20140923090009451

advertisement
Systemic risk in derivatives markets
A pilot study using CDS data
Nicholas Vause (joint work with Robleh Ali and Filip Zikes)
23 September 2014
Outline
• Trade repository data
• Indicators of systemic importance
– Contracts traded
• Indicators of systemic risk
– Market values of contracts traded
• Further work
Trade repository data
• Pittsburgh summit of G20 leaders (2009):
“All OTC derivatives contracts should be reported to trade repositories”
• Implemented in Europe under EMIR
• Reporting began in February 2014
Trade repository data in Europe
• 26 counterparty fields
– name, sector, region
– market value and collateral posted (from August 2014)
• 59 common fields
– product, notional amount, maturity
– class-specific data (e.g. reference entity for CDS)
• All classes
– Interest rates, credit, FX, equities and commodities
• 100,000s legal entities
Pilot study data
•
•
•
•
•
•
•
•
CDS data from DTCC’s Trade Information Warehouse
UK reference entities (66 single names)
Jan 2009 – Dec 2011
Market participants: 300
Gross notional: €600bn
Net notional: €25bn
2% of global CDS market
Gross market value: €15-50bn
Net market value: €1.0-2.5bn
CDS preliminaries
MV = €0
S
S sells protection on R to B
Notional amount = €10mn
B
MV = €0
B
MV = -€50K
B
MV = €0
Premium = 80bp
MV = +€50K
S
S sells protection on R to B
Notional amount = €10mn
Premium = 60bp
Variation margin = €50K
MV = €0
S
S sells protection on R to B
Notional amount = €10mn
Premium = 80bp +/- 30bp
Initial margin = €75K
Initial margin = €75K
Network preliminaries
• Directed weighted networks
Buyers
Y
3
6
Sellers
1
X
Z
2
0

0
2

3
0
6
0

1  A
0 
Systemic importance – network of contracts traded
All participants (out degree)
Proportion of participants
Top 25 participants
Number of connections
Systemic importance – centrality measures
• Degree centrality (DC)
• Eigenvector centrality (EC): Ax = x (highest )
• Betweenness centrality (BC)
Highest DC
Highest BC
Highest EC
Systemic importance – most important institutions
Out strength (sold notional, €bn)
Left eigenvector centrality
Betweenness centrality
•
•
High correlation between measures
Similar results for purchases of credit protection
Systemic stress – market values
• Positions valued using end-of-day prices
Blue = in the money
Red = out of the money
Arrows point from ‘debtors’ to ‘creditors’
• But collateral
Modelling changes in market values
CDS curves
(1, 3, 5, 7, 10)
PCA
Level factors
AR(2)-GARCH(1,1)
Slope factors
t-distributed residuals
Univariate models
of factor changes
t-copula
(correlations linked to
industry factors)
Joint model of
level changes
Joint model of
slope changes
• Jointly simulate residuals
• Compute factor changes (conditional on starting point)
• Compute CDS changes and, hence, change in market value
Individual probabilities of gains / losses (€ millions)
0.03
0.100
Var1
N(s=17.9)
0.075
0.02
0.06
Var2
N(s=5.34)
Var3
N(s=9.35)
0.050
0.01
0
100
0.03
-25
0
25
-50
0.15
Var5
N(s=19)
Var6
N(s=3.58)
0.02
0.10
0.01
0.05
-100
0.3
0
100
Var9
N(s=1.92)
0.15
0
50
-20
Var7
N(s=3.15)
20
Var8
N(s=8.5)
0.050
0.025
0.05
-20
0
20
-20
0.20
Var10
N(s=2.23)
0.2
0.15
0
20
Var11
N(s=2.96)
-50
0.10
0.10
0.1
0.1
0
50
Var12
N(s=4.45)
0.05
0.05
-20
0
-20
Var13
N(s=5.83)
0
0.50
0.025
0.25
20
Var14
N(s=0.541)
0.75
0.050
-50
-25
0
-50
Var15
N(s=2.37)
0.2
0.02
-25
0
25
-5
0
5
-25
0
25
Var16
N(s=22.1)
0.01
0.1
-50
0
0.10
0.2
0.075
Var4
N(s=2.26)
0.1
0.02
0.025
-100
0.2
0.04
0
20
40
-100
0
100
Joint probabilities of losses
•
Joint probabilities of pairs of dealers each breaching their 95% VaRs
– Independence = 0.25% (i.e. 5% x 5%)
– Perfect dependence = 5%
D1
D2
D3
D4
D5
D6
D7
D8
D7
D10
D11
D12
D13
D14
D15
D2
0.13
D3
0.00
3.03
D4
2.13
0.03
0.00
D5
0.00
2.46
3.78
0.00
D6
0.66
0.06
0.05
0.94
0.08
D7
2.44
0.32
0.08
1.43
0.04
0.81
D8
0.00
2.80
3.88
0.01
3.56
0.07
0.11
D7
2.06
0.81
0.49
1.00
0.45
0.31
1.78
0.47
D10
0.63
0.02
0.01
0.94
0.05
0.44
0.25
0.00
0.15
D11
0.65
2.13
2.04
0.25
2.07
0.07
0.51
1.86
1.75
0.10
D12
4.08
0.13
0.00
2.09
0.00
0.64
2.65
0.00
2.07
0.38
0.64
D13
3.74
0.10
0.00
2.41
0.00
0.77
2.04
0.01
1.71
0.90
0.55
3.74
D14
0.59
1.44
1.06
0.31
0.76
0.19
0.98
1.27
0.96
0.01
0.99
0.69
0.58
D15
0.00
1.61
1.63
0.03
1.25
0.07
0.17
1.97
0.28
0.00
0.82
0.01
0.01
2.36
D16
4.35
0.11
0.00
2.43
0.00
0.75
2.53
0.00
1.94
0.71
0.60
4.27
4.05
0.61
0.01
Stress test – variation margins
•
•
Stress test = all 66 CDS premiums rise by 100bp
Variation margin from payer (rows) to receiver (columns) in € millions
Stress test – initial margins
•
•
Stress test = volatilities rise from March 2010 to May 2010 levels
Initial margin requirements in € millions
Bilateral positions
Centrally-cleared positions
Future work
• Combine derivatives data with financial accounts
• Contagious default or CVA losses in network of exposures
• Depends on capital cushions
• Much less potent given collateral reforms
• Contagious liquidity strains in network of margin payments
• Depends on liquidity cushions
• Increasing in likelihood given collateral reforms
Download