Liquidity bounds for prefunded delivery-versus-payment and the cost of fragmented settlement assets
Why
Official-sector reports on tokenization keep saying atomic settlement is expensive in liquidity because it gives up netting, and the DTCC's 98% netting figure gets read as a fifty-fold increase in funding. I wanted to check whether that comparison is even the right one for a system where sale proceeds are reusable the moment they settle, and to put a number on the real difference.
Found
- What a participant needs under atomic settlement is the running maximum of its net outflow, not its gross purchases. For balanced flow that is a small constant above the net requirement: about 1.11× at 99% confidence and 1.07× at 99.9%, regardless of trade count or volatility. Heavy-tailed, directional and clustered flow shrink it further.
- The gross-to-net ratio grows with the square root of the number of trades. That is the only reason the 98% statistic looks alarming.
- The real costs sit elsewhere. Splitting the cash leg across K non-fungible settlement assets raises prefunding by about √K, roughly 3× for six issuers, and that penalty applies to net settlement too. Micro-batching recovers little: minute-scale batches close about 3% of a gap that is itself about 10%.
- So the case for or against atomic settlement should be argued on counterparty risk, finality, the securities leg and settlement-asset fungibility, not on netting.