A Model of Network Formation for the Overnight Interbank Market - - PowerPoint PPT Presentation

A Model of Network Formation for the Overnight Interbank Market

A Model of Network Formation for the Overnight Interbank Market

Mikhail Anufrieva Andrea Deghib Valentyn Panchenkoc Paolo Pind

a University of Technology Sydney b Università degli Studi di Siena c University of New South Wales, Sydney d Università Bocconi

seminar at Bocconi University, Milan 16 October 2017

1 / 57

A Model of Network Formation for the Overnight Interbank Market

Outline

1

Introduction Interbank Market

2

Model Variables and payoffs Solution Equilibrium Network Configurations Interest Rate

3

Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Policy Implications

Systemic Risk and Network Approach

4

Conclusion

2 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction

Plan

1

Introduction Interbank Market

2

Model Variables and payoffs Solution Equilibrium Network Configurations Interest Rate

3

Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Policy Implications

Systemic Risk and Network Approach

4

Conclusion

3 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction Interbank Market

Overnight interbank market

unsecured interbank loans to balance short-term liquidity typically over one night bilateral agreements seting quantity and interest rate

• ver-the-counter (for US Fed funds) or transparently

centrally organized (e.g., e-MID) directly related to Monetary Policy and financial stability

helps banks to manage their reserves, offset liquidity shocks CBs target prevalent interbank rate freeze of IB market during GFC negatively impacted financial system

reveals information about what banks expect about credibility of other banks and overall market conditions

4 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction Interbank Market

e-MID: electronic Market for Interbank Deposit

5 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction Interbank Market

Theoretical Literature of Overnight Markets

Poole (JF, 1968)

corridor systems with reserve requirements banks are uncertain about their reserves

Afonso and Lagos (ECTA, 2015)

banks meet at random and may negotiate profitable trade if opportunity arises banks are certain about targeted reserve balances and there is no risk of default intermediation emerges in the equilibrium

Heider, Hoerova, Holthausen (WP, 2009)

version of Diamond and Dybvig (1983) model competitive market in presence of a counter-party risk due to asymmetry in information market can break down with banks hoarding cash

6 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction Interbank Market

One day Lending/Borowing on eMID

7 / 57

A Model of Network Formation for the Overnight Interbank Market Introduction Interbank Market

This Model:

banks face uncertainty about their reserves and have an

• ption to access the interbank market

banks form bilateral agreements about loan size and interest rate there is a default risk concept of pairwise stable equilibrium network there will be no intermediation in the equilibrium and a possibility of market breakdown

8 / 57

A Model of Network Formation for the Overnight Interbank Market Model

Plan

1

Introduction Interbank Market

2

Model Variables and payoffs Solution Equilibrium Network Configurations Interest Rate

3

Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Policy Implications

Systemic Risk and Network Approach

4

Conclusion

9 / 57

A Model of Network Formation for the Overnight Interbank Market Model

The model

10 / 57

A Model of Network Formation for the Overnight Interbank Market Model

Intuition: a trading network

Consider any connected network of trading opportunities

• ver some goods.

In equilibrium every node will have the same marginal rate of substitution between goods: sort of contagion? Even more so if the network is endogenous: and it will be generically connected

11 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Environment

We model one day (night) Central Bank and N banks Central Bank maintains a Corridor System: Ti reserve holdings on bank i rp penalty rate at which any bank can borrow from the Central Bank rd deposit rate of the Central Bank rp > rd Banks manage their reserves (“cash”) and use overnight market for lending and borrowing

12 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Variables

Exogenous variables si initially expected reserve holdings of bank i Fi CDF of an error made in prediction of the reserves si + εi actual reserve holdings (net of interbank market) qi probability of default on interbank loan of bank i (known to all) Endogenous variables Banks lend money to each other to achieve Ci expected reserve holdings of bank i afer interbank trade As a result banks create a directed network g = (L, r) with L lending matrix, ℓij ∈ R+ is an amount lent by i to j for each ℓij > 0 we have an interest rate rij ≥ 0

13 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Reserve Holdings

With interbank lending: projected reserves of bank i are Ci(g) ≡ si −

• k∈N ℓik
• Lending

+

• m∈N ℓmi
• Borrowing

. actual reserve holdings of bank i (si + εi) −

• k∈N ℓik +
• m∈N ℓmi = Ci(g) + εi

14 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Expected Payoffs

Expected payoff with the Central Bank: πCB

i

= − Ti−Ci

−∞

• Ti−(Ci+εi)
• rpdFi(εi)+

∞

Ti−Ci

• (Ci+εi)−Ti
• rddFi(εi)

Expected marginal rate of bank i from CB transactions is Wi := rd + (rp − rd) Fi(Ti − Ci) .

15 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Expected Payoffs: Interbank market

Banks can enter the interbank market: borrowers can get lower interest rate lenders can get higher interest rate

Assumption (Exogenous Default)

Bank i has a probability of default on loan, qi ∈ [0, 1]. It is independent of banks’ reserves and known to every bank. Defaults of different banks are independent events.

Assumption (Consequences of Default)

If a borrower defaults, then the lender will not receive its loan (neither principal nor interest) back. If a lender defaults, the borrower still has to pay the principal and loan.

16 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Expected Payoffs: Interbank market

Assumption (Consequences of Default)

If a borrower defaults, then the lender will not receive its loan (neither principal nor interest) back. If a lender defaults, the borrower still has to pay the principal and loan. πIM

i

=

• k∈N

rikℓik(1 − qk)

• interest due from

repaying borrowers

−

• k∈N

ℓikqk

• losses because of

defaulted borrowers

−

• m∈N

ℓmirmi

• interest due to

lenders

17 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Expected Payoffs: Summary

Bank maximizes the total expected profit: qiB + (1 − qi)πi, or, equivalently, πi = πCB

i

+ πIM

i

= = − Ti−Ci

−∞

• Ti − (Ci + εi)
• rpdFi +

∞

Ti−Ci

• (Ci + εi) − Ti
• rddFi

+

• k∈N

rikℓik(1 − qk) −

• k∈N

ℓikqk −

• m∈N

ℓmirmi Expected marginal rate of bank i from CB transactions is Wi := rd+(rp−rd) Fi(Ti−Ci) = rp Fi(Ti−Ci)+rd 1−Fi(Ti−Ci)

• ,

from borrowing −rmi from lending rik(1 − qk) − qk

18 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Pairwise Stable Networks

For every possible network g, banks can compute their total expected profits, πi. Which network do we expect to see?

19 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Pairwise Stable Networks

For every possible network g, banks can compute their total expected profits, πi. Which network do we expect to see?

Definition

Network g = (L, r) is a pairwise stable equilibrium if: a pair of banks (a borrower and a lender) set the quantity/rate (L, r) of their loan such that it is Pareto

• ptimal for them given the rest of the network

if L = 0, there no link and r is not defined

19 / 57

A Model of Network Formation for the Overnight Interbank Market Model Variables and payoffs

Pairwise Stable Networks

In pairwise stable equilibrium g = (L, r): 1). no single bank can remove an existing link in which it is involved and be beter off; 2). no couple of banks with link ℓij > 0, can change their agreement to (˜ ℓij,˜ rij) (keeping the rest of the network the same) making one of the banks beter off without making another bank worse off; 3). no couple of banks without link can find an agreement with ˜ ℓij > 0 that would make both banks beter off. This includes the definition of Jackson and Wolinsky (1996): In pairwise stable equilibrium no pair of banks has an incentive to create a new link no single bank has an incentive to remove an existing link

20 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Solution

21 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Payoff Improving Links

when link ij is added to g, we calculate payoff improvement ∆g+ij

i

(ℓ, r) = πi(g + ij) − πi(g) and ∆g+ij

j

(ℓ, r) = πj(g + ij) − πj(g) define feasibility sets for lender FL

i→j(g) =

• (ℓ, r) : ℓ > 0, r ≥ 0, ∆g+ij

i

(ℓ, r) > 0

• and for borrower

FB

j←i(g) =

• (ℓ, r) : ℓ > 0, r ≥ 0, ∆g+ij

j

(ℓ, r) > 0

• 22 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Feasibility Sets

1 2 3 4 0.02 0.03 0.04 0.05 0.06 l r

23 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Properties of Feasibility Sets

1

feasibility set of the lender lies above function hL : R+ → R: FL

i→j(g) =

• (ℓ, r) : ℓ > 0, r > hL(ℓ)
• .

Function hL is strictly increasing and hL(0) = qj + Wi(0) 1 − qj .

2

feasibility set of the borrower lies below function hB : R+ → R: FB

j←i(g) =

• (ℓ, r) : ℓ > 0, r < hB(ℓ)
• .

Function hB is strictly decreasing and hB(0) = Wj(0) .

24 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Proposition (Link Creation)

Consider two banks i and j with ℓij = 0 in the network. Loan from i to j that makes both banks beter off exists if and only if qj + Wi(0) < (1 − qj)Wj(0). if qj = 0 (no default risk), then we have Wi(0) < Wj(0) ⇔ Fi(Ti − Ci) < Fj(Tj − Cj)

25 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Pareto Efficiency in Bargaining

Network g = (L, r) is a pairwise stable equilibrium if: 2). no couple of banks with link ℓij > 0, can change their agreement to (˜ ℓij,˜ rij) (keeping the rest of the network the same) making one of the banks beter off without making another bank worse off;

26 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Pareto Efficiency in Bargaining

Network g = (L, r) is a pairwise stable equilibrium if: 2). no couple of banks with link ℓij > 0, can change their agreement to (˜ ℓij,˜ rij) (keeping the rest of the network the same) making one of the banks beter off without making another bank worse off; −

∂πi ∂ℓij/∂πi ∂rij = − ∂πj ∂ℓij/ ∂πj ∂rij

• qj + rd + (rp − rd)Fi(Ti − Ci) = (1 − qj)
• rd + (rp − rd)Fj(Tj − Cj)
• qj + Wi = (1 − qj)Wj .

26 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Contract Curve

1 2 3 4 0.02 0.03 0.04 0.05 0.06 l r

27 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Contract Curve

Proposition (Determination of Loan Amount)

Consider an arbitrary network g, and any two banks i and j with non-empty intersection of feasibility sets FL

i→j with FB j←i.

Then there exists a unique ℓij > 0 that is consistent with requirement 2 from Definition of the pairwise stable equilibrium.

28 / 57

A Model of Network Formation for the Overnight Interbank Market Model Solution

Contract Curve

Proposition (Determination of Loan Amount)

Consider an arbitrary network g, and any two banks i and j with non-empty intersection of feasibility sets FL

i→j with FB j←i.

Then there exists a unique ℓij > 0 that is consistent with requirement 2 from Definition of the pairwise stable equilibrium.

Corollary

In every pairwise stable network, for any two banks i and j with a link ℓij > 0, it holds that qj + Wi = (1 − qj)Wj.

28 / 57

A Model of Network Formation for the Overnight Interbank Market Model Equilibrium Network Configurations

No two-sided contracts

A situation like this requires: Wi = (1 − qj)Wj − qj and Wj = (1 − qi)Wi − qi In a Pairwise Stable Equilibrium banks i and j cannot borrow to and lend from each other at the same time.

29 / 57

A Model of Network Formation for the Overnight Interbank Market Model Equilibrium Network Configurations

No loops

A situation like this requires: Wi = (1 − qj)Wj − qj ≤ Wj ≤ · · · ≤ Wi In a Pairwise Stable Equilibrium with at least one bank with a positive default risk, there are no directed loops.

30 / 57

A Model of Network Formation for the Overnight Interbank Market Model Equilibrium Network Configurations

No intermediation

A situation like this requires: Wi = (1 − qj)Wj − qj ≤ Wj = (1 − qk)Wk − qk so i must also lend to k and then Wj = Wk. In a Pairwise Stable Equilibrium the bank with positive default risk, cannot be an intermediator.

31 / 57

A Model of Network Formation for the Overnight Interbank Market Model Equilibrium Network Configurations

No separate components

A situation like this with Wi > Wk requires: Wk < Wi = (1 − qj)Wj − qj so k must also lend to j but then Wk = (1 − qj)Wj − qj‼! In a Pairwise Stable Equilibrium the separate components cannot coexist unless all lenders have the same W.

32 / 57

A Model of Network Formation for the Overnight Interbank Market Model Equilibrium Network Configurations

Result

Proposition (Characterization of PSE)

A Pairwise Stable Equilibrium is such that banks are partitioned in three groups: isolated banks, borrowers and lenders. Borrowers and lenders form generically a unique component where all directed paths have at most length one.

33 / 57

A Model of Network Formation for the Overnight Interbank Market Model Interest Rate

Contract Curve

What interest rate our model can predict?

1 2 3 4 0.02 0.03 0.04 0.05 0.06 r

34 / 57

A Model of Network Formation for the Overnight Interbank Market Model Interest Rate

Competitive PSE

Definition

Network g = (L, r) is a competitive pairwise stable equilibrium if it is the pairwise stable equilibrium and on every existing link ij, the amount ℓij > 0 maximizes the profits

• f both lender i and borrower j given the interest rate rij.

FOC for lending bank i: rij(1 − qj) = qj + rd + (rp − rd) Fi(Ti − Ci) FOC for borrowing bank j rij = rd + (rp − rd) Fj(Tj − Cj)

35 / 57

A Model of Network Formation for the Overnight Interbank Market Model Interest Rate

Competitive PSE

1 2 3 4 0.02 0.03 0.04 0.05 0.06 l r

36 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions

Plan

1

Introduction Interbank Market

2

Model Variables and payoffs Solution Equilibrium Network Configurations Interest Rate

3

Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Policy Implications

Systemic Risk and Network Approach

4

Conclusion

37 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Characteristics of e-MID market

Year Banks Var(rlen) Var(rbor) Prop 2 loops Prop 3 loops 1999 215 0.0030 0.0027 0.0140 0.0124 2000 196 0.0034 0.0025 0.0105 0.0079 2001 185 0.0037 0.0028 0.0163 0.0087 2002 177 0.0032 0.0024 0.0205 0.0077 2003 179 0.0031 0.0023 0.0174 0.0095 2004 180 0.0029 0.0022 0.0145 0.0064 2005 176 0.0029 0.0021 0.0175 0.0083 2006 177 0.0027 0.0020 0.0133 0.0074 2007 178 0.0029 0.0021 0.0095 0.0035 2008 173 0.0034 0.0024 0.0058 0.0016 2009 153 0.0035 0.0024 0.0043 0.0000 Overall 350 0.0032 0.0024 0.0131 0.0067

38 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Network

39 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Craig and Von Peter (JFI, 2014) Conditions for an ideal CP structure:

1

Core banks lend to each other

2

Periphery banks do not lend to each other

3

Core banks lend to at least one periphery bank

4

Core banks borrow from at least one periphery bank

40 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

One year Lending/Borowing on eMID

41 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Division on Core and Periphery

42 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Theoretical Models of Core-Periphery

Goyal and Vega-Redondo (2007, JET)

middleman gets a surplus

van der Leij et al (2014, WP)

2-stage: undirected “potential” relationships are formed trading over t = 1, 2, . . . , ∞ due to liquidity shocks core-periphery may occur under heterogeneity

Babus and Hu (2016, JFE) Farboodi (2015, WP) Castiglionei and Navarro (2016, WP) Wang (2016, WP) Galeoti and Goyal (2010, AER)

43 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Extension 1: Repeated Version

Idea: Banks play the same one-shot game every day, but their si’s differ from day to day. All other parameters are the same. Assume that there are two types of banks: “large” and “small”. Large banks have many operations and, therefore, the intertemporal variance of s’s is large. Small banks have few operations and, therefore, the variance of s’s is low. Intuition: Large banks will have larger needs for/excess liquidity more ofen than small banks.

44 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Repeated Version: Simulation

Simulation: two Large and three Small banks Ti = 0, means of si is 0, Fi ∼ N(0, 1) variance of si for L is 3 and for S is 1 qi is constant over time and is 0.001 L1 L2 S1 S2 S3 L1 311 260 266 275 L2 299 270 284 284 S1 289 293 147 151 S2 283 302 115 115 S3 283 304 133 125

45 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Extension 2: Endogenizing Default Rate

Assumption (Exogenous Default)

Bank i has a probability of default on loan, qi ∈ [0, 1]. It is independent of banks’ reserves and known to every bank. Defaults of different banks are independent events. Idea: The same seting as before, except that q is subjective: Instead of qk, bank i has own estimation, qik.

46 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Extension 2: Endogenizing Default Rate

Assumption (Exogenous Default)

Bank i has a probability of default on loan, qi ∈ [0, 1]. It is independent of banks’ reserves and known to every bank. Defaults of different banks are independent events. Idea: The same seting as before, except that q is subjective: Instead of qk, bank i has own estimation, qik. Can be interpreted as “cost from transaction”: πIM

i

=

k∈N rikℓik − k∈N qik(1 + rik)ℓik − m∈N ℓmirmi

and then it also includes the research cost about the borrower k, trust, etc.

46 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Extension 2: Endogenizing Default Rate

Assumption (Exogenous Default)

Bank i has a probability of default on loan, qi ∈ [0, 1]. It is independent of banks’ reserves and known to every bank. Defaults of different banks are independent events. Idea: The same seting as before, except that q is subjective: Instead of qk, bank i has own estimation, qik. Can be interpreted as “cost from transaction”: πIM

i

=

k∈N rikℓik − k∈N qik(1 + rik)ℓik − m∈N ℓmirmi

and then it also includes the research cost about the borrower k, trust, etc. qik gets reduced afer bank i transacted with k

46 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Repeated Version: Simulation

Simulation: two Large and three Small banks Ti = 0, means of si is 0, Fi ∼ N(0, 1) variance of si for L is 3 and for S is 1 qik,t = qik,t−1(1 − 0.005 × Itrans), with qik,0 = 0.001 L1 L2 S1 S2 S3 L1 276 294 302 317 L2 272 282 257 280 S1 312 290 77 47 S2 294 270 66 57 S3 314 299 52 77

47 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Empirics of Interbank Networks

Repeated Version: Simulation

Simulation: two Large and three Small banks Ti = 0, means of si is 0, Fi ∼ N(0, 1) variance of si for L is 3 and for S is 1 qik,t = qik,t−1(1 − 0.005 × Itrans), with qik,0 = 0.001 L1 L2 S1 S2 S3 L1 276 294 302 317 L2 272 282 257 280 S1 312 290 77 47 S2 294 270 66 57 S3 314 299 52 77 transactions P-P are substituted by P-C network exhibits disassortative mixing (Iori et al, 2008)

47 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Policy Implications

48 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Policy Implications

equal changes in the rates rp and rd (i.e., shifs of the corridor) do not affect the volume, but will increase the

• vernight market rate

the difference between rp and rd (i.e., size of the corridor) does affect the volume (higher size higher volume) increase in reserve requirements Ti will result in higher the overnight market rate increased uncertainty about reserves, σ, will result in a smaller number of transactions higher default probability qi will decrease the transaction volume and may lead to market freeze

49 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Systemic Risk and Contagion

Systemic Risk – risk of experiencing an event that threatens the well functioning system (payments, banking, financial). Mechanism of domino theory : initial (macroeconomic) shock contagion due to interdependence of financial institutions via financial or trade interlinkages (i.e., intermediation) Different from popcorn theory : bad conditions of economic system cause multiple (independent) failures

50 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Domino Theory

GFC led to an increased interest in the relationship between systemic risk and network effects

“In the current crisis, we have seen that financial firms that become too interconnected to fail pose serious problems for financial stability and for regulators. Due to the complexity and interconnectivity of todays financial markets, the failure

• f a major counterparty has the potential to severely disrupt

many other financial institutions, their customers, and other markets.” – Charles Plosser, 03/06/09 “Complex interactions among market actors may serve to amplify existing market frictions, information asymmetries, or

• ther externalities.”

– Janet Yellen, 01/04/13

51 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Popcorn Theory

“When popcorn is made, oil and corn kernels are placed in the botom of a pan, heat is applied and the kernels pop. Were the first kernel to pop removed from the pan, there would be no noticeable

• difference. The other kernels would pop anyway because of the
• heat. The fundamental structural cause is the heat, not the fact that
• ne kernel popped, triggering others to follow.

Many who believe that bailouts will solve Europe’s problems cite the

• Sept. 15, 2008 bankruptcy of Lehman Brothers as evidence of what

allowing one domino to fall can do to an economy. This is a misreading of the historical record. Our financial crisis was mostly a popcorn phenomenon. At the risk of sounding defensive (I was in the government at the time), I believe that Lehman’s downfall was more a result of the factors that weakened our economic structure than the cause of the crisis.” – Edward Lazear

52 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Network Literature on Contagion

Contagion starts with a local shock and spreads via network Allen and Gale (2000, JPE): equilibrium model where banks can place deposits to other banks and suffer from unexpected liquidity demand Elliot, Golub and Jackson (2014, AER): firms cross-own assets and suffer from discontinuous drop of value when the value hits a lower threshold Cabrales, Gotardi, and Vega-Redondo (2014): firms exchange the assets and suffer from a drop in asset value Gai and Kapadia (2010, PRSA), Glasserman and Young (2015, JBF); Acemoglu, Ozdaglar, and Tahbaz-Salehi (2015, AER): interbank network of credit-lending relationship Network is exogenous in all these contributions

53 / 57

A Model of Network Formation for the Overnight Interbank Market Discussion and Extensions Policy Implications

Popcorn vs. Domino Theory

Bipartite network is very stable from systemic point of view. Core-Periphery is unstable. qi,t = qt + q∗

i,t

Model can generate popcorn effect: if banks start to perceive high market risk qt’s, the network will have less links. Can we get a contagion? generally yes, by modeling changing correlations between qi’s, but this is exogenous to the model

54 / 57

A Model of Network Formation for the Overnight Interbank Market Conclusion

Plan

1

Introduction Interbank Market

2

Model Variables and payoffs Solution Equilibrium Network Configurations Interest Rate

3

Discussion and Extensions Empirics of Interbank Networks

Core-Periphery Structure

Policy Implications

Systemic Risk and Network Approach

4

Conclusion

55 / 57

A Model of Network Formation for the Overnight Interbank Market Conclusion

Network Theory of Financial Institutions

Key Qestions:

1

What are possible propagation mechanisms (transmission channels) of shocks?

2

What is the structure of the relevant network?

3

Given this structure, how large is the systemic risk? Which institutes are systemically important? Is high density good or bad for the systemic risk?...

4

Why does a given structure emerge?

56 / 57

A Model of Network Formation for the Overnight Interbank Market Conclusion

Network Theory of Financial Institutions

Key Qestions:

1

What are possible propagation mechanisms (transmission channels) of shocks?

2

What is the structure of the relevant network?

3

Given this structure, how large is the systemic risk? Which institutes are systemically important? Is high density good or bad for the systemic risk?...

4

Why does a given structure emerge? This Project: Endogenous network formation model for overnight lending interbank market Comparison with the e-MID market

56 / 57

A Model of Network Formation for the Overnight Interbank Market Conclusion

Extension and Future Work

Paper so far:

• ne-period model of banks deciding both amount and

interest bi-partite network in equilibrium, generically connected Core-Periphery is generated by aggregation Possible model extensions: careful modeling of “cost” of lending including possible loss, counter-party research cost, preferential relationships systemic risk due to appearance of intermediaries alternative notions of network stability endogenizing q’s (?)

57 / 57