Bank Relationships and Private Firms’ Financial Reporting Quality

Abstract

Private firms with relatively high (proprietary) costs of disclosure may benefit from a close relationship with a bank. Relationship lending is based on intertemporal contracting that assumes that the bank is able to acquire private information about the firm and, moreover, to keep this information private. For both reasons, we expect and find that private firms with fewer bank relationships exhibit lower levels of financial reporting quality. Controlling for many other factors, firms with a single bank relationship disclose their financial reports about 14 days later. The size of such firms’ financial reports is also smaller, containing approximately 8% fewer words than the median report. Firms with a single bank relationship also exhibit more earnings management, exceeding the median value of the three-year sum of absolute discretionary accruals by about 20%. The results are robust to different econometric specifications, including endogeneity concerns. They indicate that private firms choose to be opaque in the presence of fewer lending relationships.

Acknowledgements

We gratefully acknowledge the helpful comments provided by an anonymous referee and the editor in charge, Florin Vasvari. We thank Igor Goncharov, Martin Wallmeier and seminar participants at the EAA conference, Glasgow 2015, the EUFIN conferences, Regensburg 2014 and Paris 2015, the VHB conference, Vienna 2015, and the universities of Bern, Fribourg, Hanover, Lausanne and Münster for their valuable comments. We also thank Ruben Lanzerath (Bundesverband der Deutschen Volksbanken und Raiffeisenbanken) and Stefan Marotzke (Deutscher Sparkassen- und Giroverband) for their support.

Notes

1 Other reasons may include moral hazard or legal interest rate ceilings that do not permit full or adequate pricing of the credit risk involved, or considerable transaction costs required to set up the relationship.

2 AMADEUS includes data on the existence of bank relationships for a few countries (e.g. Austria, Spain), but usually not on the number of relationships. More than 96% of all independent private firms in the DAFNE database provide this information. The quota is much lower in the AMADEUS database, raising self-selection concerns. With respect to our sample selection, 99.8% of the firms disclose at least one bank relationship. Only two out of 1068 firm-year observations do not provide data on the number of bank relationships. We do not know whether in those cases no bank relationship exists or whether the firms did not want to report on that.

3 Alternatively, venture capitalists are specialists in information gathering, but often ask for considerable control rights. If a firm wants to keep control rights, relationship lending may be the only way to gain external funding.

4 Private firms have been required since fiscal year 2006 to disclose their financial reports in the electronic Federal Gazette (Bundesanzeiger). Initially, private firms were very reluctant to do so. Thus, the regulator tightened enforcement, for example, by imposing penalties and by monitoring. If a firm fails to disclose its financial report, the Federal Ministry of Justice will be notified accordingly. Although enforcement seems to be effective, some private firms remain unwilling to disclose their financial reports. See Achleitner et al. (2011, p. 69), available online at: http://www.pwc.de/de_DE/de/mittelstand/assets/sfu_studie_kapitalmarkt_e4_zA.PDF.

5 Since the DAFNE database only contains current information on the number of ‘house banks’, we gathered historical yearly data – each by end of March – on the number of ‘house banks' from DAFNE's annual backup copy.

6 Total accruals are estimated according to Daske, Gebhardt, and McLeay (2006) and Dechow (1994) as follows: , where CA = current assets, CL = current liabilities, Dep = depreciation and Prov = provisions. Since the database DAFNE generally lacks data on short-term debt (STD) and tax payables (TP), we have to omit STD and TP.

7 The German Commercial Code allows for a subsequent filing within six weeks if a firm has already missed the mandatory deadline (Section 335 III HGB). If the firm does not react within this subsequent filing period, monetary penalties apply (Section 335 IV HGB).

8 Standardized regression coefficients are traditionally called beta coefficients and are calculated according to the following formula: .

9 In addition to the three approaches mentioned in the paper, we also perform a propensity score matching. We match firms that have a single bank relationship with other firms having multiple bank relationships but were otherwise similar with regard to size, capital intensity, losses in the past, profitability, risk, tax aggressiveness and the driving distance to the closest bank (methods: nearest neighbor (n= 10 and n =50), Gaussian and Epanechnikov). Qualitative results remain the same; that is, with a larger number of bank relationships, the timeliness and amount of disclosure increase, as does earnings quality.

10 We also try other instruments such as local bank status (Puri et al., 2013) and financial debt to total assets (to sales) (Detragiache, Garella, & Guiso, 2000; Ongena et al., 2012). The economic reasoning is as follows: the more financial debt a firm has, the more likely it is that it would need to borrow from several banks. Thus, we expect a negative association between financial debt to sales and the SINGLE_B variable. However, financial debt to sales is close to the definition of leverage or to other metrics of firm risk, such that this instrument might directly affect financial reporting quality. Thus, the validity of financial debt to sales is questionable. An alternative instrument might be local bank status. In smaller cities, savings banks and co-operative banks are often the only bank; as a result, it is with those types of banks that a firm is more likely to have a single bank relationship. We would expect a positive correlation between local bank status and the SINGLE_B variable. However, local banks can also demand specific disclosure attributes from their lenders, indicating a correlation between local bank status and financial reporting quality. Consequently, the validity of the instrument is unclear. Alternative specifications of the treatment effect model (not tabulated) suggest that financial debt to sales and local bank status are not instruments with reasonable validity. We thank the reviewer for helpful comments on the validity of the instrumental variables financial debt to sales and local bank status.

11 With robust standard errors, we cannot use the Likelihood Ratio statistics of independent equations.

12 From 2009 to 2012, there were mergers between 14 savings banks resulting in 7 merged savings banks and between 22 co-operative banks (resulting in 11 co-operative banks). However, with these kinds of mergers, bank employees are typically not fired. Only 10 of our sample firms were affected by these mergers. Therefore, we did not include these mergers in the analysis. If we were to do so, the qualitative results would remain the same.

13 In November 2005, Unicredit acquired HypoVereinsbank, that is, before the period of investigation.

14 See http://www.spiegel.de/wirtschaft/fusion-mit-commerzbank-der-name-dresdner-bank-soll-verschwinden-a-575717.html.

Additional information

Funding

We gratefully acknowledge research support by the Fritz Thyssen foundation (Fritz Thyssen Stiftung für Wissenschaftsförderung) [grant number Az.10.15.1.001WW].