As noted earlier, a lag between the time at which information is made available to receiving banks and the time at which settlement takes place may have important risk implications in large­value funds transfer systems. Even in the RTGS environment, where both processing and final settlement are made in real­time, several circumstances can be identified in which the treatment of payment messages or the associated information could be a source of risk. This section looks at four different types of message flow structure.

To initiate a funds transfer the sending bank dispatches a payment message which is subsequently routed to the central bank and to the receiving bank as the system processes and settles the transfer. Arrangements for routing payment messages in the majority of RTGS systems are or will be based on a so­called V­shaped message flow structure. In this structure the full message with all the information about the payment (including, for example, the details of the beneficiary) is initially passed to the central bank and is sent to the receiving bank only after the transfer has been settled by the central bank (see Box 3).

Some G­10 RTGS systems, and particularly those that use the S.W.I.F.T. network, apply an alternative structure, known as a Y­shaped structure. In this case, the payment message is transmitted by the sending bank to a central processor. The central processor takes a subset of information that is necessary for settlement from the original message and routes this core subset to the central bank (the original message being kept in the central processor). Upon receipt of the core subset, the central bank checks that the sending bank has sufficient covering funds on its account and informs the central processor of the status of the transfer, for instance queued or settled. Once settled, the full message containing the confirmation of settlement is rebuilt by the central processor and sent to the receiving bank. The business information exchanged between the sending and receiving bank (such as the identity of the beneficiary) is therefore not known by the settlement agent.

So far only CHAPS has adopted an L­shaped structure, which is conceptually similar to a Y­shaped structure. This configuration was chosen because it could be implemented by modifying rather than rewriting the software supporting the previous DNS arrangement. The payment message dispatched by the sending bank is held at a system "gateway" attached to the sending bank's internal processing system and a subset of information contained in the original message (a settlement request) is sent to the central bank. If the sending bank has sufficient covering funds on its account, settlement is completed and the central bank sends back a confirmation message to the sending bank's gateway. Upon receipt of this confirmation (and only then), the original payment message is released automatically from the gateway of the sending bank and sent to the receiving bank.

In part, these different structures reflect differences in both the network configuration of the system and the operational role of the central bank. In both the V and Y­shaped structures, all messages from the sending bank are routed first to a central entity (S.W.I.F.T., a central processor or the central bank itself) and, after settlement, all messages to the receiving bank are sent by that central entity. By contrast, there is no central entity for delivering messages in the L­shaped structure, where message routing is based on the bilateral exchange of information between banks, reflecting the decentralised architecture of the CHAPS system. In terms of its operational role in the system, the central bank is directly involved in both the settlement and processing of payment messages in the V­shaped structure, while in the Y and L­shaped structures the process of message routing can be handled by the network operator or the banks themselves, with the central bank only acting as the settlement agent.

It is important to note, however, that these three types of structure all share the common feature that the receiving bank will receive the full payment message only after the transaction has been settled by the central bank. In these structures, therefore, the message flow structure per se cannot give rise to the possibility that the receiving bank will act upon unsettled payments.

Alternatively, RTGS systems could apply a T­shaped structure, in which the sending bank routes payment messages directly to the receiving bank (through the message carrier), with copies (made by the message carrier) sent simultaneously to the central bank. This means that the receiving bank will first receive the full, unsettled payment message immediately after the sending bank has dispatched it and will subsequently also receive a confirmation message from the central bank once settlement has taken place. The T­shaped structure has generally been viewed as being incompatible with the basic principle of RTGS that a funds transfer should be passed to a receiving bank if and only if it has been settled irrevocably and unconditionally by the central bank. The receiving bank may not easily be able to distinguish between settled and unsettled payment messages at the time of receipt of the (unconfirmed) messages; even where it can make the distinction, competitive pressures may cause it to credit the funds to the beneficiary customer on the basis of the unsettled message or it may anticipate the arrival of the funds in other ways (e.g. by reducing the liquidity it obtains from other sources). To the extent that this happens credit and liquidity risks would be generated in the system in a manner analogous to that discussed in Section I.2. As this aspect has been recognised as a drawback in the RTGS environment, no planned or existing G­10 RTGS systems are based on the T­shaped structure.

BOX 3

MESSAGE FLOW STRUCTURES