Mandatory supervisory disclosure, voluntary disclosure, and stock market liquidity: Evidence from the EU-wide stress-testing exercises

Transcription

1 Mandatory supervisory disclosure, voluntary disclosure, and stock market liquidity: Evidence from the EU-wide stress-testing exercises Jannis Bischof / Holger Daske University of Mannheim Bank Liquidity, Regulation, and Transparancy Banque de France & Toulouse School of Economics December 20, 2013 Chair of Accounting & Capital Markets University of Mannheim Mannheim, Germany Tel.: +49 (0) daske@bwl.uni-mannheim.de 1 / 34

2 2 / 34

3 Motivation Regulatory debate Federal Reserve and Bank of England in support of disclosure One reason for the success of the stress-test was the public disclosure of the results. - Bernanke, Chairman Federal Reserve System (2010) Many regulators from Continental Europe oppose disclosure (e.g., Germany) One EU official said that the decision to publish had been reached only after a big battle at the summit. - Financial Times, June 18, / 34

4 Motivation Academic debate Disciplining effects of disclosure (e.g., Flannery [2001], Bushman and Williams [2011]) versus investment inefficiencies (e.g., Goldstein and Sapra [2012]) Role of mandatory disclosure (e.g., Einhorn [2005], Leuz and Wysocki [2008]) 4 / 34

5 Research Question Immediate capital market effects of mandatory stress-test disclosures versus Incentive for change in banks behavior (feedback) Change in risk-taking strategy Change in voluntary risk reporting 5 / 34

6 Research Question DOI: / X Journal of Accounting Research Vol. 51 No. 5 December 2013 Printed in U.S.A. Jannis Bischof & Holger Daske: Mandatory Disclosure, Voluntary Disclosure, and Stock Market Liquidity: Evidence from the EU Bank Stress Tests Journal of Accounting Research, Vol. 51, No. 5, pp Mandatory Disclosure, Voluntary Disclosure, and Stock Market Liquidity: Evidence from the EU Bank Stress Tests JANNIS BISCHOF AND HOLGER DASKE Received 15 August 2012; accepted 9 September 2013 ABSTRACT We use the EU stress tests and the Eurozone sovereign debt crisis to study the consequences of supervisory disclosure of banks sovereign risk exposures. We test the idea that a mandatory one-time disclosure induces an increase in voluntary disclosures about sovereign risk in the following periods and, through the shift in the voluntary disclosure equilibrium, increases the University of Mannheim. Accepted by Douglas Skinner. We wish to thank Hamid Mehran and Abbie Smith (the conference editors), an anonymous referee, Mary Barth, Phil Berger, Ulf Brüggemann, Eti Einhorn, Miles Gietzmann, Martin Glaum, Luzi Hail, Mirko Heinle, Luc Laeven (discussant), Konrad Lang, Christian Leuz, Martien Lubberink, Scott Richardson, Johannes Voget, Alfred Wagenhofer, Regina Wittenberg-Moerman, and workshop participants at the JAR/NY Fed Conferences in Chicago and New York City, Workshop on Accounting and Economics in Segovia, Annual Meeting of the Accounting Research Committee of the Verein fuer Socialpolitik in Bayreuth, Deutsche Bundesbank, DFG Priority Program Conference in Mannheim, 2013 EAA Annual Congress in Paris, IAAER/AS-VHB Conference in Frankfurt, Goethe University Frankfurt, Frankfurt School of Finance and Management, University of Mannheim, and University of Tübingen for helpful comments on previous versions of the paper. The research project is part of the Priority Program 1578 Financial Market Imperfections and Macroeconomic Performance by the German Research Foundation ( Part of the research was carried out while Jannis Bischof was visiting Chicago Booth School of Business and Holger Daske was visiting London Business School and the University of Sydney. Jannis Bischof gratefully acknowledges the financial support from the German Academic Exchange Service (DAAD Postdoc Program) and the J.P. Stiegler Foundation. 997 Copyright C, University of Chicago on behalf of the Accounting Research Center, / 34

7 Setting EU stress tests (July 2010, July 2011, December 2011) are a pure disclosure event with regulatory consequences being left to local authorities (in contrast to SCAP / CCAR) No regulatory commitment to repeated disclosures The EBA s charter requires... stress tests... on a regular basis, but doesn t stipulate that they have to be every year. It all depends on the situation when they are appropriate, the spokeswoman said. 7 / 34

8 Setting Types of disclosures in the EU stress tests: (1) Forward-looking regulatory capital simulation under adverse economic conditions (conventional stress-test information) (2) Backward-looking exposure to Eurozone-specific sovereign credit risk (at a specific date in the past) Level of detail of mandatory sovereign risk disclosures increases over time The content of the supervisory disclosure is not required in, e.g., IFRS 7 or Basel s Pillar 3 reports 8 / 34

9 Setting Disclosure items over time July 2010 July 2011 Dec 2011 Sovereign Exposure Total EAD x x x Trading Book x x x Banking Book x x x FV Option x x AFS x x Derivatives x x Maturity Analysis x x CDS x Non-sovereign Exposure Total EAD x 9 / 34

10 Setting: July 2010 Disclosures 10 / 34

11 Setting: December 2011 Disclosures 11 / 34

12 Hypothesis (1): Feedback on Voluntary Risk Disclosure Costs of non-disclosure vs. cost of disclosure Trade-off results in firm-specific disclosure threshold (e.g., Verrecchia [1983]) If costs of first-time disclosure > costs of subsequent disclosure (Einhorn and Ziv [2008]), we should observe a change in voluntary disclosure of stress-tested firms Within the group of tested firms, disclosure choice is associated with type of news 12 / 34

13 Hypothesis (2): Effect on bank opaqueness & liquidity Disclosure theory only predicts a decrease in information asymmetry if a firm credibly commits to maintaining an increased level of transparency (Diamond and Verrecchia [1991], Baiman and Verrecchia [1996]) A one-time disclosure shock without any commitment to subsequent disclosures is unlikely to have a persistent impact on market liquidity We predict an indirect effect of the supervisory action in case of a change in a firm s voluntary credit risk disclosure practice 13 / 34

14 Sample and Data Sample of 273 financial institutions (82 stress-test participants, two control groups) Disclosure data Screening of all IFRS annual, interim and supervisory Pillar 3 reports for sovereign credit risk disclosures (2,047 reports) Content analysis 14 / 34

15 Univariate Results Univariate analysis: Reporting period immediately before versus reporting period immediately after the specific mandatory stress-test disclosures versus reporting period after the reporting period immediately after the stress-test disclosures (= control for general time trend) 15 / 34

16 Sovereign Risk Disclosures around Stress-Testing Exercises July 2010 Item (Aggregate Sovereign EAD) 16 / 34

17 Sovereign Risk Disclosures around Stress-Testing Exercises July 2011 Item (Sovereign EAD Details) 17 / 34

18 Sovereign Risk Disclosures around Stress-Testing Exercises December 2011 Item (CDS) 18 / 34

19 Sovereign Risk Disclosures around Stress-Testing Exercises Survival analysis methodology Discrete-time logistic hazard models Permits the implementation of a difference-in-differences design (e.g. Clotfelter et al. [2008]) that tests for the likelihood of a change in the voluntary reporting of the specific stress-test items during the reporting period immediately after the mandatory disclosures. 19 / 34

20 Sovereign Risk Disclosures around Stress-Testing Exercises The probability of a firm providing the initial disclosure of the specific item is given by: P(Disclosure sti = 1 X ti ) = F (α t + βx sti ) Underlying latent variable model is defined by equations (2): Disclosure sti = α t + β 1 Stress-Test Participations i + β 2 Post-Stress Test) st Stress-Test Participations i + β j Control j + ɛ where Disclosure sti = 1 if Disclosure sti > 0; 0 otherwise 20 / 34

21 Sovereign Risk Disclosures around Stress-Testing Exercises Multivariate Analysis: July 2010 Item (Sovereign EAD) 21 / 34

22 Sovereign Risk Disclosures around Stress-Testing Exercises Multivariate Analysis: July 2011 Item (Sovereign EAD Details) 22 / 34

23 Sovereign Risk Disclosures around Stress-Testing Exercises Multivariate Analysis: December 2011 Item (CDS) 23 / 34

24 Research Design Multivariate Analysis: Why do some stress-test participants not include the sovereign risk disclosures immediately in their risk reporting? Cross-sectional analysis of participants timing of initial voluntary disclosure Cox model Time until Disclosure it = β 0 + β 1 Type of News i + β j Control j + ɛ 24 / 34

25 Sovereign Risk Disclosures around Stress-Testing Exercises 25 / 34

26 Effect on Bank Opaqueness Bank opaqueness is approximated by the bid-ask spread (e.g., Flannery et al., 2004) Two quarters before and two quarters after 2010 and 2011 EBA stress test Difference-in-differences design (stress-test participants vs. two benchmark groups) Cross-sectional split to capture differences in the adaptation of voluntary reporting strategy (commitment to disclosure) 26 / 34

27 Research Design Multivariate analysis Log(Bid-Ask Spread) ti = α t + α i + β 1s Post-Stress Test Participation & Disclosure sti + β 2s Post-Stress Test Participation & No Disclosure sti + γ j Control j + ɛ 27 / 34

28 Effect on Bank Opaqueness Multivariate Analysis (Firm-Quarter Observations): 28 / 34

29 Risk-Taking Behavior around Stress-Testing Exercise Change in risk-taking strategy Highly standardized and detailed data on sovereign risk exposures for three different points in time We observe an economically huge reduction in sovereign exposure to PIIGS countries (-84.4bn Euros) between December 31, 2010 and September 30, 2011 (nine months) Reduction in exposure is associated with the disclosure of negative stress-test outcome However, identification of stress-test effect is less clear than in the disclosure setting (e.g., unobserved regulatory activity, no standardized data for control group) 29 / 34

30 Risk-Taking Behavior around Stress-Testing Exercise First approach Change in reported sovereign exposures between December 2010 and September 2011 (EBA templates) Use of simulation results to identify stress-test effect Low measurement error in exposures, however: no benchmark group of non-tested firms Second approach Change in market price sensitivities to changes in PIIGS CDS premia (10 months before and 10 months after submission of 2011 stress-test results) Data availability for benchmark groups, however: large measurement error 30 / 34

31 Risk-Taking Behavior around Stress-Testing Exercise Univariate Analysis of Reported Sovereign Exposures: 31 / 34

32 Risk-Taking Behavior around Stress-Testing Exercise Multivariate Analysis of Reported Sovereign Exposures: 32 / 34

33 Risk-Taking Behavior around Stress-Testing Exercise Multivariate Analysis of Market Price Sensitivities: 33 / 34

34 End Thank you for your attention! 34 / 34