Banking System ABM Model
How do interbank lending networks, heterogeneous bank balance sheets, and capital adequacy constraints generate systemic risk, contagion cascades, and too-big-to-fail dynamics that aggregate risk metrics miss entirely?
Graph workspace
Multi-panel time-series with network snapshot
Banking system ABMs produce stochastic time series (aggregate capital, interbank rate, default count) alongside evolving network topologies. The canonical visualization pairs a time-series panel showing the build-up and cascade with a network diagram snapshot highlighting contagion paths. This makes the connection between aggregate stress indicators and underlying network structure visible in one view.
Current results
System Capital Ratio
Aggregate capital adequacy of the surviving banking system. Falls during stress as losses erode equity. A system-wide measure that can mask concentration of distress in specific banks.
2.02%
percent
Cumulative Bank Failures
Running count of banks that have breached the capital minimum and been resolved. Jumps during cascade episodes. The time gap between failures reveals whether contagion is sequential or simultaneous. On this page's sketch it is a whole-number tally of the 20 seeded banks, so it can never exceed 20. At the shock quarter a bank fails when the credit loss exceeds its buffer above the minimum (, with the buffers spread evenly over the seeded 4-10 pp); over the next three quarters a fire-sale and hoarding markdown can fail more, and that markdown is calibrated so each preset lands on the total its own arithmetic states rather than simulated round by round. The chart plots the failure pressure behind the tally: the capital lost so far as a percent of the weakest surviving bank's buffer, where 100 means the next bank has failed.
20.00
count
Interbank Spread
Spread of the emergent interbank rate over the policy rate, equal to with and . The stress term is bounded and contributes at most 200 bp when every bank is below its capital target; the deficit-to-surplus term is not bounded - 40 bp at a ratio of 1, 400 bp at a ratio of 10 - and it diverges as the surplus side of the market disappears, so this readout has no ceiling. The real-world analog is the LIBOR-OIS spread, which was NOT near zero before the cascade: it stepped up to the 50-100 bp range in August 2007 and stayed there through the following year, then peaked near 364 bp in October 2008 after the Lehman failure.
44.40
basis points
Max DebtRank
DebtRank of the most systemically important surviving bank. Rises as smaller banks fail and the surviving network becomes more concentrated. A Max DebtRank above 0.3 means a single bank failure could destroy 30%+ of system equity.
0.9500
ratio
Controls
Min capital ratio: 0.08
How thin a bank's capital can get before regulators step in or shut it down.
Higher minimum reduces leverage and contagion losses but shrinks the interbank market as banks hold more capital buffers. System capital line sits higher but failures still cluster when shocks exceed the buffer.
Fire-sale impact: 0.1
How far securities prices drop for each unit banks sell, measured against how much the market can absorb.
A higher produces steeper securities price drops during stress, amplifying mark-to-market losses and accelerating the cascade. At the default the Lehman-preset spiral converges in four to six rounds with a cumulative price drop of about 6-9%; at it is about 12-18%. The system capital line drops faster and in visible steps rather than in one jump.
Interbank LGD: 0.6
The share of a bank's lending to another bank that is lost when that bank fails.
Higher LGD amplifies direct contagion: each bank failure imposes larger losses on counterparties, making cascades longer and more damaging. At , only indirect (fire-sale) contagion remains.
Network connectivity: 4
The average number of other banks each bank lends to or borrows from.
Low connectivity isolates failures but creates liquidity hoarding. High connectivity shares liquidity efficiently in normal times but transmits large shocks system-wide. The cascade threshold shifts nonlinearly with .
Initial shock size: 0.02
The share of the banking system's loan book wiped out by the one-time shock.
Because the loan book is 70% of assets, a shock of costs of assets against an initial equity of . A bank fails when , so at the 8% default the failure threshold runs from for the weakest bank (CAR 12%) to for the strongest (CAR 18%). Below 5.1% nothing fails; the failure count then climbs across that band rather than jumping from zero to everything, and fire-sale markdowns plus interbank contagion add the sharp step on top.