Motivation Step 1 Step 2 Results Benchmarking Conclusion Dating Systemic Financial Stress Episodes in the EU Countries Thibaut D UPREY 1 Benjamin K LAUS 2 Tuomas P ELTONEN 3 1 Bank of Canada 2 European Central Bank 3 European Systemic Risk Board The views are those of the authors and do not necessarily reflect those of Bank of Canada, the European Central Bank, the Eurosystem or the European Systemic Risk Board. Central Bank of Brazil, 9 August 2017
Motivation Step 1 Step 2 Results Benchmarking Conclusion Classifying events for a better analysis of macropru The analysis of macroprudential policies requires a chronology of systemic crises • 2008 can be safely ( ?) regarded as a systemic financial crisis • But the classification of all other events rely on expert judgement ... We provide a mechanistic identification of systemic financial stress
Motivation Step 1 Step 2 Results Benchmarking Conclusion Aim = identify systemic financial stress Low financial stress High financial stress High growth tranquil regime financial stress Low growth systemic stress recession
Motivation Step 1 Step 2 Results Benchmarking Conclusion Overview Construct 27 financial stress indices for all EU countries 1 ◮ Financial cycle research : financial stress index literature Identify systemic financial stress episodes 2 ◮ Business cycle research : identifying business cycle turning points using a suite of non-linear models • Method 1 : Univariate Markov switching with algorithm • Method 2 : Markov switching vector autoregressive model • Method 3 : Threshold vector autoregressive model
Motivation Step 1 Step 2 Results Benchmarking Conclusion STEP 1 : Construct 27 financial stress indices (in the spirit of CISS : Hollo et al., 2012) v o l a t i l i t y s t o c k s Equity sub-index I s cumulated drop in stocks e normalized in the [0;1] space using the empirical cumulative density c ′ i CLIFS t = I t ∗ C t ∗ I d t n volatility of government bond i f Bonds sub-index o ρ cumulated government bond spread s n Country-Level Index of o volatility effective exchange rate i t a Financial Stress (CLIFS) FX sub-index l e r cumulated change effective exchange rate r o c - s volatility idiosyncratic bank returns s o Bank sub-index r I t = [ I i,t . . . I j,t ] 1 ∗ 5 c cumulated drop bank stocks e s i w 1 . . . ρ i,j,t r mortgage lending spread i . . ... a C t = . . Housing sub-index P . . cumulated housing price drop ρ i,j,t . . . 1 5 ∗ 5
End of 2006 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347
End of 2007 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347
End of 2008 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347
Motivation Step 1 Step 2 Results Benchmarking Conclusion Country-Level Index of Financial Stress (CLIFS) 7 10 12 14 1 2 3 4 5 6 8 9 11 13 15 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 01/1965 01/1967 01/1969 01/1971 01/1973 01/1975 01/1977 01/1979 01/1981 01/1983 01/1985 01/1987 01/1989 01/1991 01/1993 01/1995 01/1997 01/1999 01/2001 01/2003 01/2005 01/2007 01/2009 01/2011 01/2013 01/2015 1 - first oil shock ; 2 - second oil shock ; 3 - Mexican debt crisis ; 4 - Black Monday ; 5 - crisis of the European exchange rate mechanism ; 6 - Peso crisis ; 7 - Asian crisis ; 8 - Russian crisis ; 9 - dot-com bubble ; 10 - subprime crisis ; 11 - Bankruptcy of Lehman Brothers ; 12 - 1st bailout Greece ; 13 - 2nd bailout Greece ; 14 - Election of Alexis Tsipras in Greece ; 15 - Brexit vote.
Motivation Step 1 Step 2 Results Benchmarking Conclusion Contribution of the cross-correlations 7 10 12 14 1 2 3 4 5 6 9 8 11 13 15 80 60 40 20 0 -20 -40 -60 -80 -100 -120 -140 01/1965 01/1967 01/1969 01/1971 01/1973 01/1975 01/1977 01/1979 01/1981 01/1983 01/1985 01/1987 01/1989 01/1991 01/1993 01/1995 01/1997 01/1999 01/2001 01/2003 01/2005 01/2007 01/2009 01/2011 01/2013 01/2015 1 - first oil shock ; 2 - second oil shock ; 3 - Mexican debt crisis ; 4 - Black Monday ; 5 - crisis of the European exchange rate mechanism ; 6 - Peso crisis ; 7 - Asian crisis ; 8 - Russian crisis ; 9 - dot-com bubble ; 10 - subprime crisis ; 11 - Bankruptcy of Lehman Brothers ; 12 - 1st bailout Greece ; 13 - 2nd bailout Greece ; 14 - Election of Alexis Tsipras in Greece ; 15 - Brexit vote.
Motivation Step 1 Step 2 Results Benchmarking Conclusion Does financial stress matter ? Industrial production growth per quantiles of CLIFS 8 6 4 2 0 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100 -2 -4 -6 -8 Average annual industrial production growth on the y-axis. Quantiles of the country-specific financial stress indices on the x-axis.
Motivation Step 1 Step 2 Results Benchmarking Conclusion STEP 2 : How to identify systemic financial stress episodes ? Low financial stress High financial stress High growth tranquil regime financial stress Low growth systemic stress recession
Motivation Step 1 Step 2 Results Benchmarking Conclusion Method 1 : Markov-Switching with selection algorithm Hamilton (1989) Markov-Switching framework Identify periods of high financial stress : 1 CLIFS t = µ S t + β CLIFS t − 1 + σ S t ǫ t Transition probability across regimes S t ∈ { L , H } driven by a hidden two-state Markov chain : exp ( θ p ) � � p = 1 − p 1 + exp ( θ p ) P ( S t | S t − 1 ) = exp ( θ q ) 1 − q q = 1 + exp ( θ q ) regime H when µ H > µ L , and financial stress period when : 1 financialstress = { P ( S t = H ) > 0 . 5 } Overlap with at least six consecutive months of real economic 2 stress (drop in industrial production and GDP correction)
Motivation Step 1 Step 2 Results Benchmarking Conclusion Method 2 : Markov switching vector autoregression builds on toolbox of Haroon Mumtaz Bivariate model to capture joint change in dynamics of industrial production growth ( gIPI ) and CLIFS n � � gIPI t = µ S t � β S t 1 , 1 , p gIPI t − p + β S t 1 + 1 , 2 , p CLIFS t − p + ǫ t , 1 p = 1 n � � CLIFS t = µ S t � β S t 2 , 1 , p CLIFS t − p + β S t 2 + 2 , 2 , p gIPI t − p + ǫ t , 2 p = 1 The tranquil or systemic financial stress state S t ∈ { L ; H } is unobservable : same hidden two-state Markov chain as before.
Motivation Step 1 Step 2 Results Benchmarking Conclusion Method 3 : Threshold vector autoregressive model builds on toolbox of Gabriel Bruneau Different joint dynamics above ( H ) or below ( L ) an estimated percentile of the CLIFS n � � CLIFS t = µ S t � β S t 1 , 1 , p CLIFS t − p + β S t 1 + 1 , 2 , p gIPI t − p + ǫ t , 1 p = 1 n � � gIPI t = µ S t � β S t 2 , 1 , p gIPI t − p + β S t 2 + 2 , 2 , p CLIFS t − p + ǫ t , 2 p = 1 The observed regime is given by : � H if CLIFS t − 1 ≥ τ S t = L if CLIFS t − 1 < τ where τ is estimated.
Motivation Step 1 Step 2 Results Benchmarking Conclusion Robustly identifying systemic financial stress events For each 27 countries we have up to 12 models • different framework, with different specifications, using CLIFS or the banking and housing extensions For each country, combine dummies S m , t for periods of systemic financial stress over all models m • robust to model uncertainty � m S m , t Systemic Stress Index SSI t = ∈ [ 0 ; 1 ] � m 1 m Definition of systemic financial stress : • starts when SSI t > 0 . 5 • ends when SSI t < 0 . 25
Zoom-in : Systemic Stress Indices, selected countries Motivation 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 1 01/1965 01/1969 01/1973 01/1977 Step 1 01/1981 (a) Portugal 01/1985 01/1989 01/1993 01/1997 Step 2 01/2001 01/2005 01/2009 01/2013 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 1 Results 01/1965 01/1969 01/1973 01/1977 01/1981 (b) Spain 01/1985 Benchmarking 01/1989 01/1993 01/1997 01/2001 01/2005 01/2009 01/2013 Conclusion
Motivation Step 1 Step 2 Results Benchmarking Conclusion Zoom out : Timing of systemic financial stress in 2008 IE BE UK DK IT NL AT ES PT HU CZ DE FR SE GR HR LU RO SI LT FI LV MT CY BG PL SK 01/2007 01/2008 01/2009 01/2010 01/2011 01/2012 01/2013 01/2014
Motivation Step 1 Step 2 Results Benchmarking Conclusion Zoom out further : Financial market stress and intensity of real economic stress IE No stress BE UK DK IT NL AT -10% ES PT HU CZ DE FR SE GR HR -20% LU RO SI LT FI LV MT CY BG -30% PL SK 02/1964 02/1969 02/1974 02/1979 02/1984 02/1989 02/1994 02/1999 02/2004 02/2009 02/2014
Motivation Step 1 Step 2 Results Benchmarking Conclusion No zoom : Systemic financial stress is costly Bi-product of the Threshold VAR Response of industrial production to a shock of 1% on CLIFS (black : VAR without regime change ; red : high stress ; blue : tranquil) Industrial Production Index 0 -1 -2 -3 -4 -5 -6 0 6 12 18 24 Months
Recommend
More recommend
