Agent-based Networks of Corporate Lending
Grzegorz Ha laj
European Central Bank
26/03/2015
Parts based on research with U. Kochanska (ECB) and Ch. Kok (ECB)DISCLAIMER: This presentation should not be reported as representingthe views of the European Central Bank (ECB). The views expressed arethose of the authors and do not necessarily reflect those of the ECB
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 1 / 30
MotivationRecent financial crisis: loss of trust on the interbank market; concernsabout failure of one of the key players spreading contagion; smallshocks with detrimental effectsA response from regulators: measures to mitigate the risk ⇒ highercapital standards + reducing bilateral exposures
I Large Exposure limits;I Credit Valuation Adjustment to unlock the risk in OTC exposures and
immediately reflect it in the capitalI Standard settlement practices (CCP framework)I ...but usually only interbank market modelled → a large part of the
network is neglectedOur aim:
I fill the gap in the literature to improve understanding of:F linkages (and emergence of links) between banks and the real economy
(non-bank corporate sector)F risk stemming from interconnectedness
Approach: modelling of banks’ reactions to these measures and to thechanging macroeconomic environment with links to corp sector(combining risk/return trade-offs, funding conditions...)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 2 / 30
Outline
Modeling framework – agent-based interbank+corporate networks
Four round model – endogenous formation
Interbank augmented by non-bank corporate sector (called: firms)1 offers of interbank placements based on individual optimisation of
interbank asset structures2 funding diversification3 negotiation phase: matching offers and preferred funding structure in a
bargaining game4 price (i.e. interest rate) adjustment (if demand 6= supply)
Scope for application
stress tests and dynamic balance sheet tool
assessing network effects of credit provision to the real economy(shocks from corporate sector)
parametrisation of LE and concentration limits (so far only forinterbank) and sector RWs
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 3 / 30
Literature – towards network formation
Networks in other research areas: game theory of Jackson andWolinsky (1996)
Extensions in finance – exogenous networks: game theory – optimalresponses of banks to shocks to incentives to lend Cohen-Cole (2011);Bluhm et al. (2013). Acemoglu et al. (2013): dealing with socialinefficiency of financial networks; Georg (2011) models interbankexposures as residuals of banks’ investment activities (but networkssimply drawn from a distribution)
Jackson and Watts (2002) combine stochastic games and matchingproblems to study general principles of network formation ineconomics; Acemoglu et al. (2014) create the interbank structurebased on equilibrium of lending contracts and repayments
Agent-based approach to address overly complex equilibria – Markose(2012); Grasselli (2013)
Matching (Chen, 2013); (Duffie and Sun 2012) and price formation(Eisenschmidt, 2009) ⇒ mechanisms important for us
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 4 / 30
Formation of the lending network – Endogenous networks
The aim of the project is to:
1 understand foundations of the topology of lending networks in theeconomy
2 capture sensitivity of the interbank network structures to theheterogeneity of banks (in terms of size of balance sheet, capitalposition, general profitability, counterparty credit risk) and thechanges of market and bank specific risk parameters
3 provide a framework to assess effectiveness of rule designed tomitigate systemic risk on the interbank system (esp. pertaining tocapital requirements, size and diversity of interbank exposures)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 5 / 30
4 round model – outlineThe following 4 rounds are repeated until 'all interbank assets of apredefined volume are invested (separate for interbank and bank-firmnetwork)
1 Firms make loan offers to other banks and firms which are drawnfrom a probability map: offers based on optimisation of theirinterbank asset structures and corporate lending portfolio
2 Firms formulate their preferred structure of interbank (banks) andbank (firms) funding from banks drawn in round 1: based on thediversification of the funding (rollover) risk
3 Firms enter negotiation phase: bargaining game in order to try tomatch the preferred allocation of the assets and the preferredstructure of interbank (bank) funding
4 Firms reconsider their pricing offers: firms with open funding gapincrementally adjust their offers of interest payments on new loan(optional feature, not used so far in the exposition)
At each step, assets are “matched” with liabilities incrementally
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 6 / 30
Figure 1: The sequential four round procedure of the interbank formation(formation of bank-firm links separate but analogous)
2) OPTIMISATIONPreferred asset
structure
1) OPTIMISATIONPreferred funding
structure 4) PRICEInterest rate adjustment
3) BILATERAL GAMESBargaining game
STEPS Repeateduntil all IB assetsare allocated
Unallocated IB assets and liabilities
NEW PLACEMENTSPart of unallocated IB assetsplaced in banks as deposits
creating IB linkages
4 ROUNDS
REPEATED
STEPS
Next step
Full allocation IB Network Completed
Partial allocation
INITIAL PARAMETERSAggregate IB lending / borrowing, capital, RWA, CDS
spreads, market interest rates
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 7 / 30
Prerequisites
(nodes) N banks and M non-bank firms: capital and bank borrowing+ out-degree distribution within (NACE) sectors
(exposures) Let Lij denotes the interbank (bank) placement (loan) ofbank j in bank (firm) i .
(capital position – constraint for risk-taking) total capital e andcapital e I ≤ e allocated to the interbank assets, eC ≤ e allocated tonon-bank firms; risk weights ω of exposures.
(probability map P) of interbank and bank-firm connections drawnfrom P allowing for capturing possible customer relationship betweenbanks and firms. Each bank j draws its counterparties Bk
j ⊂ N/{j},enlarging the set at each step k : Bk+1
j = Bkj ∪ Bk+1
j ;In addition, firms choose max number (mj) of banks granting loansbased on out-degree distribution, i.e. #Bk+1
j ≤ mj
(matching) at step k incremental matching of assets and liabilities:akj = ak−1
j −∑
i Lkij , where Lk is a matrix of placements at step k
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 8 / 30
1st round – Criteria for investment of interbank assets
General idea of banks’ optimising behaviour
Assumption (i): each bank maximises return from loan portfolio adjustedby risk related to interest rates and counterparts’ defaults (with apredefined risk aversion parameter) and taking into account customerrelationship, i.e. a drawn sample of banks and firmsAssumption (ii): optimisation of interbank portfolio separate fromoptimisation of non-bank corporate loan portfolio
Each bank maximises the following function of its interbank exposurebreakdown:
J(L1j , . . . , LNj) =∑i∈Bk
j
riLij − κj(σ ∗ L>·j )>Q(σ ∗ L·j) (1)
Outcome: a matrix of exposures LI ,k , whereby optimisation subject toconstraints...
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 9 / 30
...Constraints of the admissible set of strategies
The maximisation is subject to some feasibility and capital constraints.
1 budget constraint:∑
j |j 6=i Lij = akj and Ljj = 0, for a0j = aj being
exogenously determined;
2 counterpart’s size constraint: Lij ≤ lki ;
3 capital constraint:∑
i |i 6=j ωi (Lkij + Lij) ≤ e Ij − γ>(L·j + L·j);
4 large exposure limit constraint: Lij ≤ χej .What if the constraints are too stringent for a bank j? ⇒ bank j reducesits interbank lending and (technically) the optimisation is solved for akjreplaced by aki − 2∆aki , aki − 3∆aki ,... until aki − ki∆aki gives a feasible setof constraints
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 10 / 30
2nd round – funding diversification
Diversification risk gauged by default risk
Xj : =
{0 with probability pj1 with probability 1− pj
(2)
Assumption: pjs are risky (variance based on time series of CDS spreads)For a covariance matrix D2
X of X , the optimised funding risk is measured
F (Lki1, . . . , LkiN) = κF [Lki1 . . . LkiN ]D2
X [Lki1 . . . LkiN ]> (3)
Outcome: a matrix of interbank deposits LF ,k , whereby optimisation onthe admissible set:AF
i : = {y ∈ RN+|j ∈ Bk
j ⇒ yj ≤ akj and j 6∈ Bkj ⇒ yj = 0}.
REMARK: inclusion of non-bank corporate sector implies that (3) is alsosolved by non-bank firms (⇒ LF ,k is (N + M)× (N + M) matrix)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 11 / 30
3rd round – the game
Assumption: banks negotiate loans in pairs simultaneously (pair (i ′, j)knows the outcome of (i ′′, j) after both games are completed). Case
LI ,kij > LF ,kij
G kij (x) =
[U l ,k∗ij − s l ,kij · (x − LF ,kij )
] [Ua,k∗ij − sa,kij · (L
I ,kij − x)
](4)
where s l ,kij is a measure of how much bank i is willing to deviate from hisoptimal funding strategy, i.e.
s l ,kij = max
(U l ,kij (LF ,kij )− U l ,k
ij (LI ,kij )
|LI ,kij − LF ,kij |, 0
),
where U l ,kij (x) = −F (LF ,ki1 , . . . , LF ,kij−1, x , L
F ,kij+1, . . . , L
F ,kiN )
(for sa,kij analogously,... and for LI ,kij < LF ,kij similar)
Goal of the game: maximisation of G kij ; outcome → LG ,I ,k and LG ,F ,k
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 12 / 30
3rd round – correction to the game
Observation: the outcome of the game may not be consistent with thebudget constraint of the given size of banks’ aggregate interbank portfolios(i.e. ak and lk)
Correction
if∑
i LG ,I ,kij > akj then LG ,I ,kij := LG ,I ,kij ∗ akj∑
i LG ,I ,kij
if∑
j LG ,F ,kij > lki then LG ,F ,kij := LG ,F ,kij ∗ lki∑
j LG ,F ,kij
LG ,k = min(LG ,I ,k , LG ,F ,k) (element-wise)
Aggregation of the step k with outcomes of steps 1, . . . , k − 1
Lk+1 = Lk + LG ,k
Update: e I ,k+1, ak+1 and lk+1
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 13 / 30
4th round – price adjustment [optional]
After the first 3 rounds of a step k some banks may still have a gap inthe interbank funding ⇒ adjustment to the offered interest rate onnew interbank deposits to increase a chance to obtain funding in stepk + 1
If at the step k + 1 the gap amounts to gk+1i : = li −
∑j L
k+1ij then
the adjusted offered rate satisfies rk+1i = rki exp(αgk+1
i /li ).
REMARKS
in the baseline case we assume α = 0.25
no clearing mechanism implemented – would be good to have it sinceprice adjustment mechanism does not prove to be fully effective (asequential 4-round model with only one step and price adjustmentdoes not lead to full allocation of assets / liabilities)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 14 / 30
Sensitivity analysis based on a stylised setup
Assumptions
100 banks (bank ∈ {1, . . . , 100})uniform probability map (5% probability of connection for all pairs ofnodes)
uniform return (0.05) and investment risk (0.1) and funding risk (0.1)parameters
no correlation
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 15 / 30
Scenario 1: Risk of investment on the interbank becomeheterogenous (part 1)
x-axis: banks
y-axis: betweennesscentrality
light copper linescorrespond to the baselinecase; the darker the colorthe higher heterogeneity ofinvestment risk (σline
bank =
σbasebank + line ∗ dσ ∗ bank
100)
Centrality of more ‘risky’nodes decreases and the(relatively) less ‘risky’increases
Figure 2: Betweenness centrality vs changing riskof the return from interbank investment
0 20 40 60 80 100
0.0080
0.0085
0.0090
Source: own calculations
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 16 / 30
Scenario 1: Risk of investment on the interbank becomeheterogenous (part 2)
x-axis: banks
y-axis: Marginal (of laststep tranche) price
light copper linescorrespond to the baselinecase; the darker the colorthe higher heterogeneity ofinvestment risk (σline
bank =
σbasebank + line ∗ dσ ∗ bank
100)
(Unsurprisingly) price paidby banks that have moreinvestment risk increasessince they must searchlonger for funding.
Figure 3: Marginal price vs changing risk of returnfrom interbank investment
0 20 40 60 80 1000.021
0.022
0.023
0.024
0.025
0.026
0.027
Source: own calculations
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 17 / 30
Scenario 1: Risk of investment on the interbank becomeheterogenous (part 3)
x-axis: banks
y-axis: clusteringcoefficient
light copper linescorrespond to the baselinecase; the darker the colorthe higher heterogeneity ofinvestment risk (σline
bank =
σbasebank + line ∗ dσ ∗ bank
100)
Heterogeneity of risktranslates into uniformshift of number of trianglesassociated with nodes.
Figure 4: Clustering coeff. vs changing risk ofinvestment into interbank assets
0 20 40 60 80 1000.170
0.175
0.180
0.185
0.190
0.195
Source: own calculations
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 18 / 30
Scenario 2: Risk of investment on the interbank becomeheterogenous
x-axis: banks
y-axis: Betweennesscentrality
light copper linescorrespond to the baselinecase; the darker the colorthe higher heterogeneity offunding risk (σline
bank =
σbasebank + line ∗ dσ ∗ bank
100)
Centrality of more ‘risky’nodes decreases and the(relatively) less ‘risky’increases
Figure 5: Betweenness centrality vs changing riskof the funding (roll-over) risk
0 20 40 60 80 100
0.0080
0.0085
0.0090
Source: own calculations
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 19 / 30
Table 1: Overview of data inputs
Item Description Sources
CoverageBanks As identified in 2011 EBA Disclosures; 80 banks from EU countries.
+ 500 randomly generated banks based on TAEBA, Halaj and Kok (2014)+ Bankscope
Non-financial corpo-rations
Members of the benchmark equity indices in the countries coveredby EBA Disclosures and Halaj and Kok (2014); total 700 firms
Bloomberg and ECB
AttributesBanks Total assets, IB assets, securities, securities MtM, equity, CT1 cap-
ital, IB liabilitiesEBA
Banks Loans to non-fin. corporations: calculated by using avg. countryratio of such loans to TA based on the ECB (MFI) balance sheetdataset
ECB calculations
Banks Economic activity code (NACE), CDS of senior debt with 5 maturity,and long-term issuer ratings by Moody’s, Fitch and S&P.
Bloomberg
Non-financial corpo-rations
Total assets, total equity, total liabilities, NACE code, CDS spreadsof senior debt with 5 maturity, and long term ratings by Moody’s,Fitch and S&P.
Bloomberg
Non-financial corpo-rations
Loans from banks: calculated by using the average country ratioof loans to total assets of NFCs based on the ECB EA Accountsdataset.
ECB calculations
Lending relations and other supportive variablesLending relationship Defined as the number of loans with different banks; average figures
by country and NACE sector were applied based on the data providedthrough the Working Group on Credit Registers
ECB calculations
Interest rates onloans by size andcountry
Avg. interest rates on loans by size of loan and by country basedon the ESCB MIR data; categories of loans as follows: (below 0.25EUR mn), (equal or above 0.25-1 EUR mn), and (over 1 EUR mn).
ECB calculations
Expected default fre-quencies
Avg. of expected default frequencies for non-financial corp. bycountry and NACE.
Moody’s KMV and ECB
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 20 / 30
Sampling of the network
Observed nodes (banks + non-bank corporate firms) and +500generated banks
I generated banks: based on the total assets and proportional allocationof other attributes
Lending relationship:I {bank}–{firm}: based on aggregate Credit Register data
F → out-degree distribution (for each NACE sector) → the cardinalnumber of set Bk
j of firms k to which a bank j grants loans isconstrained by a number mj drawn from the out-degree distribution,i.e. #Bk
j ≤ mj
+F → probability that a bank in a given country lends to a firm from a
given country and a given (NACE) sector
I {EBA sample bank}–{EBA sample bank}: EBA disclosuresI {small bank}–{EBA sample bank}: arbitrary [small] probability of
connection (= 0.01)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 21 / 30
Figure 6: Network of non-bank corporate borrowing
12
345678910
111213141516171819202122
232425
2627282930
31323334
35 36
373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276
277278
279280
281282
283284285286287288289290291
292293
294295
296297
298299300301302303
304305
306307
308309310311312313314315316
317318
319320321322323
324
325
326327
328329330331332
333334
335
336
337
338
339
340341
342343344345346
347348
349
350
351
352
353
354355
356
357
358
359 360
361
362
363364
365366
367368369370371372373374375
376377378379380381382383
384385386387388389390391392393394395396
397398
399400
401402
403404
405406407408409410411412413414
415416
417418
419420421422423424425426
427428
429430
431432433434435436437438439440
441442
443444445446447448
449450
451
452 453
454
455456
457458
459
460461
462
463
464465
466
467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506
507
508
509510
511512513514515
516517518
519520521522523524525
526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558
559
560
561
562
563
564565566567
568569
570571
572573574575
576
577578579
580581582
583584585586
ADS
ALV
BAS
BAY
BMWBEICONDAI
DB1
LHA
DPW
DTE
EOA
FME
FRE
HEI
HEN
IFX
SDFLXS LIN MRK
MUV
RWE
SAP
SIE
TKA
VOW
ABE
ANA
ACS
AMS
MTSDIAEBRENG
FER
FCC
GAM
GAS
GRF
IBE
ITX
IDR
IAG
JAZMAP TL5 OHL
REE
REP
SCY
TRE
TEF
VIS
AC AI
AIRALU
ALOMT
CS EN CAPCA SGOML BN
EDFEI
GSZGTOKEROR LG LR MC ORA
RI PUBRNOSAFSANSU SOL
TECFP
UL VK VIEDG VIV
A2ASTSG I
ATLAGLAZMBPEPMIBZUCNHCPREGP
ENEENI
EXOF I
FNCGTKLUXMS MB MED
PC PRYSPMSFESRGSTMTITTENTRNTOD
US WDFYOO
AGNAF
AKZMT ASMCORFURGTO
HEIAH
DSMKPNPHIPNL
RANRENRDSIM SBMTNT
UL UNAWKLZIG
ALTBES
BANMPI
CPRCFN
COMCORCTTEDPEDPESORAMFCPGAL
GLIMCP
SCOIBS
GPAIPRINAJMTLIGMAREGLNBA
PTIPTC
REDRENSVASEMORESONSONSONSNC
SLBSCP
SUCTDSSCTVAFVAFZON
ABBAERAEX
YZAGCCCGNCPLCRHDLEDCPDGOFOIFBDFDPFFYGLB
GWMGRN
HBRIFPINMIR5KDRKMRKYGKSPMERMIOOGNORMOVGPWLIPMPCIPTRPACPRPRYA
SKGTOTTVCUTVZMN
ACKAGS
ABIBEFBELCOFCOLDIE
DELDL ELIGSZGBLBEKSOLTNE
THRUCBUMI
APAMT BIP
FOYKBC
LXMREI
RTLSESSOC
SFN
AFAAFEAHLAKTAKTALN
AMEAPE
ASUACGATR
ALBALB
BASBIO
BTHCPMCGCCAVCNCCTYCTHCTLCRADIGDOVEFOELEEBCELIENDEQVETTEXLFSCFIAFLGFISFUMGLAHKSHONHUHILKICPIFAXNSKRAKSLKESKES
KELKNE
KCRLAT
LEMMMOMARMETMETMEOMUNNEONESNOKNRENDANLGOKMOLVORAOREOKDOKDORNORNOUTOTEPNAPKCPKKPONPOYQPRRAIRMRRAPRTRRUTRESREGSAGSAMSAASCLSCISTQSOPSOSSDASRVSSHSTCSTCSTESTESSKSUYTAMTTMTLVTPSTEMTLTTLS
TIETIKTRHTULTUTUPMUNRWATVACVAIVALVIKWRTWUFYTYYLE
AND
CAICWI
EVNIIALNZ
MMK
POS
OMV
RHI
SBOSTR
TKAVERVIG
VOE
WIE
ZAG
AEGIAT
EYDTATOTOCENEEEHYGDROELLELTELBEUPFFGFOYFRIGEKXAKEXAELPHTOKLMIASIKTINTINKINLAVABELKLEKORKRILAMKYLMIGMETMLSMOHMYTNIROPAPAPPPAPLAPPCQUEREVSAREPISIDOLYTENOLTEYA
PLATITTRA
SRABSL
IRB
SES
SLN TMR
VUB
GRVKRK
MEL
PET
POS TLS
ZVT
ARCJRV
MRKTVETPDBLT
EEGHAENCNOEG
PRFPKGSFGSKN
TALTKM
BRVLOK
DPKGRD
GRZKA1KCMBALGZELJMLSC
LTTZOVLAPNKAOLFRARRER
FRMRJRRKBSAFSMASCMTMA
TKBVSSVEFRRRVNF
TSHAPEERMGAP
HB LOGLHH
LPLLUIMITVCWFWW
AABELI
ALKALMALMAMBAMBMAEMAEOAHFOAATL
AURBO BNOBAVBIOBLVBOCKLEHARBIFCARCARCBRCHECHRCOLCOLCOMDNODLHDABDANDANDFDDJUDSVEACEGEERREXQEXPFEDBORFFAFLSFLUFYNGR4GABGENGERGJ GN GESGRIGRLGYL
GYLLUNHH
HARHOEHOEHVIIC
IMAJMIJDAJUTKBHKRELASLLBLLALOLLUXMATMOLMTBMNBNEUNEWNKTNDANRDNORNRDNORNRSNORNOVNZYNTRNUNOSSOJBPNDPARPAAPRIRECREARELRIARILRBLROCROCRBRRTXSALSANSASSBSSCHSIFSIMSKASKJSMASOLSPGSPNROVSTRMIGTDCTIVTKDTOPTOP
TPSTORTOTTRITRYUIEVELVWSVJBVIBVIIVIP
WDHZEA
ADNADMAGKAMEAAL
ANTARM
AHTABF
AZNAV/
BABBA/
BG/BLTBP/BATBLNBSYBT/BNZBRBCPICCLCNACCHCPGCRHDGEEZJEXPFREGFS
GKNGSK
GLEHMS
HL/IMI
IMTIHG
IAGITR
ITVSBRJMAKGFLANLGELSEMKSMGGMROMND
NG/NXTOML
PSOPSN
PFCPRU
RRSRB/
RELRSLREXRIORR/RDSRDSRMGRSASABSGESDRSVTSHPSN/SMISPDSSESTASL/
TATTSC
TPKTT/TLW
ULVUU/VODWEIWTBWMHMRWWOSWPP
ANYAPPPAN
DANFHB
MTEMOL
PANRABRIC
SYN
6PM
CW FIM
GCLGO GHMHSB
IHI
IHG
LOM
MLT
MIAMTP
MDSMSIMDIPZC
RS2
SFC
TML
AKSDET
DNOFOEGJFGOGMHG
NHYNAS
OPE
ORK
PGS
PRSREC
RCLSCHSDRSTLSTB
SUBTELTGS
YAR
ACP
BHWPEO
BZWEURGTCLTS
JSW
KER
KGH
LWBMBK
OPLPGEPKNPGN
PZU
SNS
TPE
ABBALF
ASSAZN
ATCATCBOLELUGET
HMBINV
LUPMTGNOKSAN
SCVSEC
SKASKFSSASCASWM
TELERITLSVOL
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 22 / 30
Applications – policy implications
Event-driven contagion (realised)
Deterioration of credit quality in a given sector (NACE) – corporate loan lossestrigger contagion
Risk weights – policy tool to limit exposures to risky sectors (realised)
Specific sectors can be targeted to force banks to use more capital for more riskysectors
Large Exposure limits – compactness of the networks (realised)
lower bilateral exposures allowed ⇒ more connections
Network reactions to adverse market conditions (planned)
passing macro scenarios via dynamic BS model (Ha laj, 2013):baseline macro scenario ⇒ optimising behaviour of banks ⇒ change in banks’ preferred
aggregate interbank lending and borrowing ⇒ endogenous formation of the interbank under
specified regulatory regime ⇒ adverse macro shock ⇒ banks defaults ⇒ contagion
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 23 / 30
Figure 7: Contagion simulation
BE HU LU PT
AT CY DE DK ES FI FR GB GR IE IT NL NO SE SI
Consumer, Non-cyclical [7]Consumer, Cyclical [8]Non-bank financial [11]
Consumer, Non-cyclical [1]Energy, Basic materials [2]
Contagion mechanism – cascade triggered by a deterioration of creditquality of loan portfolios to companies in a given NACE sector(manufacturing in DE) imposing 5% PD and 50% LGD
“Spectral” graph shows impact of the contagion losses of 500+ banks(the darker the bar, the higher the fraction of capital wiped out bycontagion)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 24 / 30
Figure 8: Contagion simulation for different deterioration of credit quality
5% 10% 15% 20% 25% 30% 35% 40% 45% 50% 55% 60% 65% 70% 75% 80% 85% 90% 95%
AT DE DE DE DE DE DE DE DE DE DE DE DE DE DE DK DK ES ES FI FR GB GB GB GR IT IT IT NL NO PL SI
100%
Contagion mechanism – cascade triggered by a deterioration of credit qualityof loan portfolios to companies in a given NACE sector for (y-axis)PD ∈ {5%, 10%, . . . , 100%} and 50% LGD
“Spectral” graph of contagion losses of 500+ banks (the darker the bar, thehigher the fraction of capital wiped out by contagion)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 25 / 30
Figure 9: Risk-weight policy impact on contagion triggered by deteriorationcredit quality
60%b
70%b
80%b
90%b
60%a
70%a
80%a
AT DE DE DE DE DE DE DE DE DE DE DE DE DE DE DK DK ES ES FI FR GB GB GB GR IT IT IT NL NO PL SI
90%a
“Spectral” graph of contagion losses of 500+ banks (the darker the bar, thehigher the fraction of capital wiped out by contagion)
a) baseline case (PD∈ {60%, . . . , 90%})b) risk weights increased by a factor of 1.5 for all exposures to the
manufacturing sector in Germany
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 26 / 30
Figure 10: Defaults of banks in the cascade of contagion spreading triggered bylosses in the portfolio of loans to the manufacturing sector in DE
3%
5%
7%
9%
11%
13%
15%
17%
19%
21%
23%
AT DE DE DE DE DE DE DE DE DE DE DE DE DE DE DK DK ES ES FI FR GB GB GB GR IT IT IT NL NO PL SI
25%
Defaults of banks triggered by banks failing to pay back their obligations asa result of losses related to decreasing credit quality of manufacturing loanportfolio in (counterfactual example!) Germany
Each bar indicates a defaulting bank under LE∈ {3%, 5%, . . . , 25%}Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 27 / 30
Figure 11: Second round defaults of banks in the cascade of contagion spreadingtriggered by losses in the portfolio of loans to the manufacturing sector in DE
3%
5%
7%
9%
11%
13%
15%
17%
19%
21%
23%
AT DE DE DE DE DE DE DE DE DE DE DE DE DE DE DK DK ES ES FI FR GB GB GB GR IT IT IT NL NO PL SI
25%
Defaults of banks triggered by banks failing to pay back their obligations asa result of losses related to decreasing credit quality of manufacturing loanportfolio in GermanyEach bar indicates a defaulting bank only because their debtors defaultedunder LE∈ {3%, 5%, . . . , 25%}
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 28 / 30
Where may the predefined volumes of interbank assets andliabilities come from? Endogenous balance sheet
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 29 / 30
Conclusions
Endogenous interbank networks give an important insight into therole of banks’ investment and funding strategies in shaping theinterbank market and non-bank firms’ funding channels. The simple,mechanistic cascade models are too simplistic in assuming that banksdo not react to actions of other interbank participants and marketconditions.
It is easier to introduce heterogeneity of agents if the networkapproach is taken rather than macroeconomic (e.g. generalequilibrium) framework.
In the proposed framework, we are able to analyse different policymeasures addressing the systemic risk – their ultimate impact on themarket structure and efficiency in reducing the contagion risk.
Still, a more consistent clearing mechanism would be warranted.
The model needs to be calibrated to the observed interbank / lendingnetworks. How far are we from the truth? (Target 2 data to be used)
Grzegorz Ha laj (ECB) Systemic Risk and Financial Networks, IPAM 26/03/2015 30 / 30