Docstoc

4173_4156_Presentation_20101019_ClaudiaChampagne_presentation

Document Sample
4173_4156_Presentation_20101019_ClaudiaChampagne_presentation Powered By Docstoc
					Is it contagious?
Presentation by Claudia Champagne, Ph.D.

Université de Sherbrooke
Groupe de recherche en finance appliquée (GReFA)
IFM2
Outline
1.   Contagion: definition and importance
2.   Contagion: observations
3.   Contagion and systemic risk measures
     a) Network approach
     b) Co-movement models
      i. Co-exceedances
      ii. Co-risk
4.   Conclusion
5.   Questions
                          C. Champagne, October 2010   2
1.1 Definitions
   Financial contagion: transmission of a
    financial shock in one entity to other
    independent entities
    ◦ For credit risk: affects the joined default
      probability
   Systemic risk: risk which can influence
    the stability of the financial system as a
    whole
    ◦ Concerned with the joint distribution of the
      losses of all market participants

                                 C. Champagne, October 2010   3
1.2 Why measure contagion?
   Important element of credit and systemic
    risk
    ◦ Need to specify the degree of association
      or codependence between the firms
      To obtain a conditional probability of default or
       joined default probabilities
      To assess spillover risk between institutions




                                    C. Champagne, October 2010   4
1.2 Why measure contagion? (2)
   Increased interdependence and
    globalization of the financial system lead
    to a problem of too-connected-to-fail
◦ « Too big to fail » vs « Too connected to
  fail »




                             C. Champagne, October 2010   5
1.2.1 Too-connected-to-fail?
    Surcharge for systemic risk
     ◦ Acharya (2009): capital surcharges based on
       interbank correlation of returns
     ◦ Brunnermeier et al. (2009) and Chan-Lau
       (2010): capital surcharges based on co-
       movements of banks’ risks (e.g. co-value-
       at-risk)
     ◦ Bank of England (2009): capital surcharges
       based on measures of institutions’ and
       markets’ degree of « exuberance »

                               C. Champagne, October 2010   6
2. Contagion - observations
1.   Default contagion (or systemic impact) is
     not about the size of the firm
2.   Contagion goes beyond direct
     interfirm exposure
     ◦ Many studies using interbank lending find
       weak evidence of contagion…
     ◦ Geneva Report (August 2008): « contagion
       effects have been studied [previously] by
       central banks and found to be negligible » …

                               C. Champagne, October 2010   7
2. Contagion – observations (2)
3.       Indirect links or exposures (e.g. CDS) can
         generate default contagion in non-
         traditional ways
     ◦    Example: AIG was not a direct counterparty to
          Lehman Brothers
4.       Homogeneity can also generate default
         contagion
     ◦    Similarity of assets (loans, securities, etc.)
     ◦    Same business model or accounting practices,
          etc.
5.       Contagion can be within a country and
         cross-border

                                   C. Champagne, October 2010   8
3. Systemic risk measures
   IMF Report (2009): 4 complementary
    approaches to assess direct and indirect
    systemic linkages:
    ◦ Network approach
    ◦ Co-movement models
      Co-exceedances analysis (structural model)
      Co-risk analysis
      Etc.
    ◦ Default intensity model
    ◦ Distress dependence matrix model

                             C. Champagne, October 2010   9
3.1 Network approach
   Mapping of relationships between firms
    ◦ Use of graph theory
   A network of connections is modeled as a
    graph:
    ◦ N nodes which represent financial market
      participants
    ◦ Ties (links) which represent the connections
      between nodes
   The analysis goes from whole to part; from
    structure to relation to individual firm
                                C. Champagne, October 2010   10
3.1.1Canadian network




                 C. Champagne, October 2010   11
3.1.2 Canada - U.S. Network




                  C. Champagne, October 2010   12
3.1.3 Canada – U.S. – Europe network




                     C. Champagne, October 2010   13
3.1.4 Network metrics
Network     # actors   Degree   Density   Centralization   Clustering    Geodesic   Small-world
                                                           coefficient   distance    statistic
Canadian

2005-2009     16        10      66.67%       22.86%          99.2%         1.37        1.40

1995-2000     12        4.83    43.94%       45.45%          84.0%         1.37        2.40

USA

2005-2009     746      63.93    8.58%        65.96%          80.3%        2.068        7.21

1995-2000     614       59.8    9.76%        62.88%          78.1%        2.128        5.91

Europe

2005-2009     354      58.98    16.71%       60.12%          81.6%        1.892        3.73

1995-2000     261      45.79    17.61%       65.59%          82.4%        1.873        3.65

Global

2005-2009    1933      80.65    4.18%        44.78%          80.2%        2.258       14.67

1995-2000    1584      72.08    4.55%        58.94%          77.7%        2.228       13.20

                                                      C. Champagne, October 2010              14
3.1.5 Ego-centered metrics
Actor                Degree   Bonacich Power   ARD       Eigenvector      Betweenness


Canadian network
Scotia               0.867        4.623        0.933         0.409           0.042
BMO                  0.867        4.561        0.933         0.403           0.138
TD, RBC, CIBC, BNC   0.800        4.528        0.900         0.400           0.004
Global network
Mizuho Financial     0.489       302.927       0.744         0.142           0.099
Group
JP Morgan            0.478       296.195       0.737         0.161           0.045
RBS                  0.470       291.387       0.734         0.145           0.057
ING                  0.451       279.206       0.723         0.158           0.036
CIBC (29)            0.261       161.561       0.625         0.113           0.007
Scotia (30)          0.257       158.997       0.626         0.112           0.007
TD (47)              0.207       128.223       0.599         0.094           0.006
RBC (57)             0.183       113.157       0.586         0.087           0.005
BMO (78)             0.148        91.359       0.567         0.078           0.002
BNC (620)            0.045        27.568       0.515         0.036           0.000


                                                       C. Champagne, October 2010       15
3.1.6 Network approach advantages
   Identifies systemic linkages and
    vulnerable countries or institutions
    ◦ Quantifies potential capital losses at
      country or institution level
 Identifies systemically important
  institutions
 Can track potential contagion paths
    ◦ Provides metric on domino effect induced by
      alternative distress events

                              C. Champagne, October 2010   16
3.2 Co-movement models
 Measure of interconnectedness risk
 Not based on a specific channel of
  contagion
    ◦ Co-movements can be captured in security
      prices and risk measures
   Capture firms’ codependance risk from
    direct and indirect linkages



                             C. Champagne, October 2010   17
Problems with correlation
   Measure of dependence in the center of
    the distribution
       Not suitable for the examination of systemic risk or default risk
        (i.e. tail risk)

 Detects only linear dependence between
  2 variables
 Does not distinguish contagion from
  common shocks affecting firms
  simultaneously

                                            C. Champagne, October 2010      18
3.2.1 Co-exceedances
   Gropp & Moerman (2004):
    ◦ Model based on the extreme value theory
      (EVT) framework
      Captures tail risk
      Considers non-linearities
    ◦ Use distance-to-default, ln(∆DD) as measure
      of risk
      Could use other measure such as CDS spreads
    ◦ Co-exceedance: number of periods in which
      more than one institution is in the tail of the
      distribution (i.e. bottom and top 5%)

                                   C. Champagne, October 2010   19
3.2.1 Co-exceedances
   Co-exceedances model applied to Canada – U.S.:

                     Number of (co-)exceedances in the bottom tails

                      ≥5        4       3       2         1              0

    Canada (11)        11       0       6      13         39         204

    US (20)            0       12      19      48         79         115

    Canada-US (31)     10      12      28      23         0          200




                                            C. Champagne, October 2010       20
3.2.2 Co-risk analysis
 Adrian & Brunnermeier (2009); Chan-Lau
  (2010)
 Interconnectedness risk for different
  quantiles (risk regimes)
 How the default risk of one institution
  reacts to changes in the default risk of
  another institution after controlling for
  common risk factors
    ◦ Can be estimated using quantile regression

                              C. Champagne, October 2010   21
AIG and Lehman Brothers CoRisk
    Relationship between CDS spreads




Source: IMF (2009)




                              C. Champagne, October 2010   22
4. Conclusions
   Contagion is a major component of credit and
    systemic risks
    ◦ Needs to be measured correctly
    ◦ Consequences are important
       E.g. capital surcharges
   Proposed approaches are complementary rather
    than substitutes
    ◦ Differing inputs and outputs
    ◦ Choices can also depend on usage:
       Loan pricing?
       Capital surcharge?
   Results show that measures need to be global
                                  C. Champagne, October 2010   23
5. Questions?


         C. Champagne, October 2010   24

				
DOCUMENT INFO
Shared By:
Categories:
Stats:
views:4
posted:7/9/2011
language:French
pages:24