Capital Review - Setting an output floor for the IRB approach

Setting an output floor for the IRB approach FSO Committee 7th November 2018

2
Purpose

• Previous FSO paper (#7752196) outlined different options for setting
  an output floor on IRB banks’ risk-weighted assets, linked to standardised outcomes.
• The paper recommended a combination of a higher IRB scalar and an
  output floor as a better approach to addressing the disparity between IRB and standardised outcomes than using an output floor alone.
• FSO asked for a report back with further discussion on the rationale for
  increasing the IRB scalar.

3
Risk weighted assets calculation
IRB: Capital requirement determined by banks’ estimates of probability of default (PD), loss given default (LGD), exposure at default (EAD), maturity (M), and regulatory correlation factor (R) Standardised: Capital requirement is a combination of prescribed risk weight for each type of counterparty and exposure amount IRB 𝑘𝑘𝐼𝐼𝐼𝐼𝐼𝐼 = 1.06 × 𝑓𝑓(𝑃𝑃𝑃𝑃, 𝐿𝐿𝐿𝐿 , 𝐸𝐸𝐸𝐸 , 𝑅𝑅, 𝑀𝑀) Standardised 𝑘𝑘𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 = risk weight × exposure amount

4
What is the IRB scalar?
𝑘𝑘𝐼𝐼𝐼𝐼𝐼𝐼 = 𝟏𝟏. 𝟎𝟎𝟎𝟎 × 𝑓𝑓(𝑃𝑃𝑃𝑃, 𝐿𝐿𝐿𝐿 , 𝐸𝐸𝐸𝐸 , 𝑅𝑅, 𝑀𝑀)

• It was not known where banks’ final IRB numbers would land vis-à-vis
  Basel I or the new standardised approach when Basel II was being introduced. The Basel Committee included a scaling factor for all exposures subject to IRB, which gave supervisors a level of comfort about the capital impact of the new framework.
• Basel II finalised the scalar at 1.06, and this has been unchanged
  since 2006. We also set the scalar at 1.06 in New Zealand.
• The scalar applies equally across all IRB exposures. This means it
  increases overall capital from the IRB approach while fully preserving IRB’s risk differentiation (relative capital amounts are unchanged, within and across asset classes and banks).

5
Output floor dilemma
We want to prevent unacceptably low IRB outcomes FSO made an in principle decision to impose a floor on RWA produced by the IRB approach, so that total RWA cannot be below X% of the RWA that IRB banks would calculate if they instead used the standardised approach. A review principle is that “where there are multiple methods for determining capital requirements, outcomes should not vary unduly between methods”. We also want to retain a risk sensitive framework FSO decided to retain the IRB approach because of the improved risk differentiation it offers, compared to the standardised approach. Another review principle is that “capital requirements should be set in relation to the risk of bank exposures”. But the standardised approach is not very risk sensitive in practice The problem we face is that the current standardised approach to which we would tie an output floor has, in practice, very rudimentary risk differentiation when compared to the IRB approach. Based on the QIS estimates, 94% of the four IRB banks’ credit exposures would be slotted into three risk weight buckets (35%, 40%, 100%). In effect, the standardised approach is a quasi-leverage ratio.

6
How to reduce the gap?

• The current IRB approach delivers around 75% of
  standardised outcome, on average.
• • Only using the output floor to reduce the gap
  between IRB and standardised outcomes would require a very high calibration (e.g. 90%+).
• But at this level of calibration, all IRB banks’
  marginal capital requirements are effectively determined by the standardised approach, since it is the binding constraint.
• This is inconsistent with the principle that the
  framework should be risk sensitive, given the known limitations of the current standardised approach. RBNZ s105 RBNZ s105

7
How do the objectives and tools interact?
Reduce gaps between IRB and standardised Preserve a risk-sensitive capital framework Reduce gaps across IRB banks Only use output floor Only use scalar Use a combination of scalar and output floor More intensive monitoring, supervision and enforcement of IRB (e.g. regular benchmarking)

8
Scalar and floor combined

• A combination of a more moderately calibrated scalar and output floor
  is a better way of achieving the desired outcome than relying only on one tool or the other to reduce the gap to standardised.
• Increasing the scalar from today’s 1.06 could do most of the heavy
  lifting in terms of reducing the gap between average IRB and standardised outcomes, while fully preserving the risk differentiation of the IRB approach.
• The output floor would then serve as a backstop to raise the RWA of
  any outlier banks, if that is still needed once a higher scalar is applied.

9
Example calibration

• Each of these example calibrations delivers the same average outcome (IRB at
  ~85% of standardised RWA), but with different implications for the relative outcomes of each bank. 60 65 70 75 80 85 90 95 100 60 65 70 75 80 85 90 95 100 Status quo Tight floor IRB RWA as % of Standardised RWA Tight floor (85%) 60 65 70 75 80 85 90 95 100 60 65 70 75 80 85 90 95 100 Status quo High scalar High scalar (1.2) 60 65 70 75 80 85 90 95 100 60 65 70 75 80 85 90 95 100 Status quo Moderate scalar Moderate floor Moderate scalar (1.15) and floor (80%)

10
Why use the scalar

• Since for most IRB banks the output floor would not be the binding
  constraint, marginal capital requirements are still determined by the IRB approach (calibrated higher with the scalar). This means that changes in underlying risks will still be reflected in IRB banks’ RWAs, to the extent that the IRB models pick up this change in risk.
• If we only use the output floor, so that it is the binding constraint on IRB
  banks’ capital, the risk insensitivity of the standardised approach means that total RWAs would be less likely to respond to a change in underlying risks.

11

12
Extra slide:
How would each option respond to an increase in underlying risk? • e.g. what happens if more high DTI mortgage lending increases banks’ underlying risks by 10%?

• Assume RWA calculated by IRB
  models grows by 10%, since IRB is risk-sensitive.
• Under a “tight floor” option, the
  output floor to standardised is the binding constraint, so RWA of most banks does not respond to increased underlying risk.
• Under the other two options,
  RWA increases for most banks because the IRB outcome (with higher scalar applied) is the binding capital requirement. 60 65 70 75 80 85 90 95 100 105 110 60 65 70 75 80 85 90 95 100 105 110 Status quo Tight floor Increase in risk Status quo High scalar Increase in risk Status quo Moderate scalar and floor Increase in risk Tight floor High scalar Moderate scalar and floor IRB RWA as % of Standardised RWA