Federal Reserve Bank of New York

Staff Reports

The Term Structure of Inflation Expectations

Tobias Adrian
Hao Wu

Staff Report no. 362

February 2009

This paper presents preliminary findings and is being distributed to economists

and other interested readers solely to stimulate discussion and elicit comments.
The views expressed in the paper are those of the authors and are not necessarily
reflective of views at the Federal Reserve Bank of New York or the Federal
Reserve System. Any errors or omissions are the responsibility of the authors.

Electronic copy available at: http://ssrn.com/abstract=1338125

The Term Structure of Inflation Expectations
Tobias Adrian and Hao Wu
Federal Reserve Bank of New York Staff Reports, no. 362
February 2009
JEL classification: G10, G12

Abstract

We present estimates of the term structure of inflation expectations, derived from an

affine model of real and nominal yield curves. The model features stochastic covariation
of inflation with the real pricing kernel, enabling us to extract a time-varying inflation
risk premium. We fit the model not only to yields, but also to the yields’ variance-
covariance matrix, thus increasing identification power. We find that model-implied
inflation expectations can differ substantially from break-even inflation rates when
market volatility is high. Our model’s ability to be updated weekly makes it suitable
for real-time monetary policy analysis.

Key words: affine term structure models, inflation expectations, stochastic volatility,
asset pricing, monetary policy

Adrian: Federal Reserve Bank of New York (e-mail: tobias.adrian@ny.frb.org). Wu: Princeton
University (e-mail: hzwu@princeton.edu). The authors thank Arturo Estrella, Emanuel Moench,
Simon Potter, Anthony Rodrigues, Liuren Wu, and seminar participants at the Federal Reserve
Bank of New York, the European Central Bank, and Lehman Brothers for helpful comments.
The views expressed in this paper are those of the authors and do not necessarily reflect the
position of the Federal Reserve Bank of New York or the Federal Reserve System.

Electronic copy available at: http://ssrn.com/abstract=1338125

1 Introduction

Central banks monitor financial market developments and economic data releases at
relatively high frequencies. In the U.S., for example, Federal Reserve economists brief
monetary policy makers at a weekly frequency, and the Federal Reserve’s trading desk
monitors market developments in real time. Understanding the linkages between asset
price movements and macroeconomic developments is one of the key objectives of these

monitoring efforts.
The evolution of breakeven inflation rates–the difference between nominal and
real yields at different maturities–is an especially important indicator for the conduct
of monetary policy, as breakeven inflation rates can be interpreted as measures of

inflation expectations. Since the seminal work of Kydland and Prescott (1977) and
Barro (1983), monetary economists have emphasized the importance of longer term
inflation expectations, and some studies suggest that the containment of long term
inflation expectations is the most important objective in conducting monetary policy

(see Woodford, 2003 and Bernanke et al., 2001 for summaries).

Breakeven inflation rates measure inflation expectations with a number of biases,
some of which have been previously documented (see Barr and Campbell, 1997, Elsasser
and Sack, 2004, and Gurkaynak, Sack, and Wright, 2007). First, inflation linked bench-

mark securities are typically less liquid than nominal on-the run Treasuries. Second,
the coupons of nominal and real securities with similar maturities are often different,
leading to differences in duration.
The focus of the current paper is the difference between expected inflation and
breakeven inflation due to inflation risk. We analyze the term structure of breakeven

inflation computed as the difference between a zero coupon, off-the-run nominal Trea-

Electronic copy available at: http://ssrn.com/abstract=1338125

sury yield curve and a zero coupon, real TIPS curve.1 This breakeven term structure
is sometimes called the term structure of implied inflation. Implied inflation is a better
measure of inflation expectations than breakeven inflation, as it adjusts for liquidity
differences and for differences in durations.2 However, breakeven inflation rates of
zero-coupon off-the-run curves (i.e., implied inflation) still are not pure measures of
inflation expectations. This is because the absence of arbitrage implies that the dif-
ference between zero-coupon nominal and real yields can be decomposed into three
components:

Breakeven Inflation = Expected Inflation + Inflation Risk Premium + Convexity

The literature commonly adjust for the convexity effect (see Elsasser and Sack, 2004).
However, the adjustment of breakeven inflation for the inflation risk premium requires
the estimation of a term structure model.
In this paper, we develop an affine term structure model that captures the dynamics
of real and nominal yields curves, as well as the evolution of their variance-covariance

matrix. This is important, as the inflation risk premium is proportional to the condi-
tional covariance of the real pricing kernel and inflation. In order to increase the power
for identifying the inflation risk premium, we match both the term structure of the
yield curves, and the term structure of variances and covariances.
We find a relatively small and stable inflation risk premium. The order of magnitude
of the inflation risk premium is comparable to other recent estimates in studies that
use inflation protected bonds over similar sample periods, but it is smaller and less
1
The nominal term structure is from Gurkaynak, Sack, and Wright (2006), the real term structure
is from Gurkaynak, Sack, and Wright (2007).
2
In the remainder of the paper, we use the terminology "implied inflation" and "breakeven inflation"
interchangably, as our yield curves are off-the-run zero coupon curves.

2
variable than estimates that use nominal bonds and inflation over longer time periods
(see Buraschi and Jiltsov, 2005, and Ang, Bekaert, and Wei, 2006).
An additional contribution of this paper is to present estimates of an arbitrage free
term structure model using weekly data. Much of the academic literature estimates

term structure models at lower frequencies (using monthly, quarterly, or annual data).
However, for the purpose of monetary policy decision making, it is desirable to have
estimates of inflation expectations at higher frequencies. For example, in a particular
week, payrolls could exceed expectations, the price of oil could decline, and the dollar

depreciate. How such news changes inflation expectations at different maturities has a
potentially important impact on policy decisions.
We use only real and nominal yields to estimate the term structure of inflation ex-
pectations. Our requirement to match the variance-covariance matrix of the real and
nominal yield curves provides us with enough identification power to estimate expected

inflation. In comparison, some recent work incorporates estimates of inflation expec-

tations from survey data to achieve identification (see Joyce, Lildholdt, and Sorensen,
2007, Hördahl and Tristani, 2007, and D’Amico, Kim, and Wei, 2007). We view those
papers as complementary identification strategies. The advantage of our approach is

that our model provides estimates of inflation expectations using only financial market
data. It can furthermore updated in real time.
We fit our model to an estimated variance-covariance matrix of the real and nominal
yield curves. Alternatively, options data could be used to obtain information about
second moments. For example, Goldstein and Collin-Dufresne (2002) use interest rate
caps and floors to fit an affine term structure of the nominal yield curve that features
stochastic volatility. The advantage of our approach is that our data is readily available.
To obtain an implied covariance between nominal and real yields, one would have to

3
have a time series of options on inflation swaps, which is less liquid, and not always
traded.
The remainder of the paper is organized in five sections. In Section 2, we discuss the
nominal and real yield curves and derive the relationship between breakeven inflation

and expected inflation from no arbitrage. We present our model in Section 3, and our
estimation results in Section 4. In section 5, we provide robustness checks, and Section
6 concludes.

2 Breakevens and the Inflation Risk Premium

2.1 The nominal and real yield curves

Breakeven inflation provides a crude measure of inflation expectations that can be

tracked at high frequencies. For example, in Figure 1, we plot the difference between
the 10-year on-the-run Treasury yields and the 10-year on-the-run TIPS yields in five

minute intervals for the first three weeks of March 2007. In the sample period, we can
see that breakeven inflation reacted to the release of nonfarm payrolls on March 9, and
the release of the consumer price index on March 16.3
However, using breakevens to measure inflation expectations is problematic. On-

the-run Treasuries are more liquid than benchmark TIPS. On average, the difference
between the on-the-run and the off-the-run yield is 6 basis points since the beginning
of 2003, with a daily standard deviation of 2.2% (see also Fleming, 2003, and Kr-
ishnamurthy, 2002, for analysis of the on-the-run/off-the-run Treasury yield spread).

In periods of financial market turbulence, the on-the-run/off-the-run spread tends to

3
Fleming and Remolona (1999) analyze the impact of economic news on nominal Treasury securities
in a systematic fashion, and Beechey and Wright (2008) conduct a similar analysis for the TIPS market.

4
widen. However, financial market turbulences are times when inflation expectations
potentially change, and when variances and covariances and hence risk premia tend
to change. An increase in the on-the-run/off-the-run spread during those times tends
to understate inflation expectations relative to breakeven inflation computed from on-

the-run bonds.
In addition, coupons of nominal Treasuries and TIPS with similar maturities are
often different. A 10-year breakeven spread typically has a duration that is shorter than
ten years, and the difference in duration of the nominal and the real yield introduces a

bias. The average wedge between par and zero coupon yields with a ten year maturity
is 44 basis points since the beginning of 2004.
In this paper, we use the zero-coupon, off-the run term structure of nominal Trea-
sury yields computed by Gurkaynak, Sack, and Wright (2006), and the zero coupon the
real term structure computed from TIPS real yields by Gurkaynak, Sack, and Wright

(2007). The yield data is daily and spans from January 3, 1999 to August 31, 2008. We
estimate the model at a weekly frequency, using the last day of each week to construct
our dataset, providing us with 505 weeks. We plot the yield curves of weekly real and
nominal yields for maturities 3, 4, 5, 6, 7, 8, 9, and 10 year in Figure 1, and provide

summary statistics in Table 1.

2.2 The inflation risk premium

We denote the maturity of a bond by τ . The price of a nominal bond that pays $1 at
time t + τ is Ptτ . The price of a real bond that pays $1 at time t + τ is Rτt . The Daily
Reference Level of the CPI is denoted by Qt . The absence of arbitrage implies that

there exists a discount factor Mt such that (see, for example Dybvig and Ross, 1987,

5
for an exposition of the Fundamental Theorem of Arbitrage Pricing):



Mt+τ Qt Mt+τ
Ptτ = Et , Rtτ = Et (1)
Mt Qt+τ Mt

We denote the continuously compounded yield of the nominal bond ytτ = − (1/τ ) ln (Ptτ ),
and of the inflation linked bond by rtτ = − (1/τ ) ln (Rτt ). We further denote the log-
arithmic inflation rate by π t , and the rate of change of (the negative of) the pricing

kernel by mt :

mt+1 = − (ln Mt+1 − ln Mt ) and π t+1 = ln Qt+1 − ln Qt (2)

Using these expressions (2) in equation (1) gives:

τ ytτ = − ln Et [exp (−Στs=1 mt+s − Στs=1 π t+s )] (3)

τ rtτ = − ln Et [exp (−Στs=1 mt+s )] (4)

We follow the affine term structure literature (see Piazzesi, 2003, and Singleton, 2006
for recent surveys) and assume that shocks to inflation and the real pricing kernel
are conditionally normal. We can then use the properties of the moment generating

6
function of the normal distribution to rewrite the last two equations as:

1
ytτ − rtτ = Et (Στs=1 π t+s ) (5)
   τ
  
Breakeven Expected
Inflation Inflation
1 1
+ Covt (Στs=1 πt+s , −Στs=1 mt+s ) − V art (Στs=1 π t+s )
τ   2τ  
Inflation Risk Convexity
Premium Adjustment

Breakeven inflation thus consists of three components: expected inflation, the infla-
tion risk premium, and a convexity adjustment. Intuitively, the real bond is insulated
from CPI inflation as it is indexed, so the nominal bond compensates investors for the
expected inflation until maturity. In addition, investors are compensated for the risk

that inflation varies in the future. The inflation risk premium is positive if inflation
covaries negatively with the pricing kernel M . In a consumption based asset pricing
framework, the pricing kernel is related to the growth rate of the marginal utility of
consumption. Inflation tends to be high when consumption growth is high, and the
marginal utility of consumption is low. So the negative of the pricing kernel tends to
covary positively with inflation, and we would expect the inflation risk premium to
be positive. The convexity adjustment is proportional to the conditional variance of
inflation.

7
2.3 The term structure of yield second moments

Equation (5) shows that the inflation risk premium is proportional to the covariation
between future inflation and the future real pricing kernel. In order to identify this co-

variation precisely, we match the variance-covariance matrix of nominal and real yields.
Second moments cannot be observed directly, but they can be estimated precisely when
the frequency of observed yields is high (see Merton, 1980). We use the multivariate
GARCH model proposed by Engle and Kroner (1995) to estimate the dynamics of the

variance-covariance matrix for nominal and real yields, for each maturity.
     
2 2 2
 σ̂ yt+1 σ̂ yr
t+1  ′ ′ 
(σ̂ yt ) σ̂ yr
t  ′ 
ε̂yt+1 ε̂yt+1 ε̂rt+1 
 2
 =A0 A0 +A1   A1 + A2  2
 A2
σ̂ yr
t+1 σ̂ rt+1 σ̂ yr
t (σ̂ rt )2 ε̂yt+1 ε̂rt+1 ε̂rt+1

where ε̂yt+1 is the residual of a regression of nominal yields on lagged nominal and real
yields (maturity by maturity), and ε̂rt+1 are the residuals of a regression of real yields
on lagged real and nominal yields (again maturity by maturity). Furthermore, (σ̂ yt )2 =
r 2
V art (yt+1 ), σ̂ yr
t = Cov t (yt+1 , rt+1 ), (σ̂ t ) = V ar t (rt+1 ). Because the evolution of the

variance-covariance matrix is according to a quadratic form, it is assured to be positive

definite. We provide summary statistics of the estimated variances and covariances in
Table 2.

3 Modeling Inflation Expectations

3.1 State variables and pricing kernel specification

We allow for one inflation factor π t and two real factors m1t and m2t such that mt =
m1t +m2t . We further model the stochastic evolution of the variance-covariance matrix of

8
the factors explicitly, denoting the conditional variance of πt+1 by (σ πt )2 , the conditional
covariance of π t+1 and m1t+1 by σ πm
t
1
, etc. The vector of state variables is:

′
Xt = πt m1t m2t (σ πt )2 σ πm
t
1
σ πm
t
2
(σ m1 2
t ) (σ m 2 2
t )

with dynamic evolution:

Xt+1 = µ + ΦXt + Σt ǫt+1 where ǫt+1 ∼ N (0, I8 ) (6)

vec (Σt Σ′t ) = S0 + S1 Xt (7)

We provide more detail of the specification in appendix A.

Following Duffie and Kan (1996), we model the pricing kernel allowing for flexible
prices of risk. In particular, we specify the rate of change of (the negative of) the log
pricing kernel mt from equation (2) as:

1
mt+1 = rt1 + λ′t Σ−1 ′−1 ′ −1
t Σt λt + λt Σt ǫt+1 (8)
2

where λt is the time-varying market price of risk, and rt is the one-period real short
rate, both of which are affine functions of state variables:

λt = λ0 + λ1 Xt , rt1 = δ 0 + δ 1 Xt (9)

3.2 No-arbitrage pricing restrictions

Because the evolution of state variables is conditionally Gaussian, and prices of risk
are conditionally Gaussian, nominal and real yields ytτ and rtτ are affine functions of
Xt . In particular, by replacing (6), (8), and (9) into (3), we show in appendix B that

9
yields can be expressed as:
 
 ytτ  1 τ 1 τ
  = C0 + C1 Xt (10)
rtτ τ τ

where the coefficient matrices C0τ and C1τ depend on the parameters that govern the dy-

namic evolution of the state variables. It directly follows from (10) that the conditional
variance-covariance matrix of nominal and real yields is also affine:
 
τ τ τ
 V art yt+1 Covt yt+1 , rt+1  1 τ τ 1 τ τ
vec   = 2 (C1 ⊗ C1 ) S0 + 2 (C1 ⊗ C1 ) S1 Xt
τ
Covt yt+1 τ
, rt+1 τ
V art rt+1 τ τ
(11)
The requirement for the term structure to be arbitrage free thus not only imposes
consistency of pricing across the yield curve, but also consistency of the term structure

of variance-covariance matrices across maturities.

3.3 Estimation method

We estimate the model via maximum likelihood and obtain the state variables from
the Kalman filter. The state space representation of our model is:

Yt = C0 + C1 Xt + υ t (12)

Xt+1 = µ + ΦXt + Σt ǫt+1 (13)

vec(Σt+1 Σ′t+1 ) = S0 + S1 Xt (14)

′
where Yt = yt rt V art−1 (yt ) Cov t−1 (yt , rt ) V art−1 (rt ) and C0 and C1 stack
coefficients of (10) and (11) across maturities. We treat the variance-covariance matrix

10
of nominal / real yield pairs as observable. We assume that the pricing error υ t is
normally distributed with constant, diagonal covariance matrix R. Based on state
space representation in (8), we filter the factors according to the Kalman filter:

  
X̂t = µ + ΦX̂t−1 + Kt Yt − C0 − C1 µ + ΦX̂t−1 (15)

where Kt is the Kalman gain (see Hamilton, 1994). Given estimates of the latent

factors X̂t , the parameters Θ = {µ, Φ, S0 , δ 0 , δ 1 , λ0 , λ1 } can be estimated by maximum

likelihood, based on the conditional distribution of Yt |Yt−1 for each observation. The
 
conditional distribution of Yt is N Ŷt |Yt−1 , ΩYt with Ŷt |Yt−1 = C0 + C1 X̂t−1 |Yt−1 and
ΩYt = C1 V ar (Xt |Yt−1 ) C1 + R, and we assume that the variance-covariance matrix of

the observation errors, R, is constant and diagonal. The log likelihood function is then:

T  ′ 
    
−1

L (Θ) = − log ΩYt  + Yt − Ŷt |Yt−1 ΩYt Yt − Ŷt |Yt−1 (16)
t=1

4 Estimating Inflation Expectations

4.1 Pricing

Table 3 presents the parameter estimates. We use annualized percent yields in the esti-
mation, so the long-run mean of estimated inflation is the first element of − (B1 )−1 B0 ,
which is 1.9% annual. We reject a unit root for the inflation process, but do find in-
flation to be persistent. This corresponds to the average CPI inflation over the sample

period (recall that we do not use any actual inflation data). The two factors of the
real pricing kernel m1 and m2 correspond to the level and slope factors of the real term
structure. We normalize m2 to have zero mean.

11
 We match yields and second moments well. This can be seen in Figures 3-7. In
Figures 3 and 4, we plot the actual and fitted yields of the nominal and real term
structures. In Figures 5, 6, and 7, we plot the actual and fitted (annualized) real and
nominal variances and covariances. We give the summary statistics of pricing errors in

Table 5. We find small errors for yields and the variance-covariance of yields.
In the existing literature, term structure models are usually fitted to match yields.
In this paper, we the variance-covariance matrix of yields as well. This procedure pro-
vides us with greater confidence about the accuracy of our estimated inflation risk pre-

mium. If we do not impose the constraint that second moments should be matched, a
three factor model (with two real and one nominal factor and constant second moments)
produces small pricing errors. Instead of using estimated volatility to fit the model,
some have included option price data to gauge information about second moments (see
Goldstein and Collin-Dufresne, 2002 and Bikbov and Chernov, 2005). Unfortunately,

option data on TIPS securities is not readily available.

4.2 Factors

The filtered factors of the pricing kernel are plotted in Figures 8 and 9. The real
kernel m declined in 2001 and 2002, and again in the second half of 2007. In a sim-
ple consumption based asset pricing model, m is proportional to the growth rate of

consumption. In more elaborate habit formation models, m is proportional to the

deviation of consumption growth from a moving average of consumption growth (the
habit). The latter model of the real kernel might be consistent with our estimated m,
but we do not investigate this route further (see Wachter, 2006 for a term structure

model with habit formation).

The filtered inflation factor π increases in 2003-2004, and declines from an average

12
2.4% in 2005 to an average of 1.8% since the beginning of 2007. In the period from
1999-2001, the inflation factor π is rather low–this is likely due to the low liquidity of
the TIPS market in the first three years (see subsection 5.2 for further discussion).
The variance of the real pricing kernel σ 2m is higher than the variance of inflation
σ 2π , as can be seen in Figure 9. Real volatility was particularly high in 2001-2003, and
then again towards the end of 2007, corresponding to periods of low m. Real volatility
is thus high when m (and hence real interest rates) are low. Inflation variance is
particularly high in the early part of the sample (1999-2002 with an average of 1%),

and is only 30 basis points since the beginning of 2004. The higher volatility of the
inflation factor again likely reflects some of the illiquidity of the TIPS market in the
early sample. Average covariance of the real pricing kernel and inflation is negative,
giving rise to a positive inflation risk premium (Equation 5).

4.3 Expected inflation and forward inflation

Expected inflation can be computed from the parameters of (6):

1  
Et (Στs=1 π t+s ) = [1 0 0...] C̃0τ + C̃1τ Xt (17)
τ
 
where C̃0τ = C̃0τ −1 + I8 + C̃1τ −1 µ, C̃1τ = I8 + C1τ −1 Φ, C̃00 = C̃10 = 0.

We plot the 5-10 year breakeven forward rates together with the expected inflation
forward rates in Figure 10, and the 4-5 and 9-10 year breakeven and expected inflation
forward rates in Figure 11. Expected forward inflation is less volatile than breakeven
forward inflation, especially for longer maturities.
Differences of breakeven and expected inflation are more pronounced for forward

rates. This can be seen in Figure 11, where we plot the 4-5 year, 5-10 year, and 9-

13
10 year forward inflation rates for breakevens and expected inflation. The difference
between 5-10 year breakeven and 5-10 year inflation was 31 basis points in the second
half of 2007, and 27 basis points since the beginning of 2003. The difference between
expected inflation and breakeven inflation is less pronounced for shorter maturities

such as the 4-5 year forward, as risk premia are smaller at shorter horizons.
We show a comparison of our estimated π to current core and total CPI inflation
in Figure 12. Since 2003, it appears that the inflation that we extract from the term
structure is closer to the core CPI than to total CPI. This reflects the fact that the

difference between core and total CPI is transitory, so it does not affect expected
inflation at the maturities above three years which we use in our analysis.
The adjustment to breakeven inflation (i.e. the difference between breakeven infla-
tion and expected inflation) correlates highly with market based measures of implied
volatility. This can be seen in Figures 14 and 15. In Figure 14, we plot the model

implied adjustment to breakeven inflation together with the Merrill Lynch Option Im-
plied Volatility Index (MOVE). The move index is a simple average of the Treasury
implied volatility of 2-year, 5-year, 10-year, and 15-30-year exchange traded options.
We can see that the breakeven adjustment is high when Treasury implied volatility is

high. In Figure 15, we plot the model implied adjustment together with the S&P 500
implied volatility (VIX) computed by the Chicago Board Options Exchange (CBOE).
We again find a high correlation between the adjustment, and implied volatility.

14
5 Robustness

5.1 Testing for the number of factors

Our baseline specification has eight factors: two factors of the real pricing kernel that
capture the level and slope of the real term structure, one inflation factor to model

inflation expectations, and the variances and covariances of the real and nominal vari-
ables. Compared to most results in the literature, this is a relatively large number of
factors. However, the variances and covariances are pinned down by our requirement
that the model fit the variance-covariance matrix of nominal and real yields across

maturities.
We have estimated a number of alternative specifications. In Table 6, we report
a test against an alternative model with five factors: one real factor, inflation, and
the variances of the real kernel and inflation, as well as their covariance. We can see

that the five factor model is rejected against the eight factor alternative at the 1%
level. This result arises as one factor for the real term structure is not sufficient to
model the dynamics of level and slope. Figure 2 shows that the real term structure
has noticeable movements in both level and slope, and that those movements are not
perfectly correlated. Thus yield pricing errors increase substantially when only one real

factor is used.
We also estimate alternative models with ten factors: two real factors, two inflation
factors, and their variances and covariances (not reported). We do not find that the
pricing performance improves significantly in the ten factor model, and thus use the

eight factor model in our baseline specification in order to preserve parsimony. We have
also estimate specifications without time varying variances and covariances. We achieve
a similar fit in terms of yield errors, but do not fit second moments, per construction.

15
5.2 Structural break tests

Elsasser and Sack (2004) point out that the TIPS market was relatively illiquid for
a number of years. The liquidity in the TIPS market biases real rates upwards, thus

artificially compressing breakeven inflation rates. In order to see how the illiquidity
might change our estimates, we fit the model separately before 2002 and since the
beginning of 2002 (thus splitting the sample into a three year and a six year period).
We use 2002 as a break point as Elsasser and Sack argue that liquidity in the inflation

protected market was comparable to the liquidity of the off-the-run Treasury market
since then.
We do find a structural break at the beginning of 2002, as reported in Table 7.
However, the changes in parameters in the two separate sample periods is small. Im-

portantly, the filtered factors do not change substantially.

Our reason for using the 1999-2008 sample period in the baseline estimates–and
not the shorter 2002-2008 sample–is that the longer period captures a whole business
cycle. Recall that the U.S. economy was in a recession at the beginning of 2001. If

we start estimation in 2002, we exclude the recession from the sample, and pick up a
trend in the slope of the real curve (see Figure 2).

5.3 Inflation as observable

The indexation of TIPS to the Consumption Price Index (CPI) introduces a predictable
component in yield changes. This predictable component is sometimes called "carry",
and a carry adjustment of yields is undoing this predictability. The interest and prin-

cipal payments for TIPS are linked to the non-seasonally adjusted urban CPI (CPI-
UNSA) with a three-month lag. The CPI is published every month. The daily reference

16
index (DRI) for TIPS payoff and pricing calculation is computed based on the CPIU-
NAS values with two- and three-month lags (M2 and M3) as,

Daily Reference Index =Three-Month CPI Lag

+ (Number(Today−1)
of Days in Month)
(Two-Month CPI Lag − Three-Month CPI Lag).

TIPS principal is adjusted by multiplying the principal at issuance by the DRI at

maturity and then dividing it by the DRI at issuance date. The adjusted principal is
paid at maturity. The principal payment at maturity is

Daily Reference Index

$ Par Value× Daily Reference Indexm a tu rity d a te
issu a n ce d a te

where the ratio of the two DRIs is often referred to as the index ratio.
We estimate specifications where the DRI is included in the observation equations

(estimation results of such specifications is not reported here). We adjust for the carry
effect by modeling the indexation lag explicitly. We do not find substantial differences
in our estimates of the term structure of expected inflation, so we omit it from the
current paper.

6 Conclusion

We propose a novel methodology to extract the term structure of inflation expectations

from the term structures of nominal and real interest rates. Our contribution is to fit
an arbitrage free affine model not only to yields, but also their conditional variance-
covariance matrix. We find that an eight factor model with two real factors, one
inflation factor, and five variance-covariance factors fits both first and second moments
of the term structures well.

17
 Our model can be updated daily, making it suitable for market monitoring. We
do find that there can be substantial differences between model implied inflation ex-
pectations, and breakeven inflation rates. These differences are highly correlated with
market volatility measures such as the MOVE Treasury implied volatility index, or

the VIX equity implied volatility index. Intuitively, as implied volatility increases, risk
premia increase, and breakevens tend to overpredict inflation expectations.

18
A Compact state-space form of 8-factor model

Let Zt = [π t m1t m2t ]′ . The transition equation for Zt follows a VAR(1) with a condi-
tional variance-covariance matrix Ωt :


Zt+1 − Zt = B0 + B1 Zt + Ωt εt+1 with εt ∼ N (0, I3 ) (18)

We model the evolution of the variance-covariance matrix Ωt of εt as a multivariate

stochastic process:

vec (Ωt+1 ) = B̃0′ ⊗B̃0′ +B̃1′ ⊗B̃1′ vec (Ωt )+S̃0′ ⊗S̃0′ vec (ε̃t ) with p·vec (ε̃t ) ∼ N (0, I5 ) (19)

where p and pinv are selection matrices:

 
 1 0 0 0 0 
 
 0 1 0 0 0 
 
   
 
1 0 0 0 0 0 0 0 0  0 0 1 0 0 
   
   
 0 1 0 0 0 0 0 0 0   0 1 0 0 0 
   
   
p=
 0 0 1 0 0 0 0 0 0 
 pinv = 
 0 0 0 1 0 

   
   
 0 0 0 0 1 0 0 0 0   0 0 0 0 0 
   
 
0 0 0 0 0 0 0 0 1  0 0 1 0 0 
 
 
 0 0 0 0 0 
 
 
0 0 0 0 1

Note that p · pinv = I5 .

19
 To obtain the compact state space (6), we stack equations (18) and (19):

 √ 
Ωt 0
   
Xt = [Zt p · vec (Ωt )]′ Σt =   
0 p S̃0′ ⊗ S̃0′ pinv

   
 B0   B1 + I3 0 
µ=    Φ=   
B̃0′ ⊗ B̃0′ vec (I3 ) 0 p B̃1′ ⊗ B̃1′ pinv

 
 εt 
and ǫt =  . We denote the elements of Ωt by:
ε̃t

 
(σ πt )2
1 2

 σ πm
t σ πm
t 
  2 
Ωt = 
 σt
πm1 m1
σt 0 
 (20)
  2 
2 2
σ πm
t 0 σm
t

 2
where (σ πt )2 is the conditional variance of π t+1 , σ m
1
t the conditional variance of

′
1 2 2 2
mt+1 , etc, so p · vec (Ωt ) = (σ t ) σ t
π πm 1 πm
σt 2 m
(σ t ) (σ t )
1 m 2 .

20
B No-arbitrage restrictions

We make the following guess for yields:

ytτ = τ1 C0y
τ
+ τ1 C1y
τ
Xt
(21)
rtτ = τ1 C0r
τ
+ τ1 C1r
τ
Xt

For real yields, we have:

1  
rtτ = − ln Et exp −mt+1 − (τ − 1) rt+1
τ −1
(22)
τ

Replacing the guess for the yield function:


τ τ 1 1 ′ −1 ′−1 ′ −1 τ −1 τ −1
C0r + C1r Xt = − ln Et exp −rt − λt Σt Σt λt − λt Σtǫt+1 − C0r − C1r Xt+1
2

Using the properties of the moment generating function of the normal distribution and
collecting terms gives:

τ τ τ −1 τ −1 1 τ −1 τ −1
C0r + C1r Xt = δ 0 + C0r + C1r (µ − λ0 ) − C1r ⊗ C1r S0

2 
1 τ −1 τ −1 τ −1
+ − C1r ⊗ C1r S1 + C1r (Φ − λ1 ) + δ 1 Xt
2

Matching terms gives:

τ τ −1 τ −1 1 τ −1 τ −1
C0r = C0r + C1r (µ − λ0 ) − C1r ⊗ C1r S0 + δ 0 (23)
2
τ τ −1 1 τ −1 τ −1
C1r = C1r (Φ − λ1 ) − C1r ⊗ C1r S1 + δ 1 (24)
2

21
For nominal yields, we have:

1
τ ytτ = − ln Et [exp (−Στs=1 mt+s − Στs=1 π t+s )] (25)
τ

Or:
τ τ
 τ −1 τ −1

C0y + C1y Xt = − ln Et exp −mt+1 − φXt+1 − C0y − C1y Xt+1

where φ = [1 0 0 ...]. Then we find:

τ τ τ −1 τ −1 1 τ −1 τ −1
C0y + C1y Xt = δ 0 + C0y + C1y + φ (µ − λ0 ) − C1y + φ ⊗ C1y + φ S0

2 
τ −1 1 τ −1 τ −1
+ C1y + φ (Φ − λ1 ) − C + φ ⊗ C1y + φ S1 + δ 1 Xt
2 1y

Matching coefficients, and we find:

τ τ −1 τ −1 1 τ −1 τ −1
C0y = C0y + C1y + φ (µ − λ0 ) − C + φ ⊗ C1y + φ S0 + δ 0 (26)
2 1y
τ τ −1 1 τ −1 τ −1
C1y = C1y + φ (Φ − λ1 ) − C + φ ⊗ C1y + φ S1 + δ 1 (27)
2 1y

Note that if the yields are rescaled by a factor γ, say they are expressed as percent

per annum so that γ = 1200, the quadratic term above must be multiplied by γ −1 . For
notational convenience we stack:

C0τ = [ C0y
τ τ ]′
C0r

C1τ = [ C1y
τ τ ]′
C1r

22
C References

Ang, Andrew, Geert Bekaert, Min Wei, 2007, "The Term Structure of Real Rates and
Expected Inflation", National Bureau of Economic Research Working Paper 12930.
Bernanke, Ben S., Thomas Laubach, Frederic S. Mishkin, and Adam S. Posen, 2001,
Inflation Targeting: Lessons from the International Experience, Princeton University
Press.

Barr, David G., and John Y. Campbell, 1997, "Inflation, Real Interest Rates,
and the Bond Market: A Study of UK Nominal and Index-Linked Government Bond
Prices", Journal of Monetary Economics 39, pp. 361—383.
Barro, Robert J., 1983, "Inflationary Finance under Discretion and Rules", Cana-

dian Journal of Economics 16, Issue 1, pp.1-17.

Beechey, Meredith, and Jonathan Wright, 2008, "The High-Frequency Impact of
News on Long-Term Yields and Forward Rates: Is It Real?," unpublished manuscript,
Johns Hopkins University.

Bibkov, Ruslan, and Mikhail Chernov, 2005, "Term Structure and Volatility: Lessons
from the Eurodollar Markets," Columbia University Working Paper.
Buraschi, Andrea, and Alexie Jiltsov, 2005 "Inflation Risk Premia and The Expec-
tations Hypothesis", Journal of Financial Economics 75, pp. 429 - 490.

Campbell, John Y., Adi Sunderam, and Luis Viceira, 2007, "Inflation bets of De-
flation Hedges? The Changing Risks of Nominal Bonds", Harvard University Working
Paper.
Chen, Ren-Raw, Bo Liu, and Xiaolin Cheng, 2005, "Inflation, Fisher Equation, and
the Term Structure of Inflation Risk Premia: Theory and Evidence from TIPS," SSRN

Working Paper 809346.

23
 Collin-Dufresne, Pierre, and Robert S. Goldstein, 2001, "Do Bonds Span Fixed
Income Markets? Theory and Evidence for ‘Unspanned’ Stochastic Volatility," Journal
of Finance, pp. 1685-1730.
D’Amico, Stefania, Don Kim, and Min Wei, 2006, "Tips from TIPS: The Informa-

tional Content of Treasury Inflation-Protected Security Prices," SSRN Working Paper

960869.
Dai, Qiang, and Kenneth Singleton, 2000, "Specification Analysis of Affine Term
Structure Models," Journal of Finance 55, pp. 1943-1978.

Dai, Qiang, and Kenneth Singleton, 2002, "Expectations Puzzle, Time-Varying Risk
Premia, and Affine Models of the Term Structure," Journal of Financial Economics
63, pp. 415-441.
Dai, Qiang, and Kenneth Singleton, 2003, "Term Structure Dynamics in Theory
and Reality", Review of Financial Studies 16, pp. 631-678.

Duffee, Gregory R., 2002, "Term Premia and Interest Rate Forecasts in Affine
Models," Journal of Finance 57, pp. 405-443.
Duffie, Darrell and Rui Kan, 1996, "A Yield-factor Model of Interest Rates", Math-
ematical Finance 6, pp. 379- 406,

Durham, J. Benson, 2006, "An Estimate of the Inflation Risk Premium using a
Three-Factor Affine Term Structure Model", Board of Governors of the Federal Reserve
Finance and Economics Discussion Series 2006-42.
Dybvig, Philip H., and Stephen A. Ross, 1987, “Arbitrage,” The New Palgrave: a
Dictionary of Economics, pp. 100-106.
Elsasser, Robert, and Brian P. Sack, 2004, "Treasury Inflation-Indexed Debt: A
Review of the U.S. Experience," Economic Policy Review 10, pp. 47-63.
Engle, Robert, and Kenneth Kroner, 1995, "Multivariate Simultaneous GARCH,"

24
Econometric Theory 11, pp. 122-150.
Evans, Martin, 2003, "Real Risk, Inflation Risk, and the Term Structure," Eco-
nomic Journal 113, pp. 345-389.
Fleming, Michael J., 2003, "Measuring Treasury Market Liquidity," Federal Reserve

Bank of New York Economic Policy Review 9, pp. 83-108.

Fleming, Michael J., and Eli M. Remolona, 1999, "Price Formation and Liquidity in
the U.S. Treasury Market: The Response to Public Information," Journal of Finance
54, pp. 1901-15.

Gurkaynak, Refet S., Brian Sack, and Jonathan H. Wright, 2006, "The U.S. Trea-
sury Yield Curve: 1961 to the Present," Board of Governors of the Federal Reserve
Finance and Economics Discussion Series 2006-28.
Gurkaynak, Refet S., Brian Sack, and Jonathan H. Wright, 2007, "The TIPS Yield
Curve and Inflation Compensation," Board of Governors of the Federal Reserve Finance

and Economics Discussion Series 2007.

Gong, Frank F. and Eli M. Remolona 1997, “A Three-factor Econometric Model of
the US Term Structure,” Federal Reserve Bank of New York Staff Reports 19.
Hamilton, James D., 1994, Time Series Analysis, Princeton University Press.

Hördahl, Peter, and Oreste Tristani, 2007, "Inflation Risk Premia in the Term
Structure of Interest Rates," European Central Bank Working Paper 734.
Joyce, Mike, Peter Lildholdt, and Steven Sorensen, 2007, "Extracting Inflation
Expectations and Inflation Risk Premia from the Term Structure: a Joint Model of
the UK Nominal and Real Yield Curves," Bank of England Working Paper.
Kim, Don, 2004, “Inflation and the Real Term Structure,” Stanford University
Dissertation.
Krishnamurthy, Arvind, 2002, "The Bond/Old-Bond Spread," Journal of Financial

25
Economics 66, pp. 463-506.
Kydland, Finn, and Edward C Prescott, 1977, "Rules Rather Than Discretion: The
Inconsistency of Optimal Plans," Journal of Political Economy 85(3), pp. 473-91
Merton, Robert C., 1980, "On Estimating the Expected Return on the Market: An

Exploratory Investigation," Journal of Financial Economics 8, pp. 323-61.

Nelson, Daniel B., 1990, "ARCH Models as Diffusion Approximations," Journal of
Econometrics 45, pp. 7-38.
Piazzesi, Monika, 2003, "Affine Term Structure Models," Handbook of Financial

Econometrics.
Singleton, Kenneth J., 2006, Empirical Dynamic Asset Pricing, Princeton Univer-
sity Press.
Wachter, Jessica A., 2006, "A Consumption-Based Model of the Term Structure of
Interest Rates," Journal of Financial Economics 79, pp. 365-399.

Woodford, Michael, 2003, Interest and Prices: Foundations of a Theory of Monetary

Policy, Princeton University Press.

26
 Nominal Yields Real Yields
Maturity Mean Median Std Max Min Mean Median Std Max Min
3 3.97 4.02 1.30 6.74 1.39 1.94 1.94 1.27 4.23 -0.47
4 4.15 4.24 1.16 6.70 1.74 2.15 2.11 1.18 4.34 -0.16
5 4.31 4.38 1.04 6.69 2.08 2.30 2.20 1.10 4.37 0.12
6 4.47 4.45 0.95 6.71 2.40 2.43 2.25 1.04 4.38 0.38
7 4.61 4.53 0.87 6.72 2.70 2.50 2.28 0.98 4.37 0.61
8 4.74 4.61 0.81 6.77 2.96 2.56 2.32 0.94 4.36 0.81
9 4.86 4.69 0.76 6.81 3.20 2.62 2.35 0.90 4.36 0.99
10 4.97 4.77 0.72 6.83 3.41 2.66 2.37 0.86 4.35 1.14

Table 1: Summary Statistics for Yields.

The table reports summary statistics for nominal, zero coupon, off-the-run
Treasury yields and real, zero coupon TIPS yields for maturities 3, 4, 5, 6, 7,
8, 9, and 10 years. The data is weekly from 1/11/1999-8/31/2008. Source:
Board of Governors of the Federal Reserve. The nominal term structure is

from Gurkaynak, Sack, and Wright (2006), the real term structure is from
Gurkaynak, Sack, and Wright (2007).

27
 Nominal Yield Variance (×102 ) Real Yield Variance (×102 )
Maturity Mean Median Std Max Min Mean Median Std Max Min
3 2.02 1.91 0.82 4.98 0.94 1.96 1.54 1.53 8.63 0.31
4 1.98 1.88 0.76 4.70 0.92 1.39 1.15 0.78 4.42 0.30
5 1.92 1.86 0.72 4.48 0.89 1.19 0.99 0.65 3.69 0.29
6 1.83 1.78 0.69 4.24 0.83 1.08 0.90 0.61 3.24 0.27
7 1.74 1.68 0.61 3.81 0.83 0.98 0.83 0.58 2.95 0.25
8 1.66 1.63 0.59 3.49 0.80 0.91 0.76 0.55 2.75 0.23
9 1.59 1.56 0.54 3.22 0.81 0.83 0.69 0.50 2.49 0.22
10 1.53 1.52 0.51 3.10 0.79 0.78 0.66 0.47 2.43 0.20
Yield Covarianc (×102 )
Maturity Mean Median Std Max Min
3 1.31 1.12 0.77 4.18 0.33
4 1.24 1.02 0.66 3.64 0.34
5 1.17 0.93 0.61 3.26 0.34
6 1.09 0.86 0.57 2.86 0.38
7 1.01 0.81 0.52 2.57 0.34
8 0.95 0.78 0.50 2.55 0.31
9 0.88 0.73 0.46 2.40 0.28
10 0.84 0.70 0.43 2.27 0.27

Table 2: Summary Statistics for Variances and Covariances of Yields.

28
 Coefficient Estimate Std. Err.
   
0.0077 0.0221
B0 =  0   . 
0 .

   
−0.0010 0.0003 −0.0072 0.0085 0.0036 0.0100
B1 =  0.0031 −0.0002 0   0.0026 0.0019 . 
0.0036 0.0006 −0.0116 0.0119 0.0047 0.0144
   
0.5512 0 0 0.2036 . .
B̃1 =  0 0.6760 0   . 0.13436 . 
0 0 0.4945 . . 0.3876
   
0.06161 0 0 1.0081 . .
S̃0 =  0 −0.0037 0   . 3.100 . 
0 0 0.0049 . . 1.160
   
0.0807 0.0988
 1.2922   0.0801 
δ1 =    
 0.9952   0.1784 
05 .

Table 3: Parameter Estimates.

29
 λ0 = 0.0084 −0.0039 −0.0015 0.0021 0.0015 −0.0051 −0.0003 0.0025

Std.Err. = 0.0687 0.1482 0.1614 1.2515 6.4519 3.6793 9.3959 1.6002

 
−0.0058 −0.0029 0.0048 01×5
 −0.0031 −0.0020 0.0144 01×5 
λ1 =  
 −0.0030 −0.0028 0.0028 01×5 
05×1 05×1 05×1 05×5
. 
0.00844 0.0035 0.0103 .
 0.00257 0.0019 0.0022 . 
Std.Err. =  
 0.0118 0.00480 0.0143 . 
. . . .

Measurement Error
Ryield = 0.00046 1.3. × 10−5
Rvol = 1.9 × 10−6 8.5 × 10−7

Table 4: Parameter Estimates (Continued).

30
 Nominal Yield Real Yield
Maturity Mean Median Std Mean Median Std
3 0.0068 0.0129 0.0990 -0.0339 -0.0344 0.1342
4 -0.0202 -0.0176 0.0691 -0.0069 -0.0068 0.0689
5 -0.0176 -0.0166 0.0476 0.00078 -0.0049 0.0652
6 -0.0031 -0.0004 0.0351 0.0023 -0.0037 0.0590
7 0.0113 0.0090 0.0343 0.0016 0.0015 0.0518
8 0.0185 0.0171 0.0429 0.0002 0.0013 0.0503
9 0.0156 0.0127 0.0565 -0.0017 0.0015 0.0554
10 0.0015 -0.004 0.0727 -0.0046 0.0004 0.0649

Nominal Variance(×104 ) Real Variance(×104 )

Maturity Mean Median Std Mean Median Std
3 11.11959 10.3784 5.3897 49.5037 22.6688 71.1979
4 1.6215 2.4298 8.7564 -4.5012 0.6644 17.8840
5 -2.5428 -0.9189 8.4537 -12.9125 -3.7495 24.6054
6 -5.1399 -5.0284 5.6064 -12.5788 -4.7463 21.2169
7 -5.5393 -4.4742 3.3265 -9.7451 -4.4325 15.8819
8 -2.6838 -2.8632 3.5931 -6.2577 -2.5856 12.2774
9 2.1412 1.7740 6.5962 -3.6854 -1.6542 9.3206
10 8.4337 6.4254 9.5207 0.6464 1.9233 7.5621

Covariance(×104 )
Maturity Mean Median Std
3 -0.9134 -1.9248 16.4993
4 -5.8432 -4.1209 5.6884
5 -6.0529 -3.4845 7.4071
6 -4.8311 -3.7114 5.0572
7 -1.6575 -1.9004 2.8968
8 2.6626 1.1844 5.5238
9 6.4907 3.9956 8.9856
10 11.6731 7.7593 11.7283

Table 5: Pricing Errors.

31
 Maximum Likelihood Number of Number of
Model Value (Mean) Parameters Observations
5-factor -290.66 23 505
8-factor 111.22 37 505

Tests Chi Square Significance Degrees of Freedom

8-factor / 5-factor 405898.80 *** 14

Table 6: Testing for the Number of Factors.

32
 Maximum Likelihood Number of Number of
Model Value (Mean) Parameters Observations
8-factor Model 111.22 37 505
8-factor Model Before 2002 105.17 37 157
8-factor Model After 2002 118.06 37 348

Tests Chi Square Significance Degrees of Freedom

Structural Break in 2002 2860.33 *** 37

Table 7: Testing for a Structural Break.

33
 Figure 1: Figure 1: On-the-Run 10-Year Breakeven Inflation.

The figure plots the difference between the on-the-run 10-year Treasury and
TIPS yields in five minute intervals for March 1, 2007 - March 24, 2007.
Source: Bloomberg.

34
7 7
Nominal Yields:
Real Yields:
6 6

5 5

4 4

3 3

2 2

1 1

0 0

-1 -1
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 2: Nominal and Real Yields

The figure plots nominal, zero coupon, off-the-run Treasury yields and real,

zero coupon TIPS yields for maturities 3, 4, 5, 6, 7, 8, 9, and 10 years. The

data is weekly from 1/11/1999-8/31/2008. Source: Board of Governors of
the Federal Reserve. The nominal term structure is from Gurkaynak, Sack,
and Wright (2006), the real term structure is from Gurkaynak, Sack, and

Wright (2007).

35
8 8
Nominal Yields:
Fitted Yields:
7 7

6 6

5 5

4 4

3 3

2 2

1 1

0 0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 2: Figure 3: Actual and Fitted Nominal Yields.

The figure plots the nominal yields from Figure 2 together with the yields
predicted by the term structure model.

36
5 5
Real Yields:
Fitted Yields:
4 4

3 3

2 2

1 1

0 0

-1 -1
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 4: Actual and Fitted Real Yields.

The figure plots the real yields from Figure 2 together with the yields pre-
dicted by the term structure model.

37
2.5 2.5
Nominal Variances:
Fitted Variances:

2 2

1.5 1.5

1 1

0.5 0.5

0 0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 5: Actual and Fitted Conditional Variances of Nominal Yields.

38
2.5 2.5
Yield Covariances:
Fitted Covariances:

2 2

1.5 1.5

1 1

0.5 0.5

0 0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 6: Actual and Fitted Conditional Variances of Real Yields.

39
 2.5 2.5
Yield Covariances:
Fitted Covariances:

2 2

1.5 1.5

1 1

0.5 0.5

0 0
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 7: Actual and Fitted Conditional Covariance of Nominal and Real Yields.

40
4 4
pi m
3 3

2 2

1 1

0 0

-1 -1

-2 -2

-3 -3
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 8: Estimated Inflation Factor π and Real Pricing Kernel m.

41
1.9 1.9
sigma_pi2 sigma_m2 sigma_mpi

1.4 1.4

0.9 0.9

0.4 0.4

-0.1 -0.1

-0.6 -0.6

-1.1 -1.1

-1.6 -1.6
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

Figure 9: Pricing Factor Variances and Covariances (Annualized).

42
 Forward Breakeven Inflation 5-10 Year Forward Expected Inflation 5-10 Year
Unexplained Forward Inflation Risk Premium 5-10 Year
4 4

3 3

2 2

1 1

0 0

-1 -1
2004 2005 2006 2007 2008

Figure 10: Forward Breakeven 5-10 Year and Forward Expected Inflation 5-10 Year.

43
 Forward Breakeven Inflation 4-5 Year Forward Expected Inflation 4-5 Year

Forward Expected Inflation 9-10 Year Forward Breakeven Inflation 9-10 Year
4 4

3.5 3.5

3 3

2.5 2.5

2 2

1.5 1.5
2004 2005 2006 2007 2008

Figure 11: Forward Breakeven and Forward Expected Inflation 4-5 Year, 9-10 Year.

44
6 6
CPI Annual % Change

Core CPI Annual % Change

5 Current Inflation Estimated From 5

Yield Curves (End of month)

4 4

3 3

2 2

1 1
2004 2005 2006 2007 2008

Figure 12: Comparison to Current Inflation.

45
190 50
MOVE (Left Axis)

170 Annualized Forward Inflation Risk Premium 5-10 Year

40

150

30
130

110
20

90

10
70

50 0
2004 2005 2006 2007 2008

Figure 13: Breakeven Adjustment and MOVE Implied Volatility.

46
35 40
VIX (Left Axis)
Annualized Forward Inflation Risk Premium 5-10 Year
30

30

25

20 20

15

10

10

5 0
2004 2005 2006 2007 2008

Figure 14: Breakeven Adjustment and VIX Implied Volatility.

47