Bank Of Greece Sa · ATHEX:TELL
Bank Of Greece Sa — Bank for International Settlements — Investor Presentation (March 2013)
38 slides from Bank Of Greece Sa's investor presentation, Mar 22, 2013
Slide 1: Characterising the financial cycle and implications for stress testing
Characterising the financial cycle and implications for stress testing Kostas Tsatsaronis* (joint work with M Drehmann and C Borio) Bank for International Settlements Seminar at the Bank of Greece Athens, 22 March 2013 * The views expressed here are my own and do not necessarily reflect those of the
Slide 2: Motivation and Objective
Motivation and Objective Time vs cross sectional dimensions of systemic risk Macroeconomics needs to understand the financial cycle Establish the stylised facts • Data groundwork, no theory A reasonable quantification of the financial cycle Messages for macro stress testing • Medium-term c
Slide 3: Characterising the financial cycle: overview
Characterising the financial cycle: overview Two very different approaches: A. Statistical frequency-based filtering (continuous) B. Algorithmic dating of peaks and troughs (discrete) Common messages from panel of countries and series Analysis of individual series : • What are their cyclical p
Slide 4: Related work
Related work Burns and Mitchell (1946) : pioneers of bottom-up cycle characterisation Mendoza & Terrones (2008), Claessens et al (2011) dating approach Aitken et al (2011) filtering approach Schularick and Taylor (2009) , Detken and Smets (2004) , Goodhart and Hofmann (2008) on interactions
Slide 5: Filtering approach
Filtering approach Band-pass filter (Christiano and Fitzgerald (2003)) • Annual growth rates (no drift) • Cumulated to create cycles in log-levels Decomposed in • Short term cycles: 5 to 32 quarters long • Medium term cycles: 32 to 120 quarters long 5
Slide 6: The ratio of medium to short cycle volatility
The ratio of medium to short cycle volatility Credit and house prices are more volatile over the medium term 6
Slide 7: The ratio of medium cycle volatility pre- & post-1985
The ratio of medium cycle volatility pre- & post-1985 Amplitude of financial variables increased post-liberalisation 7
Slide 8: Turning point dating approach
Turning point dating approach Bry-Boschan (1971) and Harding-Pagan (2002) algorithm Pick local minima and maxima • Window width • Censoring to guarantee minimum length of phase Make sure that min and max alternate Two parameterisations • Short term : min phase length 2 quarters • Medium term
Slide 9: GDP Credit
am BANK FOR INTERNATIONAL SETTLEMENTS GDP Credit | 1.09 —— log level || © Peak/Trough ST > Pea [be _}1.04 = NBER recessions , [; Mill Crises + 0.99 1 0.94 Ae A | ee DO gg, 61 66 71 76 81 86 91 96 O1 O6 11 61 66 71 76 81 86 91 96 01 O6 11
Slide 10: Property Equity
am BANK FOR INTERNATIONAL SETTLEMENTS Property 66 71 76 81 86 91 96 01 06 1.20 1.15 1.10 1.05 1.00 0.95 0.90 Equity 81 86 91 1.50 1.40 1.30 1.20 1.10 1.00 0.90 0.80 1 10
Slide 11: Turning point analysis
am BANK FOR INTERNATIONAL SETTLEMENTS Turning point analysis Amplitude” Duration Expansion Contraction Expansion Contraction Cycle® Medium-term cycles In per cent Number of quarters Credit 140 +7 53 8 72 Credit/GDP 31 -8 27 13 41 House prices 48 -16 19 15 42 Equity prices 121 —48 20 12 38 AAP 57 —26
Slide 12: Turning point analysis and liberalisation
Turning point analysis and liberalisation 12
Slide 13: Commonality in cycles of financial series
Commonality in cycles of financial series Closer correlation of medium-term cycles in: • House prices and credit (60%) or credit/GDP (42%) Lower correlations of equity prices with the rest Harding-Pagan Concordance measure • % of time two series are in same phase 13
Slide 14: Concordance measures
am BANK FOR INTERNATIONAL SETTLEMENTS ao Concordance measures Credit Credit/GDP House prices | Equity Crom’ | HP | EQ | AAP | HP | EQ | AAP | EQ | AAP | AAP AU 91 56 57 84 54 60 76 42 66 75 DE 89 59 49 62 63 42 54 29 58 67 GB 93 80 53 75 85 48 68 55 75 71 JP 77 86 80 81 77 59 76 78 96 82 NO 72 72 56
Slide 15: Combining the series and methods
Combining the series and methods Frequency filters: • Add the three medium term components of filtered series Dating points: • Create a continuous variables: the median distance of solo series from turning point • Combined peaks are local minima of median distance • Each series has peak within s
Slide 16: Combined series both approaches
Combined series both approaches 16
Slide 17: Combined series both approaches
Combined series both approaches 17
Slide 18: Peaks and crises
Peaks and crises More often than not, the two methods identify peaks within five quarters from each other Jointly identified peaks are often close to crises 18
Slide 19: Peaks identified by both methods and crises
am BANK FOR INTERNATIONAL SETTLEMENTS Time to closest’ Peak Cluster Date Crises (filters) Peak (TP) width® Duration* Number of quarters Peaks identified by both methods that are close to a crisis® GB 2009 Q1 -6 5 0 10 72 SE 2009 Q1 —2 4 0 7 75 US 2007 Q3 0 0 0 12 68 JP 1992 Q2 2 -3 0 8 74 GB 1991 Q1
Slide 20: The financial and business cycles: US
The financial and business cycles: US 20
Slide 21: Greece: the short and medium credit cycles
Greece: the short and medium credit cycles 1 0.5 0 -0.5 -1 -1.5 21
Slide 22: Summary of results
Summary of results The financial cycle looks like the medium-term component of credit and property prices This cycle seems to have strengthened and lengthened post-1985 The cycle peaks align well with crises Recessions are worse when in financial cyclical downturn What does this imply for ma
Slide 23: Houston, we have a problem…
Houston, we have a problem… “The banking system’s reported financial indicators are above minimum regulatory requirements and stress tests suggest that the system is resilient” 19 August 2008, IMF, Iceland: Financial Stability Assessment – update, p 5 Quote typical of the tone in vulnerability ass
Slide 24: Five Propositions
Five Propositions I. Macro stress testing is a tool box, not a single tool II. Beware of macro stress tests as early warning devices III. Macro stress tests can greatly help in crisis management/resolution… IV. …and their additional benefits should not be underestimated…. V. …but when you do them, d
Slide 25: Beware of model limitations for early warning
Beware of model limitations for early warning Partial equilibrium exercises Likely sources of miss-specification (ie model risk for stress testing) • Log-linearity (macro model legacy) • Crises as “structural breaks” Models the antithesis of what financial stress is all about • Instability: no
Slide 26: Crises are tail events with pronounced dynamics
Crises are tail events with pronounced dynamics 2 Distribution of real GDP growth Actual and forecast real GDP growth 26
Slide 27: Crises emerge at the top of the financial cycle
Crises emerge at the top of the financial cycle The financial cycle United States Australia United Kingdom Japan 27
Slide 28: The system appears strongest when it is most vulnerable
The context: the system appears strongest when it is most vulnerable (paradox of financial instability) 5 9 Buoyant asset prices Subdued implied volatilities Price of insurance against distress 28
Slide 29: Reality can be brutal...
am BANK FOR INTERNATIONAL SETTLEMENTS Reality can be brutal... Actual and forecast real GDP growth Three-month Libor minus OIS spreads” T T R 6 — US dollar % — Euro 400 1 3 =—= Pound sterling ' 300 L 0 200 3 1 =— Actual 1 3 =—— Forecast : 100 one year ahead* 1 == Forecast this year’s -6 0 DC “16-12
Slide 30: Beware of stress tests as early warning exercises
Beware of stress tests as early warning exercises Model limitations → shocks have to be large to get any effect Initial conditions (balance sheet strength and earnings capacity) perceived as benign → shocks have to be even larger Hubris is high and prudence is low → bias to argue that analysis
Slide 31: Macro stress tests can help in crisis resolution
Macro stress tests can help in crisis resolution Hubris has given way to prudence Vulnerabilities have already crystallized • Key risk is known • Weak macro conditions allow history to be used for scenarios that are as severe as reality The important behavioural reactions that create non- line
Slide 32: A way forward?
A way forward? Enhance models • Focus on common exposures and common reactions rather than bilateral inter-linkages Add patches to model that can deal with non-linearities Use information from early warning indicators 32
Slide 33: Predicting banking crises
Predicting banking crises LOW HIGH 33
Slide 34: Loss rate on real estate loans and financial obligations ratio
am BANK FOR INTERNATIONAL SETTLEMENTS 7 — Loss rate on real estate loans ab 1h ° A VV L L n n L n n L L L n n n L 1985 1990 1995 2000 2005 2010 — Financial obligations ratio, ue household’s real estate debt 10 F L MSBD oF L 1 1 L L 1985 1990 1995 2000 2005 2010
Slide 35: Sensitivity of bank write-downs to GDP
Sensitivity of bank write-downs to GDP Δ(Bank Losses) -------------------------- Beta Δ(GDP) high Beta low DSR Juselius and Kim (2011) K* Drehmann and Juselius (2012) 35
Slide 36: Important additional benefits
Important additional benefits They force a common language about financial stability • To break the compartmentalised view of key stakeholders (macroeconomists, finance specialist, risk managers or supervisors) Plus: • Improve toolbox for financial stability analysis • Improve data availability
Slide 37: When you do them, do them right!
When you do them, do them right! Have the will to really stress the system • Should not feel to constrained by recent history Ensure buy-in by all stakeholders • Clear objectives and responsibilities for policymakers • Create incentives for market participants Do not undermine banks’ own stres
Slide 38: Thank you!
Thank you! ktsatsaronis@bis.org 38
Bank for International Settlements — Investor Presentation (March 2013) on 6ix