FRANKFURT (MNI) – The following is the second part of a verbatim
text of a technical note from the European Central Bank (ECB) on the
macroeconomic scenarios and reference risk parameters following the
publication of the EU-wide stress tests’ result
Regarding the data entered into the ECB models for PDs and LGDs,
country-level financial sector PDs were approximated using the Moodys
EDFs (expected default frequency) extracted from the Moodys KMV
database.17 Sovereign PDs were derived from CDS spreads. Retail real
estate PDs, consumer credit PDs and corporate sector PDs were obtained
from the ECB Monetary and Financial Institutions (MFI) database on
write-offs, while LGDs were extracted from Moodys LossCalc database
assuming a constant PD over time.18
To illustrate the severity of the adverse scenario, Chart 1 plots
the ranges of changes across all countries in the PDs between the
adverse scenario and the end-2009 values over 2010 and 2011, for the
four private-sector portfolios sectors considered in the credit risk
part of the exercise, and Table 4 shows the corresponding figures for
2011. Chart 2 shows the results of the same exercise as Chart 1, now for
the LGD parameters. As seen in both charts, the PDs and LGDs increase
substantially across sectors and countries under the adverse scenario
compared to end-2009 in both 2010 and 2011. In this regard, it is
important to note that the stresses on long-term interest rates that
result from the sovereign shock feed through to higher PD and LGD
levels.
3. Sovereign bond haircuts
The increase in bond yields affects the valuation of holdings of
government debt in the banks trading books,19 and in the exercise its
impact is not offset by changes in the valuation of derivative positions
(credit derivatives, interest rate swaps, etc.) that are used to hedge
the sovereign bond exposures.
For the purposes of estimating valuation haircuts, it was agreed
among participating supervisors that a five-year maturity was
representative of the approximate duration of sovereign bond holdings
held by banks in the EU. Hence, the haircuts for sovereign bonds are
computed in two steps, first by estimating five-year bond yields,
consistent with the assumptions for ten-year yields and then, in a
second step, translating these five-year yields into their corresponding
sovereign bond prices. a. Transformation of ten-year yields to five-year
yields The transformation uses the ten-year yields prevailing in the
benchmark and adverse scenarios together with the five-year yields that
were assumed to prevail in the market at the end of 2009. The changes in
five-year bond yields from 2009 to 2010 and to 2011 were set equal to
the changes (in basis points) in the ten-year yields. This method was
applied for all countries, apart from Germany, which acts as the
reference sovereign issuer with the lowest yield in the euro area.20 The
exceptions are the euro area countries where the bond markets are not
liquid or where this method would lead to a significant compression of
sovereign bond yield spreads vis- vis German bonds. For those countries
(Cyprus, Malta, Slovakia, and Slovenia) it was assumed that the
sovereign bond yield spreads over the German yields would remain
constant in the benchmark scenario. In the adverse scenario, the
five-year yields are constructed from the values in the benchmark
scenario using the same procedure as followed for the ten-year yields,
taking into account both the yield curve flattening and the sovereign
risk components. Again Germany, being the reference issuer, is assumed
to be unaffected by the elevated sovereign risk. b. Haircuts on
sovereign debt The haircuts were computed from changes in the prices of
five-year sovereign bonds under both scenarios. The parameters that are
essential for the pricing of sovereign bonds (coupons, coupon
frequencies, coupon and maturity dates) were collected from Bloomberg.
In order to eliminate potential distortions which may arise when
the bonds that are currently the most actively traded have been issued
with very high or very low coupons,21 all bonds for which market quotes
were available on Bloomberg for each country that had a remaining
maturity of 4.5 to 6.5 years were priced and the weighted average change
in their prices was used to construct the haircut. The weights in the
average are based on the outstanding amount of the bonds.
In the pricing of sovereign bonds the discounted cash-flow method
was used, in which the yields to maturity under the relevant scenario
are used to construct the discount factors. This method takes into
account the actual maturity dates, coupon dates and coupon frequencies
for the individual bonds. The haircuts are applied to the market value
of bonds at the end of 2009, separately for each year. Therefore, a bond
which was worth 100 at the end of 2009 and which has a haircut of 4% in
2010 and 6% in 2011 should be valued at 96 at the end of 2010 and at 94
at the end of 2010. The haircuts used in the exercise (Table 5) are the
future values of the outstanding sovereign bonds. The exercise is
supposed to provide the values of the bonds to be booked in the end-2010
and end-2011 accounts. This implies that a 5-year bond, representative
of the average maturity of this portfolio by banks, has a duration of
only 3 years at the end of 2011, when accounts are closed.
The haircuts can be decomposed to reflect the three main
contributing factors: the overall rise in longterm interest rates
foreseen in the benchmark macroeconomic scenario, the common upward
shift of the yield curves, and the country-specific sovereign risk shock
(Table 6). The decomposition illustrates that for some non-euro area
countries, the higher haircuts are driven primarily by the expected
increase in long-term interest rates, with the impact of the sovereign
risk shock playing a lesser role. For the purposes of illustration and
comparison with the haircuts on five-year bonds, the same calculations
were carried out for ten-year bonds (Table 7). The yields used to
calculate the haircuts on ten-year bonds are the yields provided as part
of the macroeconomic scenario that, where appropriate, include a
sovereign risk component. The haircuts on the ten-year bonds are
generally higher than the corresponding haircuts on the five-year bonds
due to the higher duration. Taking Austria as an example, the haircut on
the five-year bonds under the adverse scenario is 5.6%. The
corresponding haircut on tenyear bonds is 9.5%. For Greece, the
respective figures are 23.1% and 42.2%.
[TOPICS: M$$EC$,M$X$$$,M$$CR$,MT$$$$]