RDP 2012-06: The Impact of Payment System Design on Tiering Incentives 6. The Impact of Tiering on Risk

The benefits of tiering can come at a cost of increased credit and concentration risk. This section estimates the changes in credit and concentration risk in RITS due to increased tiering. The effect of system design on credit risk is also examined.

6.1 Credit Risk

Tiering creates a two-way exposure between a client and its settlement bank because payments are settled across the settlement bank's books, rather than directly in RITS (for which there is no credit risk). Furthermore, these payments – unlike those in RITS – may be subject to the ‘zero hour’ rule, which means that in the event of a bankruptcy, their finality can be challenged. In this section we present measures of this two-way exposure for the two system designs at either end of the liquidity-usage spectrum: the pure RTGS system and the RITS replica system.

6.1.1 Settlement bank exposures

A settlement bank's maximum intraday exposure to a client can be measured as the client's maximum intraday cumulative net payment (as opposed to receipt) position when the client settles directly in the RTGS system. This measure of settlement bank exposure should be regarded as an upper bound because a settlement bank can vary the timing of sending its clients' payments to minimise its exposure, and require clients to pre-fund settlement obligations.^[17]

We find that a settlement bank's average maximum intraday exposure to any one of the smallest 29 tiering candidates over the sample period is less than $100 million (Figure 2). While the largest maximum intraday exposure over the month is roughly three times the size of the average maximum intraday exposure, this is still quite low for the smallest 29 tiering candidates (Figure 3). Unsurprisingly, maximum intraday exposures are typically much higher among the largest 20 tiering candidates. We are unable to determine the size of the exposures that the settlement banks in our simulations would be willing to accept, as these are likely to be functions of the capitalisation and risk preferences of the individual institutions. However, we note that while the largest maximum intraday exposure of around $2 billion is sizeable, it is considerably smaller than the tier 1 capital held by each of the four largest settlement banks (over $20 billion in 2008).

Because our measure of settlement bank exposure (a client's maximum intraday cumulative net payment position) is equal to our measure of the client's liquidity usage when it participates directly in the RTGS system we expect higher settlement bank exposures in the more liquidity-intensive pure RTGS system. The difference in exposure between the two system designs varies considerably with the institution being tiered. For the median institution in the tiering group (in terms of this exposure), the average maximum intraday exposure of the settlement bank to one of its clients is 8 per cent higher in the pure RTGS system.

6.1.2 Individual client exposures

A client's maximum intraday exposure to its settlement bank can be measured using that client's maximum intraday cumulative net receipt (as opposed to payment) position when it settles directly in the RTGS system. Because a settlement bank has discretion over the timing of payments, and because it may require pre-funding from its client, these estimates should be viewed as a lower bound.

Clients' average maximum intraday exposures are typically less than $1 billion (Figure 4). The largest maximum intraday exposures are still less than $1 billion for smaller institutions, but are as high as $3.5 billion for the largest clients (Figure 5). Given that the largest clients are typically (though not always) branches of global banks, their largest exposures are still small relative to their group tier 1 capital.

Clients' exposures in the pure RTGS system are similar to those in the RITS replica system. Again, the difference in exposure between the two system designs varies considerably with the institution being tiered. For the median institution (in terms of this exposure), the average maximum intraday exposure is 2 per cent higher in the pure RTGS system.

6.1.3 Total client exposures

While a settlement bank is unlikely to face the simultaneous default of all of its clients, if a settlement bank defaults, all of its clients are exposed. To estimate the maximum total client exposure to a particular settlement bank we can sum the minute-by-minute exposures, measured using each client's cumulative net receipt position when it settled directly.^[18] As noted above, these estimates of client exposures should be viewed as lower bounds.

Each observation in Figures 6 and 7 represents the maximum aggregate loss that could occur if the settlement bank to which the n^th smallest institution tiers defaults on its obligations, and it defaults on all its obligations to any smaller institutions that also use it as a settlement bank. For example, when the 49th institution is tiered in Figure 6, the average maximum intraday exposure in total for that institution and other clients using the same settlement bank as an agent is around $4 billion in the RITS replica system. Each colour in the figure represents one of the 4 largest settlement banks.

With the exception of larger institutions that tier to the settlement bank depicted in pink, total client exposures are typically at least as high in the pure RTGS system as they are in the RITS replica system. For the median case, the average maximum intraday exposure is around 1 per cent higher in the pure RTGS system compared with the RITS replica system.

6.2 Concentration Risk

Indirect participants in a payments network send payment instructions to their settlement bank, which then acts on their behalf. Consequently, in choosing to tier the client becomes operationally dependent on its settlement bank. One might argue that larger institutions are better equipped to minimise the probability of an operational problem. However, by concentrating payment flows, tiering amplifies the consequences of an operational incident at the settlement bank – in particular, the size of the potential liquidity sink increases.

A general measure of this type of operational risk is the level of concentration in the system: the increase in settlement banks' share of payments as the level of tiering increases. Note that our measure of concentration is the share of payments sent, rather than sent and received, as generally even when a participant suffers an operational incident they can still receive payments. While a more accurate way to model the impact of tiering on the consequences of an operational incident is to simulate operational incidents in a tiered network, this is beyond the scope of this paper.

We find that our cumulative tiering scenarios result in only a modest increase in the concentration of payments being sent to RITS by the 4 largest participants. In the absence of tiering, the 4 largest participants account for around 57 per cent of all payments sent to RITS by value. If all of our 49 tiering candidates were to settle indirectly, the combined share of the 4 largest participants would rise by around 10 percentage points. Since it is unlikely that an operational incident would occur at all 4 of the largest participants simultaneously, it is more noteworthy that the largest increase for an individual settlement bank is only 4 percentage points.

An alternative measure of concentration is the value of payments that the four largest participants are collectively responsible for; that is, the value of payments sent by them to the central system plus the value of payments settled across their own books. By this measure the rise in concentration is more substantial, at just over 24 percentage points. In addition, the largest increase in share for an individual settlement bank rises by around 13 percentage points. Thus, the extent to which concentration risk is an issue depends on the relative likelihood of different types of operational outages; that is, whether outages are more likely to simply affect the ability of an institution to access the central system, or whether they are more likely to disrupt the processing of payments entirely. We do not pursue this issue further here.

Footnotes

Note that the timing of settlement in the tiered simulations may also vary depending on the liquidity available to the settlement bank. [17]

Note that exposures are not netted multilaterally. Therefore, if a client has negative exposure (that is, it owes the settlement bank), that exposure is excluded from the calculation. [18]