Systemic illiquidity in the Russian interbank market

1
Systemic illiquidity in the Russian interbank
market

• Alexei Karas
• Gleb Lanine
• Koen Schoors

2
Background

• Russia faced 3 severe interbank market crises
• August 1995
• Careless risk management
• Structural reliance on interbank market for
  financing assets
• August 1998
• Collapse of the GKO market
• Unhedged positions in currency forwards
• May/June 2004
• Mini-crisis on the interbank market
• These interbank market crises are costly
• Systemic instability
• Trust of depositors is affected
• CBR intervenes to solve the problem

3
Motivation

• Bank supervision neglects interbank linkages on
  the interbank market
• Although in its enforcement the CBR seems to
  protect money centre banks (See Claeys, Schoors,
  2007)
• We want to understand
• How vulnerable the Russian interbank market is to
  contagion
• Whether this is linked to the structure of the
  banking system
• Whether the CBRs past interventions have helped
  to stabilize the interbank market
• Whether the CBR could improve the effectiveness
  of its interventions

4
Contributions

• We consider several types of shocks
• A shock to an individual bank default
• Correlated bank defaults
• We define a new transmission channel
• Next to the traditional capital channel
• We define an innovative liquidity channel
• We show this new channel is relevant in reality
• We link this to the interbank market structure
  (through centrality measures)
• And use this to assess CBR interventions

5
The data

• We have the bank balances and income statements
  from two sources
• INTERFAX
• Mobile
• We have the bilateral interbank exposures
• A matrix of more than 1000 x 1000
• Monthly data
• For the period 1998-2004
• Covering two crises on the interbank market

6
Market participation
7
Liquidity drains on the Russian interbank market
8
The dominance of large banks II
Post 1998 crisis peak
9
Persistency of interbank relationships
10
Flight to quality in crisis time
Top lenders shift to large debtors in times of
crisis
11
Financial crises and bank healthCapital versus
liquidity
1998
2004
12
Traditional methodology
where yij the gross exposure of bank i to bank
j and ci the capital of bank i.
13
The capital channel (passive banks)

• A bank fails if the funds lost because the
  failure of debtor banks exceed her capital

14
The liquidity channel

• Consider the following dataset

where yij the gross exposure of bank i to bank
j li the net liquidity position of bank i
15
The liquidity Channel

• Define the net exposure on the interbank market
  NEi
• Then we can define the liquidity channel
• A bank fails if its net liquidity lt net exposure
  Nei
• Only if it is linked to a bank that was affected
  active banks scenario
• If there one bank attacked panic scenario

16
The empirical literature

• Empirical work on the capital channel
• Sheldon and Maurer (1998), Swiss banking system.
• Upper and Worms (2002), German banking system
• Furfine (1999) Federal funds market
• Michael (1998) London interbank markets.
• Degryse and Nguyen (2006), Belgian interbank
  market
• But often no bilateral data
• Construct bilateral data from gross exposures
• No link to balance sheet data
• The other transmission channels are neglected

17
The simulations

• We assume a loss given default of 100
• Anecdotal evidence suggests very low recovery
  rates
• We create a initial shock that kills banks
• An idiosyncratic bank shock (Kill each bank once)
  
• Random correlated defaults (10000 simulations/
  month)
• Then calculate the further round effects taking
  into account both channels of contagion
• Capital channel
• Liquidity channel

18
Correlated defaults

• Calculate individual unconditional bank failure
  probabilities using a probit model
• Generate correlated defaults using CR
• Input probability of default from logit model
• Using Bernouilli distribution to draw banks
• In each month 10000 simulations of correlated
  initial defaults as a shock

19
How do we report the results?

• What
• The share of lost banking assets
• The number of failed banks
• We calculated
• The average (but who wants to know the average
  expected damage of an earthquake)
• The worst case scenario (could be a quirk)
• The Value at Risk (95)
• The expected shortfall (95), which is the
  average of the 5 worst cases
• Here we report only the expected shortfall

20
Contagion under different scenarios
1998 crisis
2004 crisis
Added value liquidity channel
21
Contagion with alternative shocks
1998 crisis
2004 crisis
Added value liquidity channel
22
Intermediate conclusion

• The capital channel does not suffice to
  understand systemic crises in the interbank
  market
• The 1998 crisis is somewhat predicted by it
• The 2004 crisis is off the screen
• The liquidity channel is empirically relevant to
  Russia (both active banks and panic scenarios)
• The 1998 crisis is predicted
• The 2004 is also clearly predicted
• Interbank market crises may be not a domino
  effect
• But rather something like a liquidity run
• It may be important in general

23
Next step I does contagion risk help to predict
bank failure?

• Take the active bank scenario
• Rerun the simulations exogenously imposing the
  survival of a banks that failed contagiously.
• Do this for all simulations and all contagiously
  failing banks
• Compare for each bank the new losses to the
  losses of the initial simulation
• Average over simulations
• Result partial contribution to contagion of a
  given bank at a given point in time
• systemic importance or contagion risk

24
Next step I does contagion risk help to predict
bank failure?
Benchmark model with active banks
Panic scenario contagion risk
25
Next step II Is this related to interbank
market structure?

• Theoretical work by Allen and Gale (2000)
• They model the capital channel
• They find that a complete market structure can be
  proven to be the most stable one
• There is some work related to our liquidity
  channel
• Boissay (2006) has a model of financial contagion
  through trade credit
• Illiquid firm may render their suppliers illiquid
  though they were fundamentally solvent
• Empirical work
• Degryse and Nguyen look at interbank market
  structure
• Müller (2003) uses network theory (centrality
  measures)

26
(No Transcript)
27
(No Transcript)
28
(No Transcript)
29
(No Transcript)
30
Centrality as a measure of structure
31
Individual centrality measures
32
Market concentration and contagion
Over time large banks have more positions
But smaller ones
33
Is systemic importance related to centrality?
34
Next step IIIEvaluating the CBRs intervention

• Method use contagion risk simulations
• Analyze what would have happened without CBR
  liquidity injections
• Sberbank and VTB are part of CBR
• Look how the CBR behaved in reality
• Analyze the effectiveness of the behavior
• Could the systemic risk have been better
  contained by targeting different banks?
• Try to allocate the same quantity of liquidity
  and attain lower contagion risk
• Conclusion the CBR did relatively well in saving
  the crisis, but could do more in prevention

35
Evaluating CBR interventions
36
Concluding remarks

• The Liquidity channel
• is relevant to interbank systemic stability in
  Russia
• though its theoretical effects poorly understood
• Interbank market structure
• helps to explain the stability of the interbank
  market
• Helps to explain bank-specific contagion risk
• The CBR
• did not bad in solving the two last crises,
• But may do more in terms of prevention through
  influencing the interbank market structure