Dating Systemic Financial Stress Episodes in the EU Countries - - PowerPoint PPT Presentation

▶
dating systemic financial stress episodes in the eu
SMART_READER_LITE
LIVE PREVIEW

Dating Systemic Financial Stress Episodes in the EU Countries - - PowerPoint PPT Presentation

Sep 11, 2022 207 likes •464 views

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

slide-1
SLIDE 1

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Dating Systemic Financial Stress Episodes in the EU Countries

Thibaut DUPREY1 Benjamin KLAUS2 Tuomas PELTONEN3

1Bank of Canada 2European Central Bank 3European 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

slide-2
SLIDE 2

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

slide-3
SLIDE 3

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Aim = identify systemic financial stress

Low financial stress High financial stress High growth Low growth tranquil regime financial stress recession systemic stress

slide-4
SLIDE 4

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Overview

1

Construct 27 financial stress indices for all EU countries

◮ Financial cycle research : financial stress index literature 2

Identify systemic financial stress episodes

◮ 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
slide-5
SLIDE 5

Motivation Step 1 Step 2 Results Benchmarking Conclusion

STEP 1 : Construct 27 financial stress indices

(in the spirit of CISS : Hollo et al., 2012)

Equity sub-index Bonds sub-index FX sub-index P a i r w i s e c r

  • s

s

  • c
  • r

r e l a t i

  • n

s ρ

  • f

i n d i c e s I Country-Level Index of v

  • l

a t i l i t y s t

  • c

k s cumulated drop in stocks volatility of government bond cumulated government bond spread volatility effective exchange rate cumulated change effective exchange rate volatility idiosyncratic bank returns cumulated drop bank stocks mortgage lending spread cumulated housing price drop Bank sub-index Housing sub-index normalized in the [0;1] space using the empirical cumulative density Financial Stress (CLIFS) CLIFSt = It ∗ Ct ∗ I

′

t

Ct =    1 . . . ρi,j,t . . . ... . . . ρi,j,t . . . 1   

5∗5

It = [Ii,t . . . Ij,t]1∗5

slide-6
SLIDE 6

End of 2006 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347

slide-7
SLIDE 7

End of 2007 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347

slide-8
SLIDE 8

End of 2008 Dataset publicly available : http://sdw.ecb.europa.eu/browse.do?node=9693347

slide-9
SLIDE 9

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Country-Level Index of Financial Stress (CLIFS)

0.1 0.2 0.3 0.4 0.5 0.6 0.7

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

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.

slide-10
SLIDE 10

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Contribution of the cross-correlations

  • 140
  • 120
  • 100
  • 80
  • 60
  • 40
  • 20

20 40 60 80

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15

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.

slide-11
SLIDE 11

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Does financial stress matter ?

Industrial production growth per quantiles of CLIFS

  • 8
  • 6
  • 4
  • 2

2 4 6 8 0-10 10-20 20-30 30-40 40-50 50-60 60-70 70-80 80-90 90-100

Average annual industrial production growth on the y-axis. Quantiles of the country-specific financial stress indices on the x-axis.

slide-12
SLIDE 12

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 Low growth tranquil regime financial stress recession systemic stress

slide-13
SLIDE 13

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Method 1 : Markov-Switching with selection algorithm

Hamilton (1989) Markov-Switching framework

1

Identify periods of high financial stress : CLIFSt = µSt + βCLIFSt−1 + σStǫt Transition probability across regimes St ∈ {L, H} driven by a hidden two-state Markov chain : P (St |St−1 ) =

  • p =

exp(θp) 1+exp(θp)

1 − p 1 − q q =

exp(θq) 1+exp(θq)

  • regime H when µH > µL, and financial stress period when :

1financialstress = {P (St = H) > 0.5}

2

Overlap with at least six consecutive months of real economic stress (drop in industrial production and GDP correction)

slide-14
SLIDE 14

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              gIPIt = µSt

1 + n

  • p=1
  • βSt

1,1,pgIPIt−p + βSt 1,2,pCLIFSt−p

  • + ǫt,1

CLIFSt = µSt

2 + n

  • p=1
  • βSt

2,1,pCLIFSt−p + βSt 2,2,pgIPIt−p

  • + ǫt,2

The tranquil or systemic financial stress state St ∈ {L; H} is unobservable : same hidden two-state Markov chain as before.

slide-15
SLIDE 15

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              CLIFSt = µSt

1 + n

  • p=1
  • βSt

1,1,pCLIFSt−p + βSt 1,2,pgIPIt−p

  • + ǫt,1

gIPIt = µSt

2 + n

  • p=1
  • βSt

2,1,pgIPIt−p + βSt 2,2,pCLIFSt−p

  • + ǫt,2

The observed regime is given by : St = H if CLIFSt−1 ≥ τ L if CLIFSt−1 < τ where τ is estimated.

slide-16
SLIDE 16

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 Sm,t for periods of systemic financial stress over all models m

  • robust to model uncertainty

Systemic Stress Index SSIt =

  • m Sm,t
  • m 1m

∈ [0; 1] Definition of systemic financial stress :

  • starts when SSIt > 0.5
  • ends when SSIt < 0.25
slide-17
SLIDE 17

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Zoom-in : Systemic Stress Indices, selected countries

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 01/1965 01/1969 01/1973 01/1977 01/1981 01/1985 01/1989 01/1993 01/1997 01/2001 01/2005 01/2009 01/2013

(a) Portugal

0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 01/1965 01/1969 01/1973 01/1977 01/1981 01/1985 01/1989 01/1993 01/1997 01/2001 01/2005 01/2009 01/2013

(b) Spain

slide-18
SLIDE 18

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Zoom out : Timing of systemic financial stress in 2008

01/2007 01/2008 01/2009 01/2010 01/2011 01/2012 01/2013 01/2014 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

slide-19
SLIDE 19

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Zoom out further : Financial market stress and intensity of real economic stress

02/1964 02/1969 02/1974 02/1979 02/1984 02/1989 02/1994 02/1999 02/2004 02/2009 02/2014 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

  • 30%
  • 20%
  • 10%

No stress

slide-20
SLIDE 20

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)

Months 6 12 18 24

  • 6
  • 5
  • 4
  • 3
  • 2
  • 1

Industrial Production Index

slide-21
SLIDE 21

Motivation Step 1 Step 2 Results Benchmarking Conclusion

What are systemic financial stress episodes ? Not

  • rdinary recessions

Number Length GDP CLIFS mean Definition of recession : events loss pcent corr Ordinary recessions Two quarters 76 11

  • 0.79

50

  • 7

Two consecutive quarters 45 7

  • 1.52

50

  • 18

Before 2008, two quarters 57 10

  • 0.77

51

  • 16

Recessions with financial market stress Two quarters 74 18

  • 4.10

66 28 Two consecutive quarters 42 13

  • 4.17

70 31 Before 2008, two quarters 39 14

  • 1.71

72 28

slide-22
SLIDE 22

Motivation Step 1 Step 2 Results Benchmarking Conclusion

What are systemic financial stress episodes ? Sectoral decomposition for selected countries

Bi-product of the Markov-switching model

Syst fin stress Equity crash Sovereign stress Foreign exchange Banking stress Housing stress IPI drop GDP drop

(c) Portugal

Syst fin stress Equity crash Sovereign stress Foreign exchange Banking stress Housing stress IPI drop GDP drop

(d) Spain

slide-23
SLIDE 23

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Comparison of continuous stress measures with expert-based crises : AUROC

CLIFS SSI panel average panel average Detken et al. (2014) Banking 0.71 0.76 0.82 0.83 Babecky et al. (2012) Banking 0.66 0.72 0.80 0.83 Currency 0.71 0.68 0.82 0.74 Debt 0.94 0.94 0.91 0.95 Leaven and Valencia (2013) Banking 0.75 0.77 0.87 0.88 Reinhart and Rogoff (2011) Banking 0.70 0.75 0.84 0.87 Currency 0.53 0.51 0.67 0.69 Equity 0.66 0.68 0.74 0.77

slide-24
SLIDE 24

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Comparison of model-based systemic financial stress episodes with expert-based crises

Share of model Share of expert identified events also identified crises also captured by experts captured by models Spain 1.00 0.43 Portugal 1.00 0.14 Total 0.81 0.43 Mean 0.83 0.55 In particular, we capture 96% of the systemic banking crises of Leaven and Valencia (2012)

slide-25
SLIDE 25

Motivation Step 1 Step 2 Results Benchmarking Conclusion

Wrap-up

Paper combines measurement of financial stress with detection

  • f turning points for a mechanistic dating of systemic financial

stress episodes Upsides :

  • Get model-implied systemic financial stress periods
  • Integrate real and financial cycle dynamics (=systemic)
  • Consistent with most expert-based datasets
  • Robust to alternative measures of financial stress
  • Robust to model uncertainty
  • Robust to event reclassification once new data arrive

Downsides :

  • Hard to capture causal relation between financial stress and real

economic stress Follow up work : "A new database for financial crises in European countries", ECB Occasional Paper No. 194, July 2017.