Bank market power and interest rate setting: why consolidated banking data matter

AbstractThe literature on the effects of bank market power on access to credit has produced many results that are sometimes contradictory. Yet, all of these studies are based on unconsolidated data that ignore the national market power of banking groups. This results in an underestimation bias that this paper proposes to correct. Using a panel of more than 55,000 French firms covering the period 2006–2017, I consider a set of structural and non-structural measures of bank market power both at the unconsolidated and consolidated levels. My results strongly support the market power hypothesis which emphasizes the virtues of competition on interest rate setting. I find that bank market power increases the interest rate charged, but only when using my consolidated measures. This effect is stronger for small and risky firms and is concentrated on long-term loans. My findings highlight the need to take into account the capital linkages of subsidiaries within the same banking group in order to fully assess the implications of bank market power.Keywords: Cost of creditbank concentrationbank competitionrelationship lendingJEL codes: E43E51G01G21 AcknowledgmentsThe views expressed in this paper are those of the author and do not necessarily coincide with those of the ACPR or the Eurosystem. This work has benefited from insightful comments by participants in the ACPR research seminar. I am particularly grateful to Laurent Weill, Boubacar Camara, Jézabel Couppey-Soubeyran and Cyril Pouvelle who provided precious feedbacks at various stages of the project.Disclosure statementNo potential conflict of interest was reported by the author(s).Notes1 These banking groups are BNP Paribas, BPCE Group, the Crédit Agricole Group, Crédit Mutuel-CIC Group, Société Générale and Banque Postale.2 See Section 2 for a detailed discussion of this literature.3 Note that the dataset is composed of 15% of observations coming from industry, 14% from construction, 50% from trade, 16% from services and 5% from other sectors.4 In this regard, 80% of firms in the database can be considered as SMEs with respect to the European definition based on the number of employees (less than 250), the turnover (less than EUR 50 million) and total assets (less than EUR 43 million).5 To minimize the effect of gross outliers, I winsorize variables at the first and 99th percentiles.6 Note that the banking groups are identified using the ACPR database 'Groupes Economiques d'appartenance' (GEA) wich takes into account the subsidiaries claim by the head of the banking group. In addition to the five main national private banking groups (BNP Paribas, Société Générale, Crédit Agricole, Crédit Mutuel and BPCE), the other groups are foreign banking groups. Note that the GEA database is a time varying database that tracks the composition of all French banking groups over time to take into account mergers an acquisitions issues.7 To do so, before summing up the total assets of all banks belonging to the same banking group, I have removed the intra-group debts and receivables from the total assets of each banks.8 Note that as there is no information on intra-group incomes or expenses in the SURFI database, I assume that incomes and taxes are computed at the unconsolidated level.9 In France, the second largest banking group (BPCE) was created in 2009.10 Note that, contrary to the use of the within-firm estimator in the seminal work of Khwaja and Mian (Citation2008), my fixed effects methodology does not control for all observed and unobserved time-varying firm heterogeneity.11 As the FiBEN and SURFI databases provide yearly and quarterly information, respectively, note that firm variables are lagged by one year whereas bank variables are lagged by one quarter.12 Note that loans with maturities of over 1 year are equipment loans and leasing.13 See Beatriz, Coffinet, and Nicolas (Citation2018) for more justifications of this time interval for the French case.14 Results are not presented but available upon request.15 Note that Crédit du nord is a regional entity whose ultimate owner is 'Société Générale'.16 Note that France is divided into 13 regions representing 96 'departements'.17 Results are not presented but are available upon request.18 Results are not presented but available upon request.19 In this regard, the Hansen J-statistic implies that we cannot reject the null hypothesis that all instruments are exogenous, while the Kleibergen-Paap statistic indicates that our instruments are relevant. Besides, implementing the tests recommended by Baum, Schaffer, and Stillman (Citation2007), I can also reject the null hypothesis of weak instruments when I do not accept an actual test size above 10% (results available upon request).Additional informationNotes on contributorsThéo NicolasThéo Nicolas is a research economist at the Directorate for Research and Risk Analysis of the Prudential Supervision and Resolution Authority (The French banking supervisor), an institution attached to the Banque de France. He hold a Ph.D. in Economics from the Paris School of Economics and his main areas of interest are banking, financial regulation, corporate finance and applied Econometrics.