Payment Systems ABM Model
How do banks' individual liquidity management strategies, payment queuing decisions, and bilateral netting behavior generate systemic gridlock risk, cascading settlement failures, and the need for central bank intraday liquidity provision in large-value payment systems?
Graph workspace
Multi-panel intraday time-series with network heatmap
Payment system ABMs produce intraday time series (cumulative settlement, queue depth, reserve positions) alongside network-level metrics (bilateral flow concentration, cascade propagation). The canonical visualization pairs an intraday timeline showing settlement progress and queue buildup with a heatmap or adjacency matrix of the bilateral payment network. This makes the connection between individual bank timing decisions and system-wide gridlock visible in one view.
Current results
Settlement Completion Rate
Fraction of total daily payment value that has settled by the current tick. Should reach 100% by end-of-day in a well-functioning system; shortfall indicates settlement failures.
66.96%
percent
Queue Depth
Peak total number of payments waiting in all banks' queues at any tick of the day. Spikes indicate coordination failures where banks are mutually waiting for each other's payments. By the close the queue holds only what nobody could fund, which Settlement Failures reports.
107.00
count
Liquidity Deployment
GROSS value each bank has actually paid out by tick , scaled by its own starting reserves and averaged across banks - the per-bank turnover ratio. It rises when banks are recycling liquidity and stays flat when they are sitting on it, which is what a hoarding metric has to do. The obvious alternative - average signed deviation of reserves from their starting level - cannot do this job: payments are internal transfers, so one bank's outflow is another's inflow and the signed deviations sum to exactly the central bank's injections no matter how much settlement took place. A near-conserved quantity measures the injections, not the behaviour.
43.46
multiple of starting reserves
Settlement Failures
Count of payments still queued at end-of-day. The tail risk metric. Even one failure in a large-value system can indicate systemic stress. On this page's sketch the deadline window runs the queue down over the last two hours to the payments their senders cannot fund, a share that rises with an outage, with hoarding and with a priced credit line.
102.00
count
Controls
How much cash banks keep on hand before releasing payments; higher means they hold more.
Higher hoarding shifts settlement activity later in the day, increases peak queue depth, and raises the probability of gridlock events in the final hours.
Netting interval (min): 30
How often the system scans the queues for payment loops that offset each other.
Shorter intervals produce more frequent but smaller drops in queue depth. Longer intervals allow queues to build up and then clear in large batches, creating a sawtooth pattern.
Intraday credit cost (bps): 0
What banks pay to overdraw at the central bank; a charge makes them hold bigger cushions.
Positive intraday credit cost raises every bank's effective hoarding threshold (25bp multiplies it by 1.5, the 50bp maximum by 2.0), shifts settlement later, and raises gridlock probability. The system transitions from smooth to lumpy settlement.
Top-3 bank concentration: 0.6
How much of the day's payment value runs through the three biggest banks.
Higher concentration creates a visible hub-and-spoke pattern in the network heatmap. Gridlock triggered by a core bank outage propagates faster and deeper.
Gridlock detection threshold: 0.4
Share of payments stuck in line, with nothing settling, before the system calls it gridlock and nets again. It changes the chart only.
Changes how many gridlock episodes are counted on the chart. The extra netting pass finds no offsets the scheduled sweeps would miss, so the four readouts stay the same at every setting; this model has no central-bank liquidity injection.
Threshold decay through the day: 1
How quickly a bank gives up its cash cushion as the closing deadline nears.
A low holds the morning freeze and concentrates settlement in the deadline window, so the cumulative curve is flat then near-vertical. A high front-loads releases and bends the curve up early - the same total value settled by the close, but a visibly different path to it.
Core bank outage (hours): 0
How long the largest sender stops sending while still receiving; zero means no outage.
The disrupted bank's cumulative settlement flatlines while its own queue grows. Counterparty queues spike within 30-60 minutes as expected inflows fail to arrive, and end-of-day failures appear because the deadline override cannot release a payment the sender cannot fund.