Nomura Fixed Income Research

CDO Equity, Correlation, and IRR

I. Introduction1 March 21, 2005
Correlation remains one of the thorniest issues confronting CDO professionals. As we have reported
2
in the past, CDO senior tranches benefit from low correlation among their underlying assets while
equity tranches benefit from high correlation, when we look at expected losses. However, some
confusion about this point recently has surfaced. We hope to clear things up with this report.

CDO professionals can use different measures to gauge the potential effect on a CDO equity tranche
from changing the correlation among the credits in the tranche's underlying portfolio. Expected loss
is the most frequently used measure, but some have used expected internal rate of return (IRR) as
well. Using expected IRR produces the result that CDO equity tranches sometimes appear to benefit
from low – rather than high – correlation. Some market participants have misunderstood that result
as an apparent contradiction of conventional wisdom.

Two main factors contribute to the misunderstanding. First, the thickness of a CDO equity tranche is
a key determinant of whether its expected IRR rises or falls with increasing correlation. The thinner
an equity tranche is, the greater its tendency to have expected IRR rise with increasing correlation.
This is because a very thin equity tranche can be viewed as "pure" equity, while a thick equity tranche
essentially embodies exposures of higher seniority. Conversely, the thicker an equity tranche is, the
greater its tendency to have expected IRR decrease with rising correlation. Second, using the
directional change in expected IRR as a measure of "benefit" necessarily entails certain pitfalls. For
example, because IRR is non-linear, averaging it across different scenarios can produce distorted
results. Also, in contrast to a relatively straightforward calculation of expected losses, expected IRR
is highly sensitive to the time horizon over which it is estimated. That is an important confounding
factor that may distort results.

II. The Simplest Case: A First-To-Default (FTD) Basket

To understand the effect of correlation on the equity tranche, the simplest example one can study is a
first-to-default (FTD) basket. An FTD transaction terminates when the first default occurs in a basket
of reference credits. At that point, the protection seller (i.e., an investor) pays the loss amount
(typically the notional amount minus the recovery) to the protection buyer. An FTD survives only if no
default occurs in the basket (i.e., all credits survive). The transaction terminates in all other cases. In
assessing the risk/return profile of an FTD, we only need to evaluate the likelihood of termination or
survival.

Contacts:
1
Authors are grateful to Mark Adelson and Michael Van Bemmelen for providing valuable insights and Michiko Whetten
suggestions for this report. (212) 667-2338
2
Correlation Primer, Nomura Fixed Income Research at 2-3 (6 Aug 2004). mwhetten@us.nomura.com
Weimin Jin
(212) 667-9679
This report and others are available online at Nomura's new research website. To obtain a wjin@us.nomura.com
user id and password, please contact Diana Berezina at dberezina@us.nomura.com.
The web address is http://www.nomura.com/research/s16
Nomura Securities International, Inc.
Two World Financial Center
New York, NY 10281-1198
Please read the important disclosures and analyst certifications
www.nomura.com/research/s16
appearing on the last page.
Nomura Fixed Income Research

When correlation is high, there is a high likelihood that all credits survive or default. Only the
likelihood that all credits survive is important to an FTD, because it is wiped out in all other cases. If
the probability of no default is X, the probability of wipeout is (1-X). Of course, the return on an FTD
is higher when the transaction survives than when it terminates. However, no matter how low (or
even negative) the return is for the termination case, the average return of an FTD increases as
correlation increases, because the FTD becomes more likely to survive. (See the box below for an
illustration.)

Example: Suppose we have a FTD basket. An investor makes an initial principal investment of one
dollar. The investor receives a coupon of 10% and the principal if the transaction survives (i.e., no
defaults occur). On the other hand, if one or more defaults occur in the basket, he receives no
spread and loses the principal as well. If we assume the probability of no defaults to be X, we can
calculate the average return on the FTD as follows;

Prob.[no defaults] = X è Return = 10% (The FTD receives spread payments at maturity.)

Prob.[one or more defaults] = (1- X) è Return = -100% (The FTD terminates and the principal is
lost.)

Expected return = X(10%) + (1-X)(-100%) = (110%)X – 100%

The expected return is an increasing function of X (probability of all credits surviving), which
increases monotonically as correlation increases. The same result holds when we calculate the
expected IRR. On the other hand, the expected losses of a FTD are calculated as follows;

Prob.[no defaults] = X è Losses = 0% (The principal is paid at maturity.)

Prob.[one or more defaults] = (1- X) è Losses = 100% (The FTD terminates and the principal is lost.)

Expected losses = X(0%) + (1-X)(100%) = (1-X)100%

The expected loss is a decreasing function of X, which increases monotonically as correlation

increases. In other words, higher correlation benefits an FTD, whether we look at the expected losses
or the expected return!!!

FTD baskets are the simplest form of an equity tranche. As we showed above, an FTD basket
benefits from higher correlation among reference credits, both in terms of expected losses and
expected return. However, the relationship is slightly more complicated for a loss tranche, where the
entire loss distribution of the portfolio, not just the extreme outcomes, must be taken into a
consideration. Therefore, we must evaluate the behavior of the IRR under different loss distributions
resulting from various deal parameters. In the following sections, we briefly analyze the
characteristics of IRR as a return measure and then compare results from Monte Carlo simulations
for equity tranches as we change tranche size, spread, and time horizon.

III. The IRR Puzzle

The internal rate of return (IRR) is a popular measure among some market participants. The IRR is
the discount rate that equates the present value of a stream of cash flows and the amount of initial

(2)
 Nomura Fixed Income Research

3
investment. In the fixed income arena, IRR is the yield to maturity of a security. The IRR, despite its
4
popularity, can be misleading at times as a measure of return.

Let us illustrate this point using a simple example. Suppose an investor has two investments options;
a risk-free investment and a risky investment. The investor pays $100 today, and receives $100 in
the future if he chooses the risk-free investment. In the second choice, he receives $110 half of the
time and $90 otherwise. Clearly, both investments produce the same expected terminal value of
$100, but the investor chooses either investment based on his risk appetite. Over the 1-year time
horizon, the two investments have the same expected IRR of 0%. However, if we assume a time
horizon of 2 years, the expected IRR for the second choice becomes negative (-0.125%) and lower
than the first case. On the other hand, the IRR for the risk-free case is unchanged at 0%. Moreover,
if we look at a 6-month horizon, the relation reverses and the second investment yields a positive IRR
of +1%. (See Table 1 below.)

Table1 : IRRs of a Risky Investment

2-Year Horizon
Time --> 0 1 2 IRR Expected IRR
Investment 2 - Risky -0.125%
Case 1 (50% prob.) -100 0 110 +4.88%
Case 2 (50% prob.) -100 0 90 -5.13%
1Year Horizon
Time --> 0 1 IRR Expected IRR
Investment 2 - Risky 0.000%
Case 1 (50% prob.) -100 110 +10.00%
Case 2 (50% prob.) -100 90 -10.00%
0.5-Year Horizon
Time --> 0 ½ IRR Expected IRR
Investment 2 - Risky +1.000%
Case 1 (50% prob.) -100 110 +21.00%
Case 2 (50% prob.) -100 90 -19.00%
Source: Nomura

Table 1 indicates that an investor who bases his investment decision on the level IRR is likely to
choose the risky investment when the time horizon is shorter, simply because the IRR is higher. In
an extreme case, the investor will take a risky investment that has a 50% chance of earning $1 on a
$100 investment and a 50% chance of losing $100 tomorrow. In such a case, the expected IRR is
5
calculated to be 1,789%! This is obviously absurd.

3
More specifically, the IRR is calculated by iteration using the following equation:
N
CFt
∑ (1 + IRR )
t=0
t
=0

where CFt is the cash flow at time t and N is the time horizon. In other words, IRR is a discount rate at which the
net present value of an investment equals zero.
4
Finance textbooks generally teach that the IRR is a flawed measure for comparing investment returns, because;
(1) it favors a short -term investment over a long-term investment, (2) it assumes a reinvestment rate to be IRR
itself, not the risk-adjusted required rate of return, and (3) it has more than one value when cash flows changes
signs more than once. For a general capital budgeting situation, the net present value (NPV) is a preferred
measure.
5
If there is one cash flow besides the initial investment, IRR can be calculated from the formula above as:

CFt
IRR = t −1
CF0
where t is the date of the final cash flow. Since one day is 1/365 year (t=1/365), IRR of the first case is 3,678% (=
(101/100)^(365) –1). On the other hand, IRR of the second case is –100%. So the expected IRR is:
(.5)(3678%)+(.5)(-100%) = 1789%. This is simply a result of calculating an "annualized" return. If we calculate a

(3)
Nomura Fixed Income Research

The problem gets arguably more complicated when correlation enters the picture. For any tranche of
a credit risk portfolio, the expected IRR is calculated as the probability-weighted average of IRR for
each scenario. Graph 1 below illustrates the probability distribution of IRR for an equity tranche (with
0% attachment point and 3% detachment point) of a hypothetical portfolio, at varying levels of
correlation. The graph illustrates that as correlation increases, the distribution grows at both ends of
the graph (the highest IRR and the lowest IRR). The lowest IRR tends to be a large negative number
(close to –100%), while the highest IRR is often smaller in magnitude, capped around the spread
level.

Graph 1: Distribution of Losses and IRR for Equity Tranche

(Varying correlation)

60%
Positive IRR
(capped at spread)
40%
Probability of Defauls and IRR

20%

IRR
0%
0 (zero 1 2 3 4 5 6 7 or more
loss) (tranche
wiped out) -20%

Low correlation
Medium correlation -40%
Large, negative IRR
High correlation
IRR (right scale)
-60%
Number of Defaults

*Assuming a 125-name portfolio; the 0% -3% tranche; 5-year default probability of 1.58%; recovery rate of 40%;
annual spread of 1,500 bps, and 5-year horizon. Source: Nomura

As a result, the expected IRR can be pulled significantly lower when the IRR at the right end of the
graph becomes a large negative number. In the next section, we analyze the effect of correlation on
equity tranches of differing characteristics using simulation results.

IV. Simulation Analysis

For our analysis, we create a hypothetical CDO that references a 125-name portfolio. We assume a
6
5-year default rate of 1.58% and a recovery rate of 40% for each reference credit. After simulating
tranche losses at zero, 10% and 20% correlation using the one-factor Gaussian copula model, we
calculate the expected IRR for an equity tranche. For the base case, the investment horizon si
assumed to be 5 years. Spreads are paid quarterly. We analyze IRR for equity (first loss) tranches
of various sizes. Unlike an FTD, a loss tranche remains outstanding until losses completely wipe it
out.

In order to calculate IRR, we set up quarterly cash flows for each simulation scenario. The initial
cash flow (at t = 0) is the amount of principal investment that an investor makes, which equals the
notional amount of the tranche. The investor receives spread payments quarterly, calculated as the
quarterly spread in basis points multiplied by the outstanding notional amount for the period.

holding-period return for a given investment horizon, all the three cases above have the same expected return of
0%.
6
This is an equivalent of "Baa2" rating on the Moody's idealized default rate table.

(4)
 Nomura Fixed Income Research

For example, if no default occurs, the notional amount of the tranche remains at the initial tranche
size. An investor will receive the full amount of spreads for 5 years, and at the end of the 5 years
receives the principal amount. For this scenario, IRR will be equal to the spread level with an
adjustment for quarterly compounding (e.g., 10.38% if the tranche spread is 1,000 bps). On the other
th
hand, in a scenario where the tranche is wiped out immediately after the 10 quarter, the investor
th
receives spread for 10 quarters but the spread payment after the 10 quarter is reduced to zero. In
that case, no principal amount is received at the end of 5 years. For this scenario, IRR will be a large
negative number (e.g., -58.2% if the tranche spread is 1,000 bps).

In general, the level of IRR depends on (1) the amount and (2) the timing of each cash flow. In the
context of a CDO tranche, the amount of each cash flow depends on a tranche’s size, investment
horizon, and default probability. On the other hand, the timing of each cash flow depends on the time
horizon of the transaction and the number and timing of defaults.

Below, we analyze the relationship between a portfolio’s correlation and a tranche’s expected IRR as
we vary tranche size, spread level, and time horizon.

A. Changing Tranche Size

Graph 2 compares the expected IRR of equity tranches of various sizes. The tranche spread is
assumed to be 1,500 bps per annum. In the graph, the expected IRR is mostly higher for a low
correlation portfolio than for a higher correlation portfolio, but the lines cross around the tranche size
of 2%. To the right of this crossover point, higher correlation results in a lower expected IRR. On the
other hand, for the 0%-1% tranche, higher correlation is associated with a higher expected IRR. At
1% size, where the tranche is thinnest, the order is clear where higher correlation is associated with a
higher IRR. The result is consistent with the FTD example in the previous section. Given its thin
size, the 0%-1% tranche is similar to an FTD basket. Its return primarily depends on the likelihood of
the extreme outcome (i.e., zero default), because in all other cases (i.e., non-zero default) the 0%-1%
tranche is almost or completely wiped out.

Graph 2: Comparison of Equity Tranche Returns

(Changing tranche size and portfolio correlation)

20

15

10

5
Expected IRR (%)

0
1%

2%

3%

4%

5%

6%

7%

8%

9%

10%

-5

-10

-15

-20

-25

-30
Tranche Size

IRR (corr = 0%) IRR (corr = 10%) IRR (corr = 20%)

* A hypothetical portfolio of 125 names. The assumptions used are; recovery rate of 40%; 5-year default rate of
1.58%; time horizon of 5 years, and tranche spread of 1,500 bps. Source: Nomura

(5)
Nomura Fixed Income Research

Graph 2 suggests that whether higher correlation hurts an equity tranche’s expected IRR or not
depends on the tranche’s size. Our results suggest that IRR on a thin equity tranche increase with
7
higher correlation.

Graph 3 plots the relationship between the portfolio correlation (X-axis) and the equity tranche's
expected IRR (Y-axis). We assumed a spread of 1,500 bps for three equity tranches with different
sizes; 0%-1%, 0%-3%, 0%-5%, and the overall portfolio (i.e., 0%-100%). The graph confirms that the
expected IRR of a thin equity tranche (i.e., 0%-1% tranche) is consistent with conventional wisdom,
where the expected IRR increases for higher correlation. As before, the reverse relationship holds for
thicker equity tranches (the 0%-3% and 0%-5% tranches). Interestingly, the line is almost flat for the
overall portfolio (the dotted line; the 0%-100% tranche).

Graph 3: Comparison of Equity Tranche Returns

(Changing portfolio correlation and tranche size)

20

15

10

5
Expected IRR (%)

0
0%

2%

4%

6%

8%

10%

12%

14%

16%

18%

20%
-5

-10

-15

-20

-25

-30
Correlation

0%-1% Tranche 0%-3% Tranche 0%-5% Tranche 0%-100% Tranche

* A hypothetical portfolio of 125 names. The assumptions used are; recovery rate of 40%; 5-year default rate
of 1.58%; time horizon of 5 years, and tranche spread of 1,500 bps. Source: Nomura

B. Changing Default Probability

The level of default risk of an underlying portfolio also affects the tranche IRR. Graph 4 compares the
IRR of equity tranches of various sizes. The difference between Graph 4 and Graph 2 is that the
assumed default probability is higher for Graph 4 at 3.05% over 5 years ("Baa3" level), than for Graph
2 where we assumed a 1.58% default probability ("Baa2" level). When comparing Graph 4 and
Graph 2, we see that the crossover points for different correlation levels have moved to the right. The
20%-correlation line lies above both the 10%-correlation and zero-correlation lines for the 1% and 2%
tranche sizes. In other words, not only the 0%-1% tranche but also the 0%-2% is associated with
higher expected IRRs when correlation is higher.

7
For the overall portfolio (i.e., 0%-100%), IRRs are almost identical.

(6)
 Nomura Fixed Income Research

Graph 4: Comparison of Equity Tranche Returns -- Higher Default Risk

(Changing tranche size and portfolio correlation)

20

10

10%
1%

2%

3%

4%

5%

6%

7%

8%

9%
Expected IRR (%)

-10

-20

-30

-40

-50

-60
Tranche Size

IRR (corr = 0%) IRR (corr = 10%) IRR (corr = 20%)

* A hypothetical portfolio of 125 names. The assumptions used are; recovery rate of 40%; 5-year default rate of
3.05%; time horizon of 5 years, and tranche spread of 1,500 bps. Source: Nomura

As with Graph 3, Graph 5 plots the relationship between the portfolio correlation (X-axis) and the
equity tranche's expected IRR (Y-axis) for higher default probabilities. As before, we assumed a
spread of 1,500 bps for three equity tranches with different sizes; 0%-1%, 0%-3%, and 0%-5%.
Graph 5 confirms that the IRR of a thin equity tranche (i.e., 0%-1% tranche) exhibits a relationship
that is consistent with conventional wisdom, where the IRR increases for higher correlation. The
reverse relationship holds for thicker equity tranches (0%-3% and 0%-5% tranches).

Graph 5: Comparison of Equity Tranche Returns -- Higher Default Risk

(Changing portfolio correlation and tranche size)

10

-10
Expected IRR (%)

-20

-30

-40

-50

-60
10%

12%

14%

16%

18%

20%
0%

2%

4%

6%

8%

Correlation

0%-1% Tranche 0%-3% Tranche 0%-5% Tranche

* A hypothetical portfolio of 125 names. The assumptions used are; recovery rate of 40%; 5-year default rate
of 3.05%; time horizon of 5 years, and tranche spread of 1,500 bps. Source: Nomura

C. Changing Time Horizon

As we have briefly shown in Section III, expected IRR is affected by the time horizon of an
investment. We checked the relation using a shorter time horizon. We restructured the cash flows

(7)
Nomura Fixed Income Research

over 1 year, instead of the 5-year horizon as we initially assumed. Like Graph 2, Graph 6 compares
the expected IRRs of equity tranches of various sizes, but over a 1-year horizon. Graph 6 shows that
for the 1-year horizon, the 20%-correlation line stays above the lower-correlation lines between the
1% point and the 5% point, though by very small amounts. In other words, the IRR is higher for the
20%-correlation portfolio than for either of the lower-correlation portfolios for equity tranches of 1%
and 5% in size. The results suggest that the relation between portfolio correlation and the IRR of an
equity tranche also depends on the time horizon. We suspect that it is related to the degree of
8
compounding in the IRR calculation.

Graph 6: Comparison of Equity Tranche Returns

(1-year Horizon: changing tranche size and portfolio correlation)

15

10
Expected IRR (%)

10%
1%

2%

3%

4%

5%

6%

7%

8%

9%
-5

-10
Tranche Size

IRR (corr = 0%) IRR (corr = 10%) IRR (corr = 20%)

* A hypothetical portfolio of 125 names. The assumptions used are; recovery rate of 40%; 5-year default rate
of 1.58%; time horizon of 1 year, and tranche spread of 1,500 bps. Source: Nomura

V. Conclusion

Our analysis has shown that the relationship between correlation and the expected IRR of a CDO
equity tranche is not monotonic. In general, the expected IRR of a thin equity tranche tends to
increase for higher correlation, but the relation reverses for a thicker tranche under certain
circumstances. Other factors that affect the relationship include assumptions about default
probability and time horizon. Given the non-monotonic relationship between an equity tranche’s
expected IRR and portfolio correlation, we believe that a case-by-case analysis is warranted when
one focuses on the IRR as a measure of an investment's desirability. In the meantime, the
conventional wisdom that an equity tranche’s expected losses decline as portfolio correlation
increases remains valid.

— E N D —

8
In the capital budgeting context, the IRR is known to punish a long-term project compared to a short-term
project, even when the net present value (NPV) evaluated at the same cost of capital is equivalent for the two
projects.

(8)
 Nomura Fixed Income Research

VI. Recent Nomura Fixed Income Research

Fixed Income General Topics
• Report from Arizona 2005: Coverage of Selected Sessions of ABS West 2005 (14 Feb
2005)
• U.S. Fixed Income 2005 Outlook/2004 Review (16 Dec 2004)
• U.S. Fixed Income 2004 Mid-Year Outlook/Review (1 Jul 2004)
ABS/CDO
• CDOs-Squared Demystified (4 Feb 2005)
• Student Loan ABS 101 (26 Jan 2005)
• ABS/MBS Disclosure Update #6: 24 Steps to Tighter ABS – Regulation AB (27 Dec
2004)
• ABS Credit Migrations 2004 (7 Dec 2004)
• Home Equity ABS Basics (1 Nov 2004)
• “The Bespoke [bispóuk]” - A Guide to Single-Tranche Synthetic CDOs (17 Nov 2004)
• Tranching Credit Risk (8 Oct 2004)
• CDOs in Plain English (13 Sep 2004)
• Correlation Primer (6 Aug 2004)
• ABS/MBS Disclosure Update #5: Reactions to the Comment Letters (4 Aug 2004)
• ABS/MBS Disclosure Update #4: Issues from ASF Sunset Seminar (13 May 2004)
• ABS/MBS Disclosure Update #3: Start Your Engines – Get Ready to Comment
(10 May 2004)
• CDS Primer (12 May 2004)
MBS
• GNPL REMIC Factor Comparison (18 Jan 2005)
• GNPL Prepayments January Factor (13 Jan 2005)
• FHA/VA Monthly (11 Jan 2005)
• GNPL Prepayments November Factor (10 Nov 2004)
• Special Report ERISA Reform (29 Sep 2004)
• GNPL REMIC Factor Comparison (20 Sep 2004)
Strategy
• Back of the Envelope" Commercial Loan Extension Analysis (18-Mar-05)
• MBS Market Check-up: March Update (15-Mar-05)
• CRB Index Record Levels: Impacting current interest rates? (14-Mar-05)
• TRIA Update: Will the Government's role continue? (10-Mar-05)
• Trade Recommendation: GNMA Project Loans (10-Mar-05)
• Update of Bankruptcy Reform (10-Mar-05)
• NAR's Pending Home Sales Index: new housing indicator (09-Mar-05)
• Trade Idea: Steeper Swap Curve, Wider 10-Year Spreads (07-Mar-05)
• Agency 5% PACs looking good (04-Mar-05)
• Floored Inverse Floaters: Gaining in Popularity (28-Feb-05)
• Commercial Real Estate Update-Hotel Sector (24-Feb-05)
• Pension Fund Reform and its Implications (22-Feb-05)
• Commercial Real Estate Update-Industrial Sector (16-Feb-05)
• Commercial Real Estate Update-Office Sector (16-Feb-05)
• MBS Market Check-up: February Update (15-Feb-05)
• Trade Recommendation: FNMA vs. FHLMC Agency Bullets (14-Feb-05)
• Social Security Reform and its implications.. (11-Feb-05)
• Commercial Real Estate Update - Apartment Sector (11-Feb-05)
• Commercial Real Estate Update - Retail Sector (11-Feb-05)
• Social Security Reform and it implications… (8 Feb 2005)
• B-Notes and Mezzanine Debt: A Primer (2 Feb 2005)
• Is U.S. Housing Price Inflation properly reflected in the CPI Index? (2 Feb 2005)
• Trade Recommendation: Re-evaluate 5-Yr FHLMC Holdings (27 Jan 2005)
• MBS Relative Value: Net Carry within the Coupon Stacks (25 Jan 2005)
Corporates
• General Motors: 2005 Earnings Revision (16 Mar 2005)
• Corporate Relative Value (15 Mar 2005)
• Corporate Weekly: For the week ended 3/11/05 - 3/14/05
• Toys "R" Us: Potential Sale (10 Mar 2005)
• US Corporate Sector Review: February 2005 (9 Mar 2005)

(9)
 NEW YORK TOKYO LONDON
Nomura Securities International Nomura Securities Company Nomura International PLC
2 World Financial Center, Building B 2-2-2, Otemachi, Chiyoda-Ku Nomura House
New York, NY 10281 Tokyo, Japan 100-8130 1 St Martin's-le-grand
(212) 667-9300 81 3 3211 1811 London EC1A 4NP
44 207 521 2000

David P. Jacob 212.667.2255 International Head of Research

Nomura U.S. Fixed Income Research

David Resler 212.667.2415 Head of U.S. Economic Research Elizabeth Bartlett 212.667.2339 Analyst
Mark Adelson 212.667.2337 Securitization/ABS Research Edward Santevecchi 212.667.1314 Analyst
John Dunlevy 212.667.9298 Cross Market Strategist Tim Lu 212.667.2392 Analyst
Arthur Q. Frank 212.667.1477 MBS Research Jeremy Garfield 212.667.2158 Analyst
Louis (Trey) Ott 212.667.9521 Corporate Bond Research Diana Berezina Analyst
Weimin Jin 212.667.9679 Quantitative Research Xiang Long Quantitative Analyst
Michiko Whetten 212.667.2338 Quantitative Credit Analyst Cristian Pasarica Quantitative Analyst
James Manzi 212.667.2231 CMBS Research/Strategy Kumiko Kimura Translator
Tomoko Nago-Kern Translator

I Michiko Whetten, a research analyst employed by Nomura Securities International, Inc., hereby certify that all of the views expressed in this
research report accurately reflect my personal views about any and all of the subject securities or issuers discussed herein. In addition, I
hereby certify that no part of my compensation was, is, or will be, directly or indirectly related to the specific recommendations or views that I
have expressed in this research report.

© Copyright 2005 Nomura Securities International, Inc.

This publication contains material that has been prepared by one or more of the following Nomura entities: Nomura Securities Co., Ltd. ("NSC") and Nomura Research
Institute, Ltd., Tokyo, Japan; Nomura International plc and Nomura Research Institute Europe, Limited, United Kingdom; Nomura Securities International, Inc. ("NSI") and
Nomura Research Institute America, Inc., New York, NY; Nomura International (Hong Kong) Ltd., Hong Kong; Nomura Singapore Ltd., Singapore; Nomura Australia Ltd.,
Australia; P.T. Nomura Indonesia, Indonesia; Nomura Advisory Services (Malaysia) Sdn. Bhd., Malaysia; Nomura Securities Co., Ltd., Taipei, Taiwan; or Nomura Securities
Co., Ltd., Seoul, Korea. This material is: (i) for your private information, and we are not soliciting any action based upon it; (ii) not to be construed as an offer to sell or a
solicitation of an offer to buy any security in any jurisdiction where such an offer or solicitation would be illegal; and (iii) is based upon information that we consider reliable, but
we do not represent that it is accurate or complete, and it should not be relied upon as such. Opinions expressed are current opinions as of the date appearing on this
material only and the information, including the opinions contained herein are subject to change without notice. Affiliates and/or subsidiaries of Nomura Holdings, Inc.
(collectively referred to as the “Nomura Group”) may from time to time perform investment banking or other services (including acting as advisor, manager or lender) for, or
solicit investment banking or other business from, companies mentioned herein. The Nomura Group, its officers, directors and employees, including persons involved in the
preparation or issuance of this material may, from time to time, have long or short positions in, and buy or sell (or make a market in), the securities, or derivatives (including
options) thereof, of companies mentioned herein, or related securities or derivatives. Fixed income research analysts, including those responsible for the preparation of this
report, receive compensation based on various factors, including quality and accuracy of research, firm’s overall performance and revenue (including the firm’s fixed income
department), client feedback and the analyst’s seniority, reputation and experience. NSC and other non-US members of the Nomura Group, their officers, directors and
employees may, to the extent it relates to non-US issuers and is permitted by applicable law, have acted upon or used this material, prior to or immediately following its
publication. Foreign currency-denominated securities are subject to fluctuations in exchange rates that could have an adverse effect on the value or price of, or income
derived from the investment. In addition, investors in securities such as ADRs, the values of which are influenced by foreign currencies, effectively assume currency risk. The
securities described herein may not have been registered under the U.S. Securities Act of 1933, and, in such case, may not be offered or sold in the United States or to U.S.
persons unless they have been registered under such Act, or except in compliance with an exemption from the registration requirements of such Act. Unless governing law
permits otherwise, you must contact a Nomura entity in your home jurisdiction if you want to use our services in effecting a transaction in the securities mentioned in this
material. This publication has been approved for distribution in the United Kingdom by Nomura International plc, which is authorised and regulated by The Financial Services
Authority (“FSA”) and is a member of the London Stock Exchange. It is intended only for investors who are “market counterparties” or “intermediate customers” as defined by
FSA, and may not, therefore, be redistributed to other classes of investors. This publication has also been approved for distribution in Hong Kong by Nomura International
(Hong Kong) Ltd. NSI accepts responsibility for the contents of this material when distributed in the United States .. This publication has also been approved for distribution in
Singapore by Nomura Singapore Limited. No part of this material may be (i) copied, photocopied, or duplicated in any form, by any means, or (ii) redistributed without
Nomura's prior written consent. Further information on any of the securities mentioned herein may be obtained upon request. If this publication has been distributed by
electronic transmission, such as e-mail, then such transmission cannot be guaranteed to be secure or error-free as information could be intercepted, corrupted, lost,
destroyed, arrive late or incomplete, or contain viruses. The sender therefore does not accept liability for any errors or omissions in the contents of this publication, which may
arise as a result of electronic transmission. If verification is required, please request a hard-copy version.