dustry into “haves and have-nots.” For example, some anecdotal estimates suggest
that no more than 15 U.S. banks will choose to use either of the IRB approaches.
Moreover, capital requirements are highly sensitive to the accuracy of certain pa-
rameter values; in particular, estimates of LGD and the granularity in PD are im-
portant (see Gordy (2000) and Carey (2000)). Since credit losses are affected by
economic conditions, the model parameters should also be adjusted to reflect ex-
pected levels of economic activity. Thus, the data requirements are so substantial
that full implementation of the Advanced IRB Approach lies far in the future even
for the most sophisticated banks. And when that date comes, regulators will have
commensurate challenges in obtaining the necessary data to validate the banks’
models.
3.5 ASSESSMENT. BIS II is a potential improvement over BIS I in its sophistication
in measuring credit risk. Moreover, it moves regulatory capital in the direction of
economic capital. However, it is far from an integrated portfolio management ap-
proach to credit risk measurement. Focus on individual ratings classifications
(whether external or internal) prevents an aggregated view of credit risk across all
transactions, and regulatory concerns about systemic risk prevent full consideration
of cross-asset correlations that might reduce capital requirements further.^49 Thus,
capital requirements are likely to be higher than economically necessary when con-
sidering actual portfolio correlations^50 Moreover, incompatible approaches to assess-
ing the capital adequacy of insurance companies and other nonbanking firms may ob-
scure their impact on financial system instability. In the United States, the insurance
industry and government-sponsored enterprises (such as Fannie Mae and Freddie
Mac), and the Financial Services Authority in the United Kingdom all use a variety
of models, ranging from minimum ratios and stress test survivorship requirements to
dynamic risk-of-ruin scenario analysis, that include both the asset and liability sides
of the balance sheet in order to measure capital requirements.
The Advanced IRB Approach also contains some properties that may distort bank
incentives to manage their credit risk exposure. For example, Allen (2002a) finds that
the maturity adjustment in the Advanced IRB Approach (see equation(7)) creates per-
verse incentives when dealing with loans with maturities greater than three years such
that the loan adjustment factor decreasesthe loan’s risk weight as the loan quality
(credit rating) declines. Moreover, the Advanced IRB Approach penalizes increases in
LGD more than increases in PD. Exhibit 3.8 uses data from Altman and Saunders
(2001b) to determine the impact of increases in LGD on the Advanced IRB risk
weights for loans with maturity of three years keeping expected losses (i.e., LGD
PD) constant. For all risk buckets (for illustrative purposes only, the Standardized
Approach’s risk classifications are used), the Advanced IRB risk weights increase as
3 • 16 BIS BASEL INTERNATIONAL BANK CAPITAL ACCORDS(^49) Hoggarth, et al. (2001) show that cumulative output losses during systemic crises average 15 to 20%
of annual GDP.
(^50) That is, the IRB frameworks are calibrated to an asset correlation of 0.20, which is higher than ac-
tual correlations that averaged 9 to 10% for eurobonds; see Jackson et al. (2001). The November 2001
potential modifications to BIS II proposals incorporate a correlation coefficient that is inversely related
to the PD. However, Freixas et al. (2000) show that systemic crises may occur even if all banks are sol-
vent.