The problem I see with this is that there are perfectly valid reasons to use both debt and intermediaries in a financial system.

A disintermediated economy can be visualised as a network in which most nodes connect to only a few other nodes. Each connection has three properties: A cash flow, an outstanding debt balance and an outstanding equity balance. Each node has five properties: Its aggregate cash flow, the equity it owes to other nodes, the debt it owes to other nodes, its real capital and its clear equity - the equity that does not appear as an asset for any other entity.

If some inauspicious event happens to a node, the holders of any equity owed to other nodes will suffer a loss. If you do not allow debt balances between nodes, then any and all inauspicious events will cause losses for every node in the network. These losses will be serially diluted, of course, but since the average node has a low number of outgoing connections, this dilution is relatively slow.

Debt, in this picture, functions much as a dike against such inauspicious events: As long as the inauspicious event is within tolerances (that is, as long as it does not wipe out the equity of the node principally involved), the loss is not transmitted to the wider system. But when it fails the failure is correspondingly spectacular, because the pool of equity in which the losses can be diluted is correspondingly smaller.

Banks serve as the second tier of dikes: By intermediating credit, it detaches the credit relationship from the cash flow relationship. Rather than node A doing business with node B and extending or obtaining credit from node B, it does business with node B and extends to or obtains credit from node C.

This has two advantages: The first is that it puts node B at one remove from node A's insolvency. Node A's insolvency will have to be bad enough to wipe out node C's equity as well before node B takes losses. The second advantage is that node C will accumulate a lot of connections, and a more connected node is, ceteris paribus, less affected by any individual default.

Both of these add value for every risk-averse node in the network. Which is very nearly all of them.

As always, there is a tradeoff between failing gracefully and failing rarely. By imposing a model that forces the system to fail gracefully, you are also mandating that it fails a little bit all the time.

- Jake

Friends come and go. Enemies accumulate.