Global Viewpoint

Issue No: 06/08

February 8, 2006

GS GLOBAL ECONOMIC WEBSITE

Economic Research from the GS Institutional Portal
at https://portal.gs.com

The Bond Yield Sudoku

Francesco Garzarelli
francesco.garzarelli@gs.com

+44 (0)20 7774 5078

Michael Vaknin
michael.vaknin@gs.com
+44 (0)20 7774 1386
Sergiy Verstyuk
sergiy.verstyuk@gs.com
+44 (0)20 7774 1173

Even accounting for new macro and behavioural forces, there seems to be no escape
from the conclusion that longer-dated bond yields across the main markets are stretched,
according to our analysis. Based on fresh empirical work, we find that the problem seems
to be more acute in Europe, exacerbated by pension regulations, and comparatively less
pronounced in Japan.

1. Finding a Pattern in the Numbers

Explaining the behaviour of longer-dated bond yields through economic fundamentals has
never been simple, but in the past couple of years this has turned into a truly daunting task.
Standard econometric models relating the level of 10-yr rates to a set of macroeconomic factors,
such as inflation and growth, have broken down in what appears to be a meaningful and
persistent deviation from historical norms.
In this note, we review these issues in light of some of the arguments recently advanced to
explain the disappearance of the bond term premium. We also present new empirical work
designed to identify which countries bond yields are trading most out of line relative to those
of their peers, controlling for domestic fundamentals and for international spill-over effects.
We reach three main conclusions:

The so-called bond conundrum is not a prerogative of the US alone, but has a global
connotation. Granted, the heavy buying of US Treasuries by foreign central banks may be
also influencing the pricing of bonds in the rest of the G-3, but this is hardly the end of the
story, according to our analysis.

There is evidence that investors are adapting their medium-term interest rate expectations to
the notion that inflation uncertainty has diminished a development we think is underpinned
by valid reasons. But it is possible that this adjustment may now have gone overboard, and
turned into an attitude that is too complacent towards inflation risks.

The duration-phoria fuelled by pension fund regulation in some countries has opened a
comparatively larger valuation gap on longer-dated European bonds, with possible
international ramifications. This is the sense in which the bond conundrum could become
self-reinforcing, and potentially more destabilizing.

2. The Work-Horses Back is Broken

Our traditional bond modelling exercise has consisted of a set of single-country equations,
where 10-yr yields are related to a set of domestic macroeconomic variables. An illustration for
the United States, Japan and Germany1 is provided in the accompanying charts.
1. We model Germany instead of an aggregate EMU-3 yield in order not to incur problems associated with the advent
of EMU, which would require special consideration and detract from our main point of interest.
Important disclosures appear at the back of this document

Goldman Sachs Economic Research

Global Viewpoint

In the standard specification, the yield offered by a

benchmark 10-yr note is associated with the level of short
rates, survey-based 1-year-ahead inflation expectations,
and activity growth. As can be seen in the graphs, the
models works reasonably well up to 2004, after which
actual yields start deviating from their fitted values in all
three countries considered.

US: Baseline Model

11
10
9
8
7

It has been argued that a good portion of the valuation

gap could be attributed to the massive accumulation of
US fixed income securities by foreign central banks,
particularly those in Asia and the Middle East. This
would mean that the standard models are missing an
important explanatory variable for the behaviour of
longer-dated yields. In previous work, we have shown
that including a proxy for this increasingly important
portion of demand does indeed correct some of the
deviation between actual and fitted yields.

6
5
4
3

Actual
Baseline Model

2
90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

Germany: Baseline Model

10

3. The Role of Foreign Central Banks

The chart at the top of the next page juxtaposes the fitted
values for US bond yields on our baseline model and one
including a measure of the undervaluation of Asian
currencies against the US Dollar, taken as an indirect
gauge of the regions central banks purchases of US
Treasuries.2 As can be seen, the valuation gap is smaller
to the tune of an average 36bp over the past two years.
Accounting for official purchases, the model suggests that
10-yr US Treasuries should presently trade at 5.5%, or
around 75bp below the baseline result.

9
8
7
6
5
4
3

Actual
Baseline Model

This result warrants an important qualification, however.

The exercise we conduct above involves jointly assessing
the validity of our baseline framework and the statistical
relevance of the new variable as an explanatory factor. In
other words, the underlying model may well be no longer
valid, and Asian purchases just happen to be the right
plug. The fact that other researchers using time series
methods have produced estimates for the impact of
foreign central bank purchases ranging between 30bp and
200bp underscores the large uncertainty surrounding
these findings.3

2
90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

Japan: Baseline Model

9
8

Actual

Baseline Model

6
5

Even with this caveat, central bank buying is not able to

fully explain the US valuation gap, let alone the fact that
this has also opened up in other bond markets. The bond
conundrum has a global connotation, and global factors
other than the re-cycling of FX reserves may be at its
origin.

4
3
2
1
0
90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

2. The variable is constructed as the percentage deviation from our GS-DEER fair value for Asian crosses against the greenback. The countries considered
include Japan, China, Hong Kong, India, Indonesia, Korea, Malaysia, Philippines, Taiwan, Thailand and Singapore, and are aggregated using GDP
weights. In previous work (see: Global Markets Daily, How Much Is Asian Buying Worth, 9 December 2004) we used US Treasury International
Capital System (TIC) data as a more direct proxy for foreign official demand, but this exposes our results to a greater endogeneity bias. In fact, we
would be fitting a relationship between prices on the left-hand side and quantities on the right-hand side. For all practical purposes, the estimations carried
out using the FX-misvaluation proxy and the TIC data come broadly to the same conclusion: not all of the bond valuation gap can be reconciled.
3. See R.N. McCauley, Assessing the Asian Bid for US Bonds, BIS, Presentation to Bond Market Association, New York 20 April 2005.

Issue No: 06/08

February 8, 2006

Goldman Sachs Economic Research

10

Global Viewpoint

US: 'Asian Buying' -- A Primary, But Not

the Only Suspect

US: Lower Sensitivity to Inflation Vol May

Have Driven Yields Down

11

10

9
8

Actual

Actual

Bas eline m odel with inflation vol

Baseline model with FX misalignment

Bas eline m odel with inflation vol

(allowing for s hifts in vol s ens itivity)

90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

monetary and fiscal policies and energy price increases,

as witnessed over the past two to three years, may have
been instrumental to this learning experience, adding in
the meanwhile to pressures to lock in real returns.

4. Enter Inflation Volatility

In a recent research note, Bill Dudley argues that the
decline in the US bond term premium the additional
compensation provided by longer-dated bonds over and
above the expected future path for short rates may be
attributable to a decline in inflation volatility, both in
absolute terms and in relation to that of the industrial
cycle.4 Adapting to this change in the environment,
investors may be now demanding a lower premium to
bear inflation risks. With a greater stability of nominal
variables Bill also maintains investors may wish to
lock in real cash flows rather than face re-investment
risks, especially in the aftermath of a big financial shock,
such as the bursting of the tech bubble. The appeal of this
argument also resides in the fact that it is applicable not
only to the US, but also to the rest of the G-3. Indeed, the
decline in inflation variability is an international
development.

We calculate that the size of the coefficient on the

additional inflation volatility variable has dropped
significantly since 2004, meaning that US long bonds
have become less reactive to inflation changes. This
finding can be extended also to Japan. Controlling for this
effect through a dummy interacting with the inflation
volatility variable, we find that US 10-yr yields are
trading only 40bp, below their fundamental fair value.

5. Look Whos More Out of Whack!

However one looks at the problem, there seems to be no
escape from the suspicion that the level of longer-dated
yields across the G-3 may be too low. At least this is what
we find even after hammering the standard valuation
models in order to account for what could be new in
todays world be it Asian reserve accumulation, or a
more sanguine assessment of inflation risks. Importantly,
the problem does not appear to originate solely on US
shores.

A key question, however, is why the market has only

recently started to recognize such an important shift in the
macro landscape. After all, the uncertainty on inflation
(measured both ex-post using 5-yr rolling averages and
ex-ante, through the distribution of 1-yr ahead CPI
forecasts collated by Consensus Economics) has not
meaningfully changed since the mid-1990s, while the
drop in longer-dated real forwards has mostly occurred
over the past two years. Introducing inflation variability
as an explanatory variable in our baseline model for 10-yr
USTs does indeed improve the regression fit, but not to
the extent of closing the yawning valuation gap in place
since 2004.

If the bond yield puzzle cannot be cracked completely, it

may be at least useful to track where, in relation to
historical norms, each of the major bond markets is
trading relative to its peers, after controlling for domestic
fundamentals. This may reveal cross-country
opportunities irrespective of overall valuations.
To this end, we estimate 10-yr bond yields in the US,
Japan and Germany simultaneously, allowing for each of
them to be influenced by the others. So, for example, the
forecast of US 10-yr yield is determined by US macro
factors and by the contemporaneous fitted values of Bund

One explanation could be that investors needed a stronger

lesson to appreciate that inflation risks have been
eradicated.5 The broad stability of core inflation and
inflation expectations at a time of very accommodative

4. US Economics Analyst Low Bond Risk Premia: The Collapse of Inflation Volatility, 6 January 2006.
5. What we have in mind is similar in spirit to the Bayesian argument put forward by Thomas Sargent and Timothy Cogley in their paper The
Market Price of Risk and the Equity Premium: A Legacy of the Great Depression?. The memory of runaway inflation in the 1970s and the heavyhanded cure imposed by central banks in the 1980s may have fed exaggerated fears of price instability and radical policy reactions, leading to high
bond term premia through the 1990s. These fears gradually wore off as investors slowly adapted their expectations to changes in the macro
environment, notably the low and stable inflation under varying economic conditions.

Issue No: 06/08

February 8, 2006

Goldman Sachs Economic Research

Global Viewpoint

Actual and Model 10-yr Yields - At a Glance, end-2005

US: Simultaneous Model

10
Percent (S.E.*)

9
8
7

Germany Japan

Actual yields

4.53

3.39

Baseline model

6.25

4.82

1.90

(0.73)

(0.72)

(0.55)

5.48

Baseline w/ Asian buying

United
States

1.51

(0.65)

Actual

Baseline w/ inflation uncertainty

4
Simultaneous model

6.20
(0.74)

Ditto also w/ interaction dummy

4.90
(0.61)

2
90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

Simultaneous model

4.78

4.33

1.54

(0.34)

(0.38)

(0.39)

* Standard error of regression.

Germany: Simultaneous Model

across the individual countries over our sample period.

This means that in our 15-yr sample there are one or more
unobserved factors jointly driving the term premium on
longer-dated yields in the G-3. For example, all major
bond markets were cheap after the 1994 Fed monetary
shock, and all became rich in the aftermath of the
1997 Asian crisis and the deflationary scare that followed
that event.

10
9
8
7
6
5

Secondly, Japanese bonds currently appear to be the

closest aligned to their own domestic fundamentals and
the international yield linkage. At the opposite side of the
spectrum are German Bunds, which are grossly
expensive, while US Treasuries are only modestly overvalued. Holding all else equal, Bunds could sell off 100bp
returning to their fitted value without this impacting the
equilibrium level of JGBs. Also, according to the model,
the spread between 10-yr Treasuries and Bunds should be
trading closer to 50bp than the current 100bp.

Actual

Simultaneous model

3
2

90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

Japan: Simultaneous Model

9
8

The relative richness of German yields highlighted by

our model is material by both statistical and historical
standards. One plausible explanation is to be sought in the
pension fund actuarial solvency rules, which are forcing
the purchase of long duration securities in several
European countries in the face of limited supply. While
some of this impact may have already spilled over into
other markets, with potentially destabilizing effects
(witness the hype already surrounding the re-introduction
of the 30-yr maturity in the US), the nature of the rules,
which foster a home bias, makes the relative
misalignment more persistent and difficult to arbitrage
away.

Actual

7
6

Simultaneous model

5
4
3
2
1
0
90 91 92 93 94 95 96 97 98 99 00 01 02 03 04 05

That said, both the extreme magnitude of the relative

valuation differential, and the fact that the focus on pension
asset-liability mismatches is shifting to the US, suggest it
may be worthwhile positioning for a reduction of the gap
between Europe and other markets. Feeding the model with
our baseline macroeconomic projections reinforces this
view.

and JGB yields. If there is some force altering the

relationship between, say, Bunds and German
fundamentals, the model would pass this on to Treasuries
and JGBs (and vice-versa). Our results are summarized in
the charts above. Two elements stand out most.
Firstly, departures from fair value are typically common

Issue No: 06/08

February 8, 2006

Goldman Sachs Economic Research

Global Viewpoint

6. Investment Conclusions
New forces help explain some of the decline in global
bond yields over the past couple of years. While data
limitations, the proximity of the possible break in historical
relationships, and distortions created by pension fund
demand hinder reaching strong conclusions, the suspicion
is that long-term rates may now have fallen too far.
Controlling for both macro factors and for term premia
spill-overs from one bond market to another, our work
shows that Japanese bonds offer the best comparative value
across the G-3, and German Bunds offer the least.
Based on our analysis, three positions stand out. One is
part of our Top Trades for this year, and involves
receiving 5-yr 5-yr forward real rates in the US against
paying them in Europe. We would also pay 10-yr EUR
swaps to receive 10-yr JPY swaps, against holding
opposite positions in 2-yr rates to control for the risks in
the direction of monetary policies. The cost-of-carry on
this box trade is marginally positive. Finally, we would
hold a 1-yr into 30-yr EUR payer, struck at 4.50%.
Francesco Garzarelli, Mike Vaknin and Sergiy Verstyuk

Issue No: 06/08

February 8, 2006

Goldman Sachs Economic Research

Global Viewpoint

Technical Appendix
Our econometric approach is based on the simple concept that bond yields in the United States, Germany and Japan
are determined simultaneously, and are driven by both domestic and foreign factors. In our estimation, we rely on a
methodology similar to the GVAR-VARX* approach of Pesaran and Smith.6 Specifically, we run all bond yield
equations in a unified structure, where each countrys 10-yr rate is related to its domestic macroeconomic
fundamentals, as well as the contemporaneous global level of 10-yr rates. The global level in each countrys equation
is defined as a weighted average of foreign rates, with corresponding countries trade shares used as weights
(technically, this is implemented by imposing a linear constraint on the coefficients for foreign rates.)
Using the US as an example, we posit that 10-yr Treasury yields (10YRUS) are a function of US domestic
fundamentals (expected inflation INFUS, industrial production growth IPUS, and 3-month rates 3MUS), as well as of
yields on German Bunds and JGBs (10YRDM and 10YRJP). Repeating for Germany and Japan results in the following
structural system:
a2INF U S

(US)

10YRUS

= a1

(DM)

10YRD M

= b1

+ b2INF D M

(JP)

10YRJP

= c1

+ c 2INF JP

+ a3IP U S

a43M US

+ a510YRDM

+ a610YRJP

b3IP DM

b43M DM

+ b510YRU S

+ b610YRJP

c 3IP JP

+ c 43M JP

+ c 510YRU S + c 610YRD M

We run a two-stage least squares (2SLS) estimation procedure. In the first stage, each of the endogenous foreign 10-yr
rates on the right-hand side is instrumented by all the macro variables. Since expected inflation and 3-month rates
are also treated as endogenous variables here, we instrument them too, and our list of instrumental variables
therefore includes the lags of expected inflation and 3-month rates. In the second stage, we substitute into the
structural form equations above the proxies for the foreign 10-yr rates obtained in the first stage, and only then
estimate the structural models coefficients.
After estimating the structural model and solving for 10-yr rates, our results can be expressed in the following reduced
form (taking the equation for Japan as an example):
(JP)

10YRJP

= 1.35

+ 0.15INF JP

+ 0.01IP JP

+ 0.613M JP

- 0.34INF U S

+ 0.03IP US

+ 0.173M US

+ 0.05INF D M

- 0.01IP D M

+ 0.13M D M

As can be seen from the equation above, our predicted yields are not calculated using actual foreign yields directly.
Rather, the simultaneous equation solution amounts to solving a system of three unknowns (i.e., the 10-yr rates), so
that in the end bond yields are expressed solely as functions of the macro fundamentals.

6. Pesaran, M.H., Smith, R. 2006. Macroeconomic Modelling with a Global Perspective. Mimeo.

Issue No: 06/08

February 8, 2006

Goldman Sachs Economic Research

Global Viewpoint

Copyright 2006 The Goldman Sachs Group, Inc. All rights reserved.
This material should not be construed as an offer to sell or the solicitation of an offer to buy any security in any jurisdiction where such an offer or solicitation would be illegal. We are not soliciting any action based on
this material. It is for the general information of clients of The Goldman Sachs Group, Inc. It does not constitute a personal recommendation or take into account the particular investment objectives, financial situations,
or needs of individual clients. Before acting on any advice or recommendation in this material, clients should consider whether it is suitable for their particular circumstances and, if necessary, seek professional advice.
The price and value of the investments referred to in this material and the income from them may go down as well as up, and investors may realize losses on any investments. Past performance is not a guide to future
performance. Future returns are not guaranteed, and a loss of original capital may occur. The Goldman Sachs Group, Inc. does not provide tax advice to its clients, and all investors are strongly advised to consult with
their tax advisers regarding any potential investment. Certain transactions - including those involving futures, options, and other derivatives as well as non-investment-grade securities - give rise to substantial risk and
are not suitable for all investors. The material is based on information that we consider reliable, but we do not represent that it is accurate or complete, and it should not be relied on as such. Opinions expressed are our
current opinions as of the date appearing on this material only.
We endeavor to update on a reasonable basis the information discussed in this material, but regulatory, compliance, or other reasons may prevent us from doing so. We and our affiliates, 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, act as principal in, and buy or sell the securities or derivatives (including
options) thereof of companies mentioned herein. For purposes of calculating whether The Goldman Sachs Group, Inc. beneficially owns or controls, including having the right to vote for directors, 1% of more of a class
of the common equity security of the subject issuer of a research report, The Goldman Sachs Group, Inc. includes all derivatives that, by their terms, give a right to acquire the common equity security within 60 days
through the conversion or exercise of a warrant, option, or other right but does not aggregate accounts managed by Goldman Sachs Asset Management. No part of this material may be (i) copied, photocopied, or
duplicated in any form by any means or (ii) redistributed without The Goldman Sachs Group, Inc.s prior written consent.
The Global Investment Research Division of Goldman Sachs produces and distributes research products for clients of Goldman Sachs, and pursuant to certain contractual arrangements, on a global basis. Analysts
based in Goldman Sachs offices around the world produce equity research on industries and companies, and research on macroeconomics, currencies, commodities and portfolio strategy.
This research is disseminated in Australia by Goldman Sachs JBWere Pty Ltd (ABN 21 006 797 897) on behalf of Goldman Sachs; in Canada by Goldman Sachs Canada Inc. regarding Canadian equities and by
Goldman Sachs & Co. (all other research); in Germany by Goldman Sachs & Co. oHG; in Hong Kong by Goldman Sachs (Asia) L.L.C.; in Japan by Goldman Sachs (Japan) Ltd; in the Republic of Korea by Goldman
Sachs (Asia) L.L.C., Seoul Branch; in New Zealand by Goldman Sachs JBWere (NZ) Limited on behalf of Goldman Sachs; in Singapore by Goldman Sachs (Singapore) Pte. (Company Number: 198602165W); and
in the United States of America by Goldman, Sachs & Co. Goldman Sachs International has approved this research in connection with its distribution in the United Kingdom and European Union.This material has been
issued by The Goldman Sachs Group, Inc. and/or one of its affiliates and has been approved for the purposes of section 21 of the Financial Services and Markets Act 2000 by Goldman Sachs International, which is
regulated by the Financial Services Authority, in connection with its distribution in the United Kingdom, and by Goldman Sachs Canada, in connection with its distribution in Canada. Goldman Sachs International and
its non-US affiliates may, to the extent permitted under applicable law, have acted on or used this research, to the extent that it relates to non-US issuers, prior to or immediately following its publication. Foreigncurrency-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. In addition, options involve risk and are not suitable for all investors. Please ensure that you have read and understood
the current options disclosure document before entering into any options transactions.
Further information on any of the securities mentioned in this material may be obtained on request, and for this purpose, persons in Italy should contact Goldman Sachs S.I.M. S.p.A. in Milan or its London branch
office at 133 Fleet Street; persons in Hong Kong should contact Goldman Sachs (Asia) L.L.C. at 2 Queens Road Central; persons in Australia should contact Goldman Sachs JBWere Pty Ltd. (ABN 21 006 797 897),
and persons in New Zealand should contact Goldman Sachs JBWere( NZ) Ltd . Persons who would be categorized as private customers in the United Kingdom, as such term is defined in the rules of the Financial
Services Authority, should read this material in conjunction with the last published reports on the companies mentioned herein and should refer to the risk warnings that have been sent to them by Goldman Sachs
International. A copy of these risk warnings is available from the offices of Goldman Sachs International on request. A glossary of certain of the financial terms used in this material is also available on request.
Derivatives research is not suitable for private customers. Unless governing law permits otherwise, you must contact a Goldman Sachs entity in your home jurisdiction if you want to use our services in effecting a
transaction in the securities mentioned in this material.

Issue No: 06/08

February 8, 2006