the squared interest rate spread-metric Sb as a dependent variable. The results are similar to those in Table 3 for the simple metric and the coefficients have the same sign in all cases except for the trend regressor. B.1. GMM estimation results with squared metric. Variable
Coefficient
St error
Intercept log Euribor Inflation log HHI log Income log Trend Up to 1 year 1 to 5 years Austria Germany Spain Finland France Italy R-squared
4.479 −0.648 −0.197 1.391 −0.096 0.065 0.187 0.955 1.635 3.058 3.806 −0.376 1.991 3.903 0.490
2.170** 0.147* 0.116*** 0.708** 0.101 0.144 0.255 0.298* 1.039 1.606*** 1.137* 0.378 0.842** 1.540**
Significance level: * 1%; ** 5%; *** 10%.
References Alegria, C., & Schaeck, K. (2008). On measuring concentration in banking systems. Finance Research Letters, 5, 59–67. Andrews, D. W. K. (1991). Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica, 59(3), 817–858. Armstrong, M., & Vickers, V. (2001). Competitive price discrimination. RAND Journal of Economics, 32(4), 579–605. Berkovec, J. A., Canner, G. B., Stuard, G. A., & Hannan, T. H. (1998). Discrimination, competition, and loan performance in FHA mortgage lending. Review of Economics and Statistics, 80(2), 241–250. Bennardo, A., Pagano, M., & Piccolo, A. (2015). Multiple bank lending, creditor rights, and information sharing. Review of Finance, 19, 519–570. Bikker, J. A., & Haaf, K. (2002). Competition, concentration and their relationship: An empirical analysis of the banking industry. Journal of Banking & Finance, 26, 2191–2214. Borenstein, S. (1991). Selling costs and switching costs: Explaining retail gasoline margins. RAND Journal of Economics, 22, 354–369. Borenstein, S., & Rose, N. (1994). Competition and price dispersion in the U.S. airline industry. Journal of Political Economy, 102, 653–683. Busse, M., & Rysman, M. (2005). Competition and price discrimination in Yellow Pages advertising. RAND Journal of Economics, 36, 378–390. Cabral, L. M. B. (2000). Introduction to industrial organization. Cambridge, MA: MIT Press. Carbo, S., Humphrey, D. B., Maudos, J., & Molyneux, P. (2009). Cross-country comparisons of competition and pricing power in European banking. Journal of International Money and Finance, 28, 115–134. Cavalluzzo, K. S., & Cavalluzzo, L. C. (1998). Market structure and discrimination: The case of small businesses. Journal of Money, Credit and Banking, 30(4), 771–792. Claessens, S., & Laeven, L. (2004). What drives bank competition? Some international evidence. Journal of Money, Credit and Banking, 36, 563–583. Clerides, S. (2004). Price discrimination with differentiated products: definitions and identification. Economic Inquiry, 42, 402–412. Clerides, S., & Michis, A. (2006). Market concentration and nonlinear pricing: Evidence from detergents prices in six countries. In Working paper. University of Cyprus. Coccorese, P. (2005). Competition in markets with dominant firms: A note on the evidence from the Italian banking industry. Journal of Banking & Finance, 29, 1083–1093. Courty, P., & Pagliero, M. (2012). The impact of price discrimination on revenue: Evidence form the concert industry. Review of Economic and Statistics, 94, 359–369.
12
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Dai, M., Liu, Q., & Serfes, K. (2014). Is the effect of competition on price dispersion non-monotonic? Evidence from the U.S. airline industry. Review of Economics and Statistics, 96, 161–170. Degryse, H., & Ongena, S. (2005). Distance, lending relationships, and competition. Journal of Finance, 60(1), 231–266. DeSalvo, J. S., & Huq, M. (2002). Introducing nonlinear pricing into consumer theory. Journal of Economic Education, 33, 166–179. Dziuda, W., & Mastrobuoni, G. (2009). The euro changeover and its effects on price transparency and inflation. Journal of Money, Credit and Banking, 41, 101–129. Farinha, L. A., & Santos, J. A. C. (2002). Switching from single to multiple bank lending relationships: Determinants and implications. Journal of Financial Intermediation, 11, 124–151. Gaggero, A., & Piga, C. A. (2011). Airline market power and intertemporal price dispersion. Journal of Industrial Economics, 59, 552–577. Gary-Bobo, R. J., & Larribeau, S. (2004). A structural econometric model of price discrimination in the French mortgage lending industry. International Journal of Industrial Organization, 22(1), 101–134. Gerardi, K. S., & Shapiro, A. H. (2009). Does competition reduce price dispersion? New evidence from the airline industry. Journal of Political Economy, 117, 1–37. Hannan, T. H., Kiser, E. K., Prager, R. A., & McAndrews, J. J. (2003). To surcharge or not to surcharge: An empirical investigation of ATM pricing. Review of Economics and Statistics, 85(4), 990–1002. Leslie, P. (2004). Price discrimination in Broadway theatre. RAND Journal of Economics, 35, 520–541. Lipczynski, J., Wilson, J. O. S., & Goddard, J. (2013). Industrial organization: Competition, strategy and policy (4th ed.). Harlow: Pearson. Minnis, M. (2011). The value of financial statement verification in debt financing: evidence from private U.S. firms. Journal of Accounting Research, 49, 457–506. Minnis, M., & Sutherland, A. (2015). Financial statements as monitoring mechanisms: Evidence from small commercial loans. In Working paper. University of Chicago Booth School of Business. Mishkin, F. S., Matthews, K., & Giuliodori, M. (2013). The economics of money banking and financial markets (European ed.). Harlow: Pearson Education. Murphy, M. M., & Ott, M. (1977). Retail credit, credit cards and price discrimination. Southern Economic Journal, 43(3), 1303–1312. Oi, W. Y. (1971). A Disneyland dilemma: Two-part tariffs for a Mickey Mouse monopoly. Quarterly Journal of Economics, 85(1), 77–96. Pepall, L., Richards, D., & Norman, G. (2008). Industrial organization: Contemporary theory and empirical applications. Cornwall: Blackwell Publishing. Rochet, J.-C., & Stole, L. A. (2002). Nonlinear pricing with random participation. Review of Economic Studies, 69(1), 277–311. Rossi, S. P. S., Schwaiger, M. S., & Winkler, G. (2009). How loan portfolio diversification affects risk, efficiency and capitalization: A managerial behavior model for Austrian banks. Journal of Banking & Finance, 33, 2218–2226. Schmalensee, R. (1981a). Monopolistic two-part pricing arrangements. Bell Journal of Economics, 12(2), 445–466. Schmalensee, R. (1981b). Output and welfare implications of monopolistic third-degree price discrimination. American Economic Review, 71(1), 242–247. Schreft, S. L., & Villamil, A. P. (1992). Credit rationaing by loan size in commercial markets. Federal Reserve Bank of Richmond Economic Review, (May–June), 1–8. Sharpe, S. A. (1990). Asymmetric Information, bank lending and implicit contracts: A stylized model of customer relationships. Journal of Finance, 45, 1069–1087. Stavins, J. (2001). Price discrimination in the airline market: The effect of market concentration. Review of Economics and Statistics, 83, 200–202. Stole, L. A. (1995). Nonlinear pricing and oligopoly. Journal of Economics and Management Strategy, 4(4), 529–562. Varian, H. R. (1985). Price discrimination and social welfare. American Economic Review, 75(4), 870–875. Verboven, F. (1999). Product line rivalry and market segmentation, with an application to the pricing of optional engine power on automobiles. Journal of Industrial Economics, 47, 399–425. Verboven, F. (2008). Price discrimination: Empirical studies. In N. Stephen, Durlauf, E. Lawrence, & Blume (Eds.), The new Palgrave: A dictionary in economics. Hampshire: Palgrave MacMillan.
Contents lists available at ScienceDirect
Journal of Economics and Business
Market concentration and nonlinear pricing in European banking Antonis A. Michis ∗ Central Bank of Cyprus, P.O Box 25529, CY-1395, Nicosia, Cyprus
a r t i c l e
i n f o
Article history: Received 26 January 2015 Received in revised form 18 November 2015 Accepted 9 February 2016 Available online 23 February 2016 JEL classification: G21 L11
a b s t r a c t An econometric test is proposed to show the existence of nonlinear pricing in the European market for loans. The test incorporates a measure of industry concentration to examine the impact of market structure on the use of nonlinear pricing tactics by banks. Econometric results using a panel dataset consisting of seven European countries suggest that nonlinear pricing is associated with increasing monopoly power in European banking. © 2016 Elsevier Inc. All rights reserved.
Keywords: Lending interest rates Market concentration Nonlinear pricing
1. Introduction Price discrimination, the practice of setting different prices for the same product or service, is an important area of research in the industrial organization literature and a pricing tactic that is extensively used by companies. This practice can be implemented based on observable consumer characteristics or through the self-selection of consumers who choose different versions of the same product. Price discrimination enables firms with monopoly power to increase their profits by extracting additional consumer surplus. This extraction is achieved by servicing more consumers based on their willingness to pay for a product or service.
∗ Corresponding author. Tel.: +357 22714278; fax: +357 22378164. E-mail address: [email protected] http://dx.doi.org/10.1016/j.jeconbus.2016.02.003 0148-6195/© 2016 Elsevier Inc. All rights reserved.
2
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Numerous empirical studies have investigated the existence of various forms of price discrimination in banking. However, the existence and causes of nonlinear pricing (a form of second-degree price discrimination) in credit markets have not been previously explored in the literature. This study uses data on the European market for loans to examine two hypotheses concerning the interest rate differences between large and small loans: (i) The difference (spread) between interest rates on small loans (“Up to or equal to D 1 million”) and interest rates on large loans (“Over D 1 million”) tends to be positive, in line with nonlinear pricing schedules. (ii) Increasing market concentration (monopoly power) tends to be associated with higher interest rate spreads. Price discrimination can influence profits, consumer welfare, tax revenues and economic efficiency (see Verboven, 2008; Courty & Pagliero, 2012). Nonlinear pricing in particular is a form of price discrimination that can be associated with positive welfare effects (Pepall, Richards, & Norman, 2008). In addition, the examination of the market concentration–nonlinear pricing relationship in the market for loans can provide valuable information for explaining the level of interest rates in Europe. It is therefore of interest to examine whether nonlinear pricing tends to be associated with higher levels of market concentration. The use of nonlinear pricing tactics by banks is an important consideration for markets dominated by a few large banking organizations and characterized by distribution limitations between banks and consumers. In consolidated markets, with limited competition from smaller banks, such distribution issues might arise in the form of reduced lending to small businesses, or difficulties in financing small scale projects. When such limitations exist, total welfare can decline with nonlinear pricing. The econometric test proposed in this study is based on a panel dataset consisting of seven European countries and ten annual time periods. The differences in lending interest rates, with and without nonlinear pricing, are modelled as a function of market concentration and other country- and marketspecific characteristics that likely influence interest rates in Europe. The results, based on a fixedeffects model specification, suggest that nonlinear pricing in the European market for loans tends to be associated with increasing market concentration. The rest of the article is organized as follows: Section 2 reviews the relevant literature, and Section 3 presents the dataset that will be used in the econometric analysis. Section 4 presents the econometric model and the empirical results of the study, and Section 5 concludes. 2. Nonlinear pricing 2.1. Theory There are three main types of price discrimination. First-degree price discrimination is based on setting different prices for each consumer using knowledge of their individual characteristics. It is rarely encountered in practice since it assumes perfect information about consumers’ preferences. With second-degree price discrimination, the price per unit charged for a product or service depends on the number of units purchased. It is a frequently employed pricing tactic by utilities associated with water and electricity supply. Third-degree price discrimination is another frequently used pricing tactic. It uses market segmentation based on observable consumer characteristics in order to set different prices per segment (e.g. providing students with discounts). Nonlinear pricing is the most common form of second-degree price discrimination and describes any pricing schedule in which the prices paid by the consumers for a product or service are not proportional to the quantities they purchase. Most of the theoretical research on nonlinear pricing (as in, for example, Schmalensee, 1981a; Armstrong & Vickers, 2001) has concentrated on two-part tariffs that consist of a fixed fee, which must be paid regardless of the consumed quantity, and a variable fee, which is proportional to the consumed quantity. Two conditions are necessary for the exercise of price discrimination: (i) the existence of market power and (ii) the existence of different market segments (sub-markets), each with a different price
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
3
elasticity of demand (Lipczynski, Wilson, & Goddard, 2013, p. 374). When price discrimination exists, as in nonlinear pricing schedules, the price difference between two products that differ, for example in terms size, can be decomposed into two sources: differences in cost and differences in margins. Price discrimination exists to the extent that the observed price differences are the result of differences in margins rather than differences in cost (Verboven, 2008). However, in many markets the monopolist does not have enough information on demand to assess the ability of consumers to pay for a product or service. In such cases the monopolist can implement a menu of prices and quantities (a nonlinear pricing schedule) to induce consumers to self-select into different segments and reveal information about their demand (Pepall et al., 2008, p. 128). By offering a menu of prices and quantities, the monopolist is able to sell to more consumers that, based on their demand functions, were not willing to buy the product (or service) at the uniform price associated with the monopoly. As a result, the monopolist’s total surplus with second-degree price discrimination is higher than with uniform monopoly pricing. However, since second-degree price discrimination is based on imperfect information about consumers’ preferences, it is less profitable than first-degree price discrimination (Lipczynski et al., 2013, p. 379). Nonlinear pricing can be associated with monopolistic or oligopolistic market structures. In a monopoly context, Oi (1971) demonstrated the optimality of the two-part tariff-pricing scheme, which achieves both allocation efficiency and profit maximization. Setting the price of a product close to marginal cost ensures efficiency, while the fixed-fee component achieves profit maximization by extracting the consumer surplus. Schmalensee (1981b) and Varian (1985) further demonstrated that in a price-discriminating monopoly, total welfare (the sum of the consumer and producer surpluses) can increase, provided that the total quantity produced increases too. In an oligopoly context, nonlinear pricing has been associated with increasing competition among firms, in the form of quantity dscounts. A related effect, demonstrated by Stole (1995), is that the quality distortion induced by monopolists to increase profits decreases with increasing competition in oligopolistic markets. Furthermore, Armstrong and Vickers (2001) and Rochet and Stole (2002) showed that, under specific conditions, two-part tariffs can be nearly optimal in oligopolistic markets.
2.2. Empirical studies The empirical literature on price discrimination and non-linear pricing can be divided into three main categories: (i) methods for identifying and measuring price discrimination (e.g. Clerides, 2004), (ii) examination of the sources of price discrimination (e.g. Borenstein & Rose, 1994) and (iii) examination of the economic consequences of price discrimination (e.g. Leslie, 2004). This study is more closely related to the second group of empirical studies, since it examines the impact of competition on the use of non-linear pricing tactics by banks in Europe. Numerous empirical studies have investigated the relationship between competition and price discrimination with mixed results. For example, Borenstein (1991), Stavins (2001) and Busse and Rysman (2005) found a positive relationship (more competition tends to be associated with increases in price discrimination) in the US gasoline, airline and yellow pages advertising markets, respectively. In contrast, Verboven (1999), Gerardi and Shapiro (2009) and Gaggero and Piga (2011) found a negative relationship in the European car, US airline and UK airline markets, respectively. In some markets a non-monotonic relationship is also possible as demonstrated by Dai, Liu, and Serfes (2014) and Clerides and Michis (2006). These findings suggest that the sources of price discrimination differ by market and each case should be examined carefully. The existence of various forms of price discrimination has been the subject of many empirical studies in the banking literature. Most of these studies have concentrated on topics like racial discrimination (Berkovec, Canner, Stuard, & Hannan, 1998; Cavalluzzo & Cavalluzzo, 1998), social discrimination (Gary-Bobo & Larribeau, 2004), credit card price discrimination mechanisms (Murphy & Ott, 1977), ATM surcharging (Hannan, Kiser, Prager, & McAndrews, 2003) and spatial price discrimination (Degryse & Ongena, 2005). Nearly all of these studies suggest the existence of some form of monopolistic price-discriminatory behaviour. A gap in the existing empirical literature concerns the
4
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
lack of any empirical support regarding the use of nonlinear pricing tactics by banks and how this is associated with changes in the market structure. 3. Data Three different sources of information were combined to construct the panel dataset used in this study. The cross-sectional units in the panel consist of Austria, Finland, France, Germany, Italy, the Netherlands and Spain, and the annual time periods cover the years from 2003 to 2012. The first source of information was the European Central Bank (ECB), from which the following variables were obtained for each country and year: the Herfindahl–Hirschman index of market concentration for loans, the national income and the average annual interest rates for “loans other than revolving loans and overdrafts, convenience and extended credit card debt”, as reported by the monetary and financial institutions sector in each country. The national income variable was preferred since it includes the economic activities of the residents of a country that occur abroad. Frequently, these activities are financed through the domestic banking system. Furthermore, the national income definition does not include the consumption of fixed capital (e.g. depreciation), which is not relevant for modelling interest rate spreads. Unlike GDP per capita, the ECB publishes only absolute values for the national income variable using a common methodology for all countries in Europe. With regard to interest rates, separate variables were obtained for the following loan categories (using the ECB definitions): • • • • • •
“Up to and including D 1 million” and maturity of “Up to 1 year”. “Up to and including D 1 million” and maturity of “Over 1 and up to 5 years”. “Up to and including D 1 million” and maturity of “Over 5 years”. “Over D 1 million” and maturity of “Up to 1 year”. “Over D 1 million” and maturity of “Over 1 and up to 5 years”. “Over D 1 million” and maturity of “Over 5 years”.
Additional sources of information included the Eurostat statistical database, from which the harmonized index of consumer prices for each country and year were obtained, and the European Banking Federation, from which the annual average Euribor rate (euro interbank offered rate) was obtained for each year in the sample. Inflation in the euro area is measured by the harmonized index of consumer prices that is also used in this study. All countries in the European Union construct this index based on the same methodology, which facilitates comparisons. The index is supported by a set of legal acts and standards that minimize the possibility of conceptual discrepancies between countries. The issue of whether the euro changeover (first introduced as a legal tender in 12 countries on 1 January, 2002) had an effect on prices and inflation was investigated by Dziuda and Mastrobuoni (2009). The authors did not find any significant increase in inflation as a result of the euro changeover, although there were incidents of a distortionary effect on prices. These incidents concerned countries in which consumers reported lower transparency and more problems during the changeover process. In the econometric model of Section 4, the interest rate spread will be used as a dependent variable, calculated as the difference between loans of “Up to and including D 1 million” and those of “Over D 1 million” for the three maturity segments: “Up to 1 year”, “Over 1 and up to 5 years” and “Over 5 years”. This variable will be used to test for the existence of nonlinear pricing in the provision of loans and how this is influenced by changes in the market structure. Table 1 presents summary statistics for all of the variables that will be used in the econometric analysis. It can be observed that the average value of the interest rate spread, 0.835, is positive, which suggests that interest rates on loans of up to and including D 1 million are on average higher than interest rates on loans that exceed the D 1 million benchmark. Data limitations do not permit a disaggregated analysis by loan type, since the ECB publishes only aggregated level data. Nevertheless, the market concentration–nonlinear pricing relationship considered in this study can still be investigated with aggregated data. In Appendix A, it is demonstrated
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
5
Table 1 Summary statistics. Variable
Sample
Mean
St dev
Min
Max
Interest rate spread (%) Euribor (%) Inflation (%) Herfindahl–Hirschman index National income (EUR millions)
210 210 210 210 210
0.835 2.309 2.074 0.101 1111,897
0.581 1.267 0.939 0.097 798,947
−0.210 0.685 −0.243 0.017 144,428
3.007 4.571 4.131 0.370 2705,513
0,40 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 2003 Austria
2004
2005 Germany
2006 Spain
2007 Finland
2008
2009 France
2010 Italy
2011
2012 Netherlands
Fig. 1. Herfindahl–Hirschman indices.
with a simple example that the average interest rate spread variable can still be observed correctly with aggregated data and used in econometric models. The time development of the Herfindahl–Hirschman index by country is presented in Fig. 1. The concentration levels vary considerably across countries, most notably for Finland and the Netherlands, and there are also differences in trends (e.g. concentration is increasing in Italy but decreasing in Austria). Herfindahl–Hirschman indices at the national level are frequently used in studies concerned with bank competition in Europe, as in for example: Alegria and Schaeck (2008), Bikker and Haaf (2002), Carbo, Humphrey, Maudos, and Molyneux (2009), Claessens and Laeven (2004) and Coccorese (2005). To further investigate the variability inherent in the interest rate spread variable, Table 2 presents analytical values of this variable per loan duration segment, country and year. From the 210 spread values included in the table, the average interest rate on loans exceeding the D 1 million benchmark is higher than the average interest rate on loans lower than or equal to the D 1 million benchmark in only seven cases. All of these cases (indicated by the shaded areas in Table 2) concern loans with durations of more than five years and were from Austria (years 2004, 2008, 2009, 2011 and 2012), Germany (year 2008) and France (year 2007). As explained in Section 2, price discrimination exists to the extent that the observed price difference between two products that differ, for example in terms of size, is the result of differences in margins rather than differences in cost (Verboven, 2008). In empirical studies, price discrimination is identified either by controlling for differences in cost in econometric models or by assuming that differences in cost are zero or even negative. For this reason, the market concentration-interest rate spread
6
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Table 2 Interest rate spreads. Year Spreads for loans with duration up to 1 year 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Spreads for loans with duration between 1–5 years 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 Spreads for loans with duration more than 5 years 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012
ITA
NED
0.307 0.610 0.780 0.539 0.705 0.706 0.966 0.628 0.326 0.446
1.106 1.143 0.935 0.756 0.928 0.851 1.125 1.104 0.965 1.593
0.716 0.688 0.480 0.413 0.352 0.597 1.512 1.085 1.185 1.319
0.934 0.898 1.008 1.038 0.972 0.797 1.518 1.569 0.710 1.667
2.017 1.815 1.581 1.050 0.742 0.874 2.033 1.745 1.557 1.534
1.633 1.531 1.331 1.186 1.057 1.005 2.269 2.455 2.249 2.501
0.364 0.609 0.188 0.260 0.308 0.532 1.302 0.828 0.483 0.196
0.487 0.271 0.502 0.213 0.517 0.734 1.213 0.971 0.856 0.933
0.447 0.388 0.351 0.103 −0.148 0.023 0.815 0.593 0.340 0.433
0.828 0.776 0.932 0.911 0.576 0.329 1.093 1.148 1.190 1.483
0.633 0.643 0.378 0.433 0.541 0.784 0.531 0.991 1.057 0.900
AUS
GER
SPA
FIN
0.888 0.767 0.589 0.483 0.432 0.443 0.528 0.476 0.371 0.480
1.190 1.265 1.164 1.111 1.035 0.946 0.709 0.994 0.922 0.886
1.124 0.962 0.802 0.614 0.545 0.803 1.628 1.472 1.524 2.392
0.603 0.583 0.455 0.561 0.553 0.498 0.668 0.810 0.595 0.732
0.819 0.351 0.421 0.574 0.536 0.319 0.733 0.093 0.187 0.232
1.010 1.133 0.852 0.566 0.376 0.202 0.597 0.873 0.857 1.041
1.227 1.713 1.194 1.088 0.794 0.991 2.709 2.179 2.021 2.185
0.383 −0.159 0.152 0.257 0.158 −0.082 −0.074 0.069 −0.078 −0.210
0.463 0.341 0.359 0.200 0.034 −0.116 0.087 0.179 0.103 0.075
0.474 0.551 0.616 0.688 0.438 0.890 1.304 1.806 3.007 2.515
FRA
relationship in Section 4 is estimated after controlling for changes in the main cost determinants of loans (Euribor and inflation), changes in market demand (national income) as well as differences between countries and loan durations (fixed effects). With regard to the degree of substitutability between large and small loans (traditional analysis of price discrimination assumes a single product in the market), it should be noted that multiple bank lending relationships exist and are well documented in the empirical banking literature (see Bennardo, Pagano, & Piccolo, 2015). Several explanations have been provided for this phenomenon. For example, a firm restricted to an exclusive relationship faces additional costs in the form of ex post rents (Sharpe, 1990) and it will be required to pay a lemon’s premium if it seeks cooperation with more lenders. This is because the information and data gathered by one bank are not easily accessible by other banks, which limits their ability for accurate evaluation and screening. Other explanations proposed in the literature include the use of multiple lending relationships as a signal of credit quality and the need to diversify and share the risks associated with risky firms with lower credit ratings by many banks (see, Farinha & Santos, 2002). The existence of these multiple bank relationships suggest that frequently firms prefer to substitute large loans from one bank with multiple smaller loans from many banks. In this context, the nonlinear pricing literature can provide valuable insights into the observed average interest rate differentials between large and small loans.
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
7
The positive interest rate differential between small (“Up to and including D 1 million”) and large (“Over D 1 million”) loans should be expected to reflect primarily differences in margins. This is because large loans involve higher monitoring costs for banks that compensate for any differences in underwriting costs per euro loaned. The latter should be expected to be higher for small loans. For example, Minnis (2011) notes that banks should be expected to demand more monitoring from large firms, mainly because they are associated with large loan sizes and additional opportunities for asset substitution. This extra monitoring effort suggests that banks do not consider large loans as carrying intrinsically less risk. In a recent study, Minnis and Sutherland (2015) investigated the factors that influence the use of financial statements and tax returns as monitoring devices for commercial loans. The authors found a positive and statistically significant relationship between loan sizes and the collection of financial statements and tax returns for monitoring purposes. Therefore, higher loan amounts increase the probability of collection, which suggests higher monitoring costs. Apart from higher monitoring costs, large loans are also associated with lower opportunities for diversification and therefore higher risk. This is an additional cost associated with large loans that can increase the provisions for bad loans in a bank’s balance sheet. In an empirical study using data from the Austrian central bank, Rossi, Schwaiger, and Winkler (2009) examined how loan portfolio diversification across different industries and loan sizes affected the realized risk and the profit efficiency of banks in Austria during the period 1997–2003. In the econometric analysis, higher diversification across different industries and loan sizes was found to reduce the risk (provision for bad loans) and increase the profits (risk-adjusted returns) of loan portfolios. Consequently, large loans entail higher monitoring costs and lower opportunities for diversification, which should be expected to offset any underwriting costs differentials between small and large loans1 . In the context of this study, large loans that exceed the D 1 million benchmark are on average associated with lower interest rates, which suggests the existence of nonlinear pricing. It is therefore of interest to examine the factors influencing the level of these spreads, particularly focusing on changes in the market structure of each country, as suggested in the literature. As already noted in Section 2, a necessary condition for price discrimination is the existence of market segmentation, each with a different price elasticity of demand. Differences in interest-rate elasticties between different classes of borrowers (e.g. executives vs. blue collars) have been reported in the literature by Gary-Bobo and Larribeau (2004). Using a model that simultaneously explains the size of a loan and its associated interest rate, the authors found evidence of (first-degree) monopolistic price discrimination in French mortgage data. Blue-collar workers were found to pay higher interest rates than executives and the estimated price mark-ups were too high to reflect default risks only. The econometric results in this study suggest that the borrowers’ price-elasticity of demand for housing in France varies with occupational status, and is inversely related with the lender’s interest rate mark-ups. Furthermore, the loans provided to executives were, on average, 40% higher than the loans provided to blue collar workers. Differences in (quantity-dependent) loan interest rates between large and small borrowers were also discussed in a study by Schreft and Villamil (1992). The authors examined data from the “Survey of Terms of Bank Lending” in the US and found evidence of an inverse relationship between average interest rates and commercial loan sizes. According to the model presented by the authors, a profit maximizing lender with market power and imperfect information about borrowers’ characteristics can use second degree price discrimination as a form of credit rationing that limits borrowing. However, credit rationing is not applied to the largest borrowers. More importantly, the lender has an incentive to offer quantity-dependent loan interest rates even if credit risk and the associated fixed underwriting costs are the same for all classes of borrowers.
1 The degree of offset between the fixed underwriting costs of loans versus the size-related monitoring and diversification costs is a worthwhile topic for future research, but requires the availability of more detailed datasets on loans contracts and their restrictive covenants. Research in this area can provide valuable insights into the reasons for the observed interest rate differentials between small and large loans.
8
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
4. Econometric analysis 4.1. Model In this section, a panel data econometric model is used to examine the determinants of interest rate spreads, with particular emphasis on market concentration. To measure the difference between interest rates on loans below or equal to the D 1 million (pS ) benchmark and interest rates on loans above the D 1 million (pL ) benchmark, the following simple metrics were defined for the interest rate spread: S a = ps − pL
(1)
S b = (ps − pL )2 .
(2)
Both metrics are used as dependent variables in econometric models. The results for the model with the squared metric as the dependend variable are presented in Appendix B. The econometric model used in the analysis has the following form: Sita,b = ˇ0 +
6
ıCi Ci +
2
i=1
ıM Mj + ˇ1 log EBFt + ˇ2 INFit + ˇ3 log HHIit + ˇ4 log Yit j
j=1
+ ˇ5 log T + εit . where Ci represents six country indicator variables, except for the Netherlands, which is represented by the intercept ˇ0 ; Mj represents 2 indicator variables for the different loan maturities in the dataset, “Up to 1 year” and “Over 1 and up to 5 years”. Loans with maturity of “Over 5 years” are captured by the intercept. The remaining variables are defined as follows: EBFt is the Euribor rate based on which banks in the Eurozone exchange funds; INFit is the annual inflation rate of country i in year t, calculated from the harmonized index of consumer prices; HHIit represents the Herfindahl–Hirschman index of market concentration; and Yit is the national income of country i in year t. The model also includes a time trend variable (T ) and the error term εit . An alternative model specification that included time-fixed effects in place of the trend variable was also examined. The two model specifications generated similar results; however, the specification with the trend variable was preferred on the basis that it involved a smaller number of coefficients to be estimated and therefore a higher number of degrees of freedom. 4.2. Results The model was estimated with the GMM method, and the coefficient estimates along with their robust standard errors are included in Table 3. The standard errors were estimated with the heteroskedasticity and autocorrelation robust covariance matrix estimator proposed by Andrews (1991). The model provided a relatively good fit to the data and all of the estimated coefficients were statistically significant at the 1%, 5% or 10% level except in the cases of income, trend, Austria and Finland. The negative coefficient for the Euribor variable suggests that increases in the euro interbank offered rate tend to be associated with reductions in the interest rate spreads. The Euribor rate constitutes a basic input cost for banks and increases in this cost should be expected to reduce the size of interest rate reductions provided on large loans. To see this, consider a market with two segments, one for small loans and one for large loans. The interest rate spread is calculated as the difference in interest rates between small and large loans, due to the existence of a nonlinear pricing schedule. Therefore, the marginal price for small loans is ps (q, .), the marginal price for large loans is pL (q, .) = ps (q, .) − d (q, .) and the interest rate spread d(q, .) is a function of the loan amount. In turn, the loan amount demanded q(ie , .) is a function of the Euribor rate (ie ). By straightforward application of the chain-rule, the partial derivative of the interest rate spread with respect to the Euribor rate is ∂d/∂ie = ∂d/∂q · ∂q/∂ie . In the presence of a nonlinear pricing
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
9
Table 3 GMM estimation results. Variable
Coefficient
St error
Intercept log Euribor Inflation log HHI log Income log Trend Up to 1 year 1 to 5 years Austria Germany Spain Finland France Italy R-squared
2.589 −0.280 −0.072 0.553 −0.067 −0.021 0.254 0.508 0.467 1.181 1.497 −0.100 0.755 1.615 0.591
0.845* 0.060* 0.034** 0.253** 0.047 0.059 0.110** 0.112* 0.376 0.577** 0.400* 0.182 0.318** 0.521*
Significance level: * 1%; ** 5%; *** 10%.
schedule, the partial derivative ∂d/∂q should be expected to be positive. This is because an increase in the loan amount demanded, will tend to increase the reduction (spread) in the interest rate associated with large loans. In contrast, the partial derivative ∂q/∂ie should be expected to be negative. This is because increases in the Euribor rate will, tend to increase the interest rate on large loans and therefore reduce the loan amounts demanded. Consequently, the partial derivative ∂d/∂ie should be expected to be negative, which is consistent with the sign of the Euribor coefficient in Table 3. The coefficient for the inflation rate is also negative and statistically significant at the 5% level. Therefore, an increase in the inflation rate tends to be associated with a reduction in the size of the spreads. This trend results because increased inflation expectations tend to reduce the value of all future loan payments by households and enterprises and therefore raise the cost of lending for banks. As a result, banks working with lower profit margins are less willing to provide interest rate reductions on large loans. In the case of the Herfindahl–Hirschman Index variable, the coefficient is positive and statistically significant at the 5% level, which suggests that higher market concentration (increasing monopoly power) tends to be associated with higher interest rate spreads. This result is consistent with the literature presented in Section 2, according to which nonlinear pricing can be associated with a monopolistic market structure and the extraction of consumer surplus. With regard to the country indicator variables, it is worth noting that the average difference between interest rates on small loans and interest rates on large loans is higher in Italy, Spain and Germany. This is after controlling for changes in the other explanatory variables. In contrast, it is lower in Finland and the Netherlands. This finding suggests that increases in the actual spreads of these two countries are mainly motivated by changes in the other explanatory variables, like for example increases in market concentration and the exercise of nonlinear pricing tactics. As demonstrated in Fig. 1, Finland and the Netherlands are associated with higher concentration levels than the other countries.
5. Conclusions The econometric results of the previous section are consistent with the use of nonlinear pricing in monopolistic markets, which enables firms with market power to extract additional consumer surplus and increase their revenues. Nonlinear pricing is not necessarily harmful for welfare, even though it can be associated with a monopolistic market structure. Schmalensee (1981b) and Varian (1985) showed that total welfare can increase with nonlinear pricing, provided that the total quantity produced increases too.
10
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
However, an important distinction holds. Despite the increase in total welfare, consumer welfare in particular decreases. On the other hand, more consumers are served under nonlinear pricing, which (from the scope of financial intermediation) is important for entrepreneurship and project financing in an economy. According to Cabral (2000, p. 181), price discrimination should not be of concern, as long as there are no distribution issues between firms and consumers. It is therefore important for the monetary authorities to ensure that liquidity and project financing in the market for loans is equally accessible to the whole distribution of consumers and firms in an economy. European banking has gone through considerable changes during the last twenty years, following rapid development in deregulation, financial innovation and the process of globalization. These changes have intensified competition in financial markets and have led to a shrinkage of profit margins for banks. Mishkin, Matthews, and Giuliodori (2013, p. 256–257) note that in response to these changes the banking industry entered a phase of consolidation with strategic mergers and acquisitions. As a result of this consolidation process the total number of banking institutions in the Eurozone dropped from 11877 in 1988 to just 6842 in 2008. The formation of large banking organizations through consolidation is frequently associated with economies of scale and scope, which can increase efficiency in the industry. An additional benefit refers to the opportunity for improving diversification in loan portfolios that can contribute towards a less vulnerable banking system. However, bank consolidation is also associated with two important considerations. First, it can drive many small banks out of the market and therefore reduce lending to small businesses. Second, consolidation can increase market concentration and reduce competition in the industry with a few very large banking organizations dominating the industry. According to Mishkin et al. (2013, p. 266), such banking markets exist in Europe within national boundaries. In a concentrated market where nonlinear pricing is used by banks, distribution issues between banks and consumers might arise in the form of reduced lending towards small businesses or difficulties in financing small scale projects. In such an environment, the gains for the banks might not compensate for the reduction in consumer welfare. This possibility suggests that a worthwhile goal for public policy should be to ensure that small businesses will not face any reduction in lending and liquidity and sufficient financing for small-scale projects will continue to exist. Appendix A. Consider a block-declining tariff (a standard form of nonlinear pricing), which consists of two pricing segments separated by a quantity threshold X* . Following DeSalvo and Huq (2002), this tariff can be represented as follows:
C(X) =
P1 X,
0 ≤ X < X∗
(P1 − P2 ) X ∗ + P2 X,
X ≥ X∗
For loan amounts up to X* units (small loans) the borrower has to pay an amount equal to C(X) = P1 X, while for loan amounts higher than X* (large loans) the expenses are equal to C(X) = (P1 − P2 ) X ∗ + P2 X. Over the first tariff segment the average interest rate for small loans is constant and equal to p¯ s = P1 . In contrast, it becomes variable over the second segment and equal to p¯ L = P2 + (P1 − P2 ) X ∗ /X. The interest rate spread (d) is calculated as the difference in interest rates between small and large loans, due to the existence of a nonlinear pricing schedule in the market. Average values of the interest rate spread variable can be obtained by considering the difference between the two average interest rates as follows:
X∗ d¯ = P1 − P2 + (P1 − P2 ) X
= (P1 − P2 ) 1 −
X∗ X
If P1 > P2 and in the second pricing segment X > X* , then in a price discriminating monopoly the average interest rate spread takes positive values. Consequently, regressing this average interest rate spread variable on market concentration should be expected to provide a positive coefficient
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
11
(everything else being equal), since increasing monopoly power (higher market concentration) should be expected to be associated with increases in the average interest rate spread variable. Appendix B. The econometric model presented in Section 4 was also estimated using the squared interest rate spread-metric Sb as a dependent variable. The results are similar to those in Table 3 for the simple metric and the coefficients have the same sign in all cases except for the trend regressor. B.1. GMM estimation results with squared metric. Variable
Coefficient
St error
Intercept log Euribor Inflation log HHI log Income log Trend Up to 1 year 1 to 5 years Austria Germany Spain Finland France Italy R-squared
4.479 −0.648 −0.197 1.391 −0.096 0.065 0.187 0.955 1.635 3.058 3.806 −0.376 1.991 3.903 0.490
2.170** 0.147* 0.116*** 0.708** 0.101 0.144 0.255 0.298* 1.039 1.606*** 1.137* 0.378 0.842** 1.540**
Significance level: * 1%; ** 5%; *** 10%.
References Alegria, C., & Schaeck, K. (2008). On measuring concentration in banking systems. Finance Research Letters, 5, 59–67. Andrews, D. W. K. (1991). Heteroskedasticity and autocorrelation consistent covariance matrix estimation. Econometrica, 59(3), 817–858. Armstrong, M., & Vickers, V. (2001). Competitive price discrimination. RAND Journal of Economics, 32(4), 579–605. Berkovec, J. A., Canner, G. B., Stuard, G. A., & Hannan, T. H. (1998). Discrimination, competition, and loan performance in FHA mortgage lending. Review of Economics and Statistics, 80(2), 241–250. Bennardo, A., Pagano, M., & Piccolo, A. (2015). Multiple bank lending, creditor rights, and information sharing. Review of Finance, 19, 519–570. Bikker, J. A., & Haaf, K. (2002). Competition, concentration and their relationship: An empirical analysis of the banking industry. Journal of Banking & Finance, 26, 2191–2214. Borenstein, S. (1991). Selling costs and switching costs: Explaining retail gasoline margins. RAND Journal of Economics, 22, 354–369. Borenstein, S., & Rose, N. (1994). Competition and price dispersion in the U.S. airline industry. Journal of Political Economy, 102, 653–683. Busse, M., & Rysman, M. (2005). Competition and price discrimination in Yellow Pages advertising. RAND Journal of Economics, 36, 378–390. Cabral, L. M. B. (2000). Introduction to industrial organization. Cambridge, MA: MIT Press. Carbo, S., Humphrey, D. B., Maudos, J., & Molyneux, P. (2009). Cross-country comparisons of competition and pricing power in European banking. Journal of International Money and Finance, 28, 115–134. Cavalluzzo, K. S., & Cavalluzzo, L. C. (1998). Market structure and discrimination: The case of small businesses. Journal of Money, Credit and Banking, 30(4), 771–792. Claessens, S., & Laeven, L. (2004). What drives bank competition? Some international evidence. Journal of Money, Credit and Banking, 36, 563–583. Clerides, S. (2004). Price discrimination with differentiated products: definitions and identification. Economic Inquiry, 42, 402–412. Clerides, S., & Michis, A. (2006). Market concentration and nonlinear pricing: Evidence from detergents prices in six countries. In Working paper. University of Cyprus. Coccorese, P. (2005). Competition in markets with dominant firms: A note on the evidence from the Italian banking industry. Journal of Banking & Finance, 29, 1083–1093. Courty, P., & Pagliero, M. (2012). The impact of price discrimination on revenue: Evidence form the concert industry. Review of Economic and Statistics, 94, 359–369.
12
A.A. Michis / Journal of Economics and Business 85 (2016) 1–12
Dai, M., Liu, Q., & Serfes, K. (2014). Is the effect of competition on price dispersion non-monotonic? Evidence from the U.S. airline industry. Review of Economics and Statistics, 96, 161–170. Degryse, H., & Ongena, S. (2005). Distance, lending relationships, and competition. Journal of Finance, 60(1), 231–266. DeSalvo, J. S., & Huq, M. (2002). Introducing nonlinear pricing into consumer theory. Journal of Economic Education, 33, 166–179. Dziuda, W., & Mastrobuoni, G. (2009). The euro changeover and its effects on price transparency and inflation. Journal of Money, Credit and Banking, 41, 101–129. Farinha, L. A., & Santos, J. A. C. (2002). Switching from single to multiple bank lending relationships: Determinants and implications. Journal of Financial Intermediation, 11, 124–151. Gaggero, A., & Piga, C. A. (2011). Airline market power and intertemporal price dispersion. Journal of Industrial Economics, 59, 552–577. Gary-Bobo, R. J., & Larribeau, S. (2004). A structural econometric model of price discrimination in the French mortgage lending industry. International Journal of Industrial Organization, 22(1), 101–134. Gerardi, K. S., & Shapiro, A. H. (2009). Does competition reduce price dispersion? New evidence from the airline industry. Journal of Political Economy, 117, 1–37. Hannan, T. H., Kiser, E. K., Prager, R. A., & McAndrews, J. J. (2003). To surcharge or not to surcharge: An empirical investigation of ATM pricing. Review of Economics and Statistics, 85(4), 990–1002. Leslie, P. (2004). Price discrimination in Broadway theatre. RAND Journal of Economics, 35, 520–541. Lipczynski, J., Wilson, J. O. S., & Goddard, J. (2013). Industrial organization: Competition, strategy and policy (4th ed.). Harlow: Pearson. Minnis, M. (2011). The value of financial statement verification in debt financing: evidence from private U.S. firms. Journal of Accounting Research, 49, 457–506. Minnis, M., & Sutherland, A. (2015). Financial statements as monitoring mechanisms: Evidence from small commercial loans. In Working paper. University of Chicago Booth School of Business. Mishkin, F. S., Matthews, K., & Giuliodori, M. (2013). The economics of money banking and financial markets (European ed.). Harlow: Pearson Education. Murphy, M. M., & Ott, M. (1977). Retail credit, credit cards and price discrimination. Southern Economic Journal, 43(3), 1303–1312. Oi, W. Y. (1971). A Disneyland dilemma: Two-part tariffs for a Mickey Mouse monopoly. Quarterly Journal of Economics, 85(1), 77–96. Pepall, L., Richards, D., & Norman, G. (2008). Industrial organization: Contemporary theory and empirical applications. Cornwall: Blackwell Publishing. Rochet, J.-C., & Stole, L. A. (2002). Nonlinear pricing with random participation. Review of Economic Studies, 69(1), 277–311. Rossi, S. P. S., Schwaiger, M. S., & Winkler, G. (2009). How loan portfolio diversification affects risk, efficiency and capitalization: A managerial behavior model for Austrian banks. Journal of Banking & Finance, 33, 2218–2226. Schmalensee, R. (1981a). Monopolistic two-part pricing arrangements. Bell Journal of Economics, 12(2), 445–466. Schmalensee, R. (1981b). Output and welfare implications of monopolistic third-degree price discrimination. American Economic Review, 71(1), 242–247. Schreft, S. L., & Villamil, A. P. (1992). Credit rationaing by loan size in commercial markets. Federal Reserve Bank of Richmond Economic Review, (May–June), 1–8. Sharpe, S. A. (1990). Asymmetric Information, bank lending and implicit contracts: A stylized model of customer relationships. Journal of Finance, 45, 1069–1087. Stavins, J. (2001). Price discrimination in the airline market: The effect of market concentration. Review of Economics and Statistics, 83, 200–202. Stole, L. A. (1995). Nonlinear pricing and oligopoly. Journal of Economics and Management Strategy, 4(4), 529–562. Varian, H. R. (1985). Price discrimination and social welfare. American Economic Review, 75(4), 870–875. Verboven, F. (1999). Product line rivalry and market segmentation, with an application to the pricing of optional engine power on automobiles. Journal of Industrial Economics, 47, 399–425. Verboven, F. (2008). Price discrimination: Empirical studies. In N. Stephen, Durlauf, E. Lawrence, & Blume (Eds.), The new Palgrave: A dictionary in economics. Hampshire: Palgrave MacMillan.