游戏测试？欧盟范围内压力测试的窗口装饰和投资组合相似性

Gaming the Test? Window-dressingand portfolio similarity around theEU-wide stress tests Angelo Cuzzola, Claudio Barbieri,Grzegorz Hałaj Disclaimer:Thispaper should not be reported asrepresenting the views of the European Central Bank(ECB). The views expressed are those of the authors and do not necessarily reflect those of the ECB. Abstract This study investigates the impact of supervisory stress testing on banks’ behaviors and theirsystemic risk implications. Utilizing confidential supervisory data from the European BankingAuthority’s EU-wide stress tests in 2021 and 2023, we employ a difference-in-differences frame-work to analyze how these exercises influence portfolio management decisions among Europeanbanks. This methodology allows us to compare stress-tested banks with similar non-tested in-stitutions before and after the stress test events, isolating the effects specifically associated withthe EU-wide assessments. Our findings reveal significant patterns of anticipatory behavior, with banks strategicallywindow-dressing their capital ratios before stress tests begin.This behavior is particularlypronounced among institutions that subsequently receive the lowest scores in terms of capitaldepletion. We document that these anticipatory adjustments lead to decreased portfolio simi-larity across banks, an effect that persists after the stress tests and remains consistent acrossdifferent similarity measures. Importantly, such a decrease in similarity does not spin off intomore granular business model or country clusters, thus limiting potential systemic risk throughportfolio synchronization. Our results, while considering how financial institutions incorporate stress test considerationsinto their strategic decision-making, highlight the dual role of stress tests in enhancing individualbank resilience and reducing systemic vulnerabilities. These findings contribute to the ongoingdebate on effective banking supervision and the design of regulatory stress testing frameworks.Keywords:Stress testing, Banking supervision, Portfolio similarity, Systemic risk, Windowdressing, Financial stability JEL classification: G21, G28, E58, C23 Non-technical summary In the aftermath of the Global Financial Crisis, stress testing has emerged as a crucial tool forbanking supervisors to assess financial institutions’ resilience to severe macro-financial shocks.As these stress test exercises have gained prominence in the supervisory toolkit, they have begunto influence bank behavior in ways that extend beyond their primary purpose to assess andimprove the resilience of single banks and the banking sector as a whole. Our paper examineshow banks’ strategic responses to stress tests affect both individual risk profiles and broaderfinancial stability. To this end, we assess the impact of the EU-wide stress test — conducted jointly by the Eu-ropean Banking Authority and the ECB Banking Supervision (SSM) — on banks’ risk-takingbehavior and systemic risk combining three main data sources: detailed balance sheet informa-tion from FINREP, including granular data on regional exposures and asset classes; loan-levelinformation from AnaCredit; and confidential supervisory datasets containing detailed stress testsubmissions, intermediate and final results, and documentation of bank-supervisor interactionsduring the quality assurance process. We use this rich dataset to construct precise measures ofportfolio similarities and analyze systematically how stress test influence portfolio managementdecisions around stress tests, including ex ante positioning to mitigate the impact of forthcomingstress scenarios or ex post adjustments following the interaction with supervisors or publicationof the stress test results. Our analysis highlights three key findings. First, we document significant increase in the CET1 capital ratios primarily occurringthrough ex-ante window-dressing, with banks managing risk-weighted assets rather than capitallevels.This anticipatory behavior is particularly pronounced among institutions that subse-quently receive low scores in the exercises, suggesting that poorly performing banks engage inmore aggressive balance sheet adjustments in preparation of the stress test. Second, we find that these risk-mitigating actions lead to a decline in portfolio similarity,contributing positively to systemic risk reduction. The management actions by vulnerable banksprove beneficial for financial stability also through the lens of liquidity risk, reducing the riskof shock amplification through coordinated fire sales.Importantly, we find no evidence thatsupervisory follow-up creates conditions for banks to converge in the composition of their balancesheets.This result holds consistently when measuring similarity using portfolio shares across several dimensions, encompassing sector, region, counterparty, and type of instrument. Third, we find a decrease in similarity among banks with the same business model but thereis no signif