What Does the Yield-Curve Slope Really Tell Us?

Michael J. Howell

JFI 2018, 27 (4) 22-33

doi: https://doi.org/10.3905/jfi.2018.1.059
http://jfi.iijournals.com/content/27/4/22
This information is current as of August 4, 2018.

Email Alerts Receive free email-alerts when new articles cite this article. Sign up at:
http://jfi.iijournals.com/alerts

Institutional Investor Journals

1120 Avenue of the Americas, 6th floor,
New York, NY 10036, Phone: +1 212-224-3589
© 2017 Institutional Investor LLC. All Rights Reserved
Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018
 What Does the Yield-Curve
Slope Really Tell Us?
Michael J. Howell

T
M ichael J. Howell he f inance literature acknowl- the spread between the 10-year less 2-year
is a managing director edges that the slope of the yield note, largely because the latter is liquid and
at CrossBorder Capital
cur ve, or term structure of more ref lective of market expectations than
Ltd. in London, U.K.
[email protected] interest rates, contains valuable the 3-month T-bill yield. In this article, we
information about the future path of the challenge the robustness of the single generic
economy (Estrella and Hardouvelis [1991], long/short yield-curve spread as a business
Mishkin [1990]). Traditionally, a relatively cycle predictor. Adrian et al. [2014] demon-
f lat or inverted yield curve warns of a busi- strate that the size of these maturity spreads
ness recession some 12 to 15 months in the largely ref lect term premia. These, in turn,
future. The literature also concludes that it may be inf luenced by different types of inves-
is the size of the yield-curve slope that is tors with different preferred habitats by tenor.
important and not its change, nor whether We argue that the observed transmission
the source of change in slope derives from mechanism between the financial and real
either the long- or short-end of the term economies likely depends upon movements
structure. However, what is unclear is exactly in the distribution of these term premia and
how this transmission operates, and whether conclude that more yield-curve parameters
certain maturity combinations work better are required to fully capture the information
than others. implicit in their pattern.
The yield curve expresses the spot
USING THE YIELD-CURVE yield on a default-free sovereign govern-
SLOPE FOR BUSINESS CYCLE ment bond across a cross-section of horizons
FORECASTING that describe the notional maturity date of
each bond.1 The spot yield is the redemption
The 10-year note less 3-month T-bill yield 2 of a zero-coupon bond. It comprises
yield is used frequently as the benchmark the product of one-period forward rates,
spread. This is justif ied by researchers that is, the discount rate of a single cash
because of its supposed robustness over time, f low from a zero-coupon bond equivalent,
from evidence that the various maturity over a fixed holding period. Each spot yield
combinations are high ly cor related. ( yt m ) comprises an expected real interest rate
However, some U.S. researchers take shorter- (Rt ) plus an expected inf lation rate (πt ) over
term money market rates instead, while Fama a holding period (m) plus a nominal term (or
[1986] uses the 5-year less 1-year note spread. bond maturity risk3) premium (tpt m ). Under
Equally, many practitioners tend to prefer the expectations hypothesis (EH), the nominal

22    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

term premia are constant for all horizons (m). In the case parameter, such as the 10-year/1-year spread. The slope
of the pure expectations hypothesis (PEH), they are zero of the term structure at any tenor ref lects a combina-
across all horizons (m). According to efficient markets tion of short-term rate expectations and term premia,
theory, these risk premia should be negligible. However, with the latter becoming increasingly more important
econometric tests reject this hypothesis (see Campbell as maturity extends.5 Although various maturity com-
[1995]). The expectations theory of interest rates can be binations are highly correlated, the data show that their
described from the following equation, after making the relative importance changes over time, which may sug-
appropriate term premia assumptions: gest that time-varying preferred habitats are important.
Curvature plays a key role in our argument because
1 m −1 1 m in the extreme case of a perfectly linear yield curve,
ytm = ∑ t t +i m ∑ Et πt +i + tptm
m i =0
E R +
the slopes of all maturity combinations are the same.
i =1
It follows that their changing relative importance must
where yt m is the spot yield of a bond of maturity m at result from f luctuations in the degree of curvature over
time t; Et denotes the expectations operator; Rt is the time. We suggest that the changing pattern of bond term
real interest rate; πt is the inf lation rate; tpt m represents the premia largely explain these f luctuations. It is known
nominal bond term premium over a holding period m. from the literature that term premia play an important
The usefulness of government bond data does not role in the transmission of monetary policy (Borio and
occur because fixed-income investors are any more pre- Zhu [2012]; and Bruno and Shin [2012]). Thinking
scient or better informed than others, but because it is of these maturity combinations as largely measures of
straightforward to extract forward-looking information term premia suggests that describing the pattern of
from the Treasury yield curve. Their signals also may term premia across the term structure requires other
be purer because government bonds are less distorted parameters, such as the size and position of the curvature
by other factors, such as illiquidity and default risk, hump along the maturity axis. This may also explain
whereas private sector bonds are popularly associated why Moench [2012] favors the curvature of the term
with defaults.4 In practice, the choice of bond maturity structure as a predictor of the business cycle. Curvature
is a derived demand because fixed-income investors tend can be thought of as an “average” measure of the size of
to target duration (see Leibowitz, Bova, and Kogelman these different maturity spreads.
[2014]), which is often thought of as the effective matu- Exhibit 1 demonstrates our main result. More detail
rity. In contrast to maturity, which only considers the is given in Appendix A. This chart compares the outputs
final payment from an investment, duration gives weight from three probit models, with parameters estimated by
to all cash f lows received (paid out), for example, cou- maximum likelihood, that predict the probability of U.S.
pons, over the entire life of the asset (liability), taking recession (NBER defined) 12 months ahead. The first
into account both size and frequency of payment. For model uses the U.S. Treasury 10-year less 1-year spread
zero-coupon bonds, duration and maturity are the same. (y10 – y1); the second takes the 1-5-10 year “butterf ly”
On average over the post-World War II period, spread measure of curvature, or the yield premium of the
the yield curve has typically been upward-sloping and U.S. Treasury 5-year tenor over the average of the one
concave to the maturity axis (see Gurkaynak, Sack, and and the 10-year tenors, and the third model measures of
Wright [2006]). The term structure is often summarized the concentration of implied term premia by maturity,
using just three parameters—the level, slope, and curva- based on the position of the curvature hump in the yield
ture. Cochrane and Piazzesi [2005] show that the first curve along the maturity axis (D-star). This parameter,
three principal components explain more than 99% of which is the maturity counterpart to the size measure of
the variation across the term structure, and following curvature, is described more fully in Appendix B. The
Litterman and Scheinkman [1991], these components RMSE6 of the D-star model (4.53) is lower than that for
are interpreted as the eponymous yield-curve param- the yield-curve slope (4.68) and curvature models (4.96).
eters because of the pattern of their loadings. We argue Over the 1985–2016 period, there were 34 recessionary
in this report that the different maturity combina- months. Using a probability threshold of 50%, the yield-
tions contain valuable information about the future curve model correctly spotted six of these occurrences
macro-economy that is not captured by a single slope 12 months ahead of time; the curvature model, 4; and

Spring 2018 The Journal of Fixed Income    23

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

Exhibit 1
Probability of U.S. Recession (NBER definition) in Next 12 Months Using Three Probit Models,
1986–2017 (monthly)

Notes: The shaded areas denote U.S. recession as defined by the NBER. The thick solid line (broken line) [thin black line] describes the path of a probit
model that assigns a probability of future recession based on data on the position of the curvature hump along the maturity axis, or D-star (the slope of
the 10-year less 1-year yield curve, YC10-1) [size of curvature at the 5-year tenor as a premium over the average of the 1- and 10-year yields, Curve].
The RMSE of the D-star model (4.53) is lower than that for the yield curve slope (4.68) and curvature [4.96] models. The D-star (yield curve)
[curvature] model correctly predicted 12 (6) [4] of the 34 recessionary months a year ahead.

the D-star model, 12. The yield-curve slope model pro- Curvature, in turn, is traditionally explained by the pos-
duced three type I errors (false positives or “phantom” itive effect on long-dated bond returns of the expected
recession warnings) and 28 type II errors (false negatives mean-reversion of yields.9 Yet an equally valid reason
or “missed” recession months). The curvature model is supply and demand imbalances, assuming limits
recorded one type I and 30 type II errors. The D-star on the substitutability between bond tenors. In other
model gave 10 type I errors and 22 type II errors. The words, curvature may ref lect excess demand (supply)
percentage of incorrect predictions7 that are corrected by because so-called safe assets at different tenors along
the D-star (yield-curve slope) [curvature] model is 5.9% the maturity axis resulting in higher (lower) prices and
(8.8%) [8.8%] at the 50% threshold, rising (falling) to lower (higher) yields. Today, the canonical safe asset
17.7% (0%) [0%] at the 70% threshold. We conclude that is the 10-year U.S. Treasury note. Safe assets include
there are plausible alternative business cycle predictors all assets that are used in an information-insensitive
to the standard slope and curvature that can be derived fashion (Gorton, Lewellen, and Metrick [2012]), while
from the term structure. We conjecture this is because simultaneously meeting minimum liquidity require-
the distributional pattern and size of term premia play a ments.10 For example, a yield-curve hump positioned
key role in the monetary transmission process. at a short maturity could describe an excess demand for
safe assets at more distant investment horizons, whereas
CURVATURE OF THE TERM STRUCTURE a hump positioned at a long maturity might suggest a
greater risk appetite among investors and hence a greater
The importance of different maturity combinations desire to hold risk assets rather than safe assets. Fluc-
depends on the curvature of the term structure. A posi- tuations in the position of the curvature hump and the
tive curvature to the term structure8 tells us that the yield changing importance of different maturity combinations
curve is generally steeper between shorter maturities. occur because investors in aggregate actively alter their

24    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

average time horizons alongside changing risk appetites The short-term policy effect assumes that prices are
and liability needs, and because of movements in the sufficiently sticky that a fall in nominal rates translates
relative importance of investors with specific duration into a decrease in real interest rates (Rt ), which, in turn,
targets. In this latter case, it may be that, say, banks boosts spending and increases future economic activity.
with relatively shorter investment horizons and with less Long-term interest rates are considered to remain unaf-
need for duration dominate the bond market, or equally fected. For the longer-term inflation effect, lower short-
possible, that it is long-term pension funds with more term interest rates are also associated with a monetary
distant horizon liability targets that matter. The exis- expansion that causes future inf lation (πt ) to increase
tence of this bias that favors different preferred habitats and so acts to further reduce real interest rates. The
is plausible given our knowledge of changes in the own- net effect raises long-term yields. These, in turn, will
ership structure of the U.S. Treasury market. The dis- correlate with faster economic growth that results from
tribution of investors has changed markedly over time, the lower real interest rates and which should be also
shifting from short-maturity holders, such as commer- associated with a pickup in inf lation. In turn, the risk-
cial banks and property and casualty insurers, to inves- taking channel, associated most with the work of Borio
tors with long-term liabilities, such as pension funds and Zhu [2012], actively involves both risk and bond
and life insurance companies. In 1955, for example, the term premia. This channel is described here under two
former group held 44.1% of outstanding U.S. Treasuries, assumptions: First, government bonds represent safe
but by 2015 this percentage had shrunk to only 4.6%. assets for many investor types. Second, the supply of
In contrast, the importance of long-term funds rose from liquidity into the financial economy alters the probabili-
14.0% to 26.2% over the same period, whereas foreign ties of systemic risk because it makes it either easier or
participants in U.S. Treasuries, which tend to have a more difficult to refinance investment positions. There-
preference for mid-duration bonds, have jumped from fore, a Central Bank monetary expansion, by reducing
3.3% to 46.6% of all holdings. A crude approximation systemic risks, leads to a fall in the demand for safe
of the changing maturity preferences of investors can asset government bonds as investors switch their funds
be gauged by taking a weighted-average of the target into risk assets. The resulting excess supply of Treasuries
maturities11 of each group, scaled by their ownership of raises term premia and, because the importance of term
outstanding Treasuries. The resulting time series show premia (tptm ) increases with maturity (m), this leads to a
the preferred maturity target rises from 3.4 years in 1955 steepening of the yield curve.12 Experience gained since
to 5.6 years in 1975 to 6.9 years in 2015. the GFC suggests that this negative demand effect on
These observations about changing preferred habi- bond prices dominates any positive supply effect induced
tats, particularly given the scale of Large-Scale Asset by policymakers, through scarcity and duration effects
Purchases (LSAPs) and maturity transformation policies (see Krishnamurthy and Vissing-Jorgensen [2011]),
undertaken by the U.S. authorities since the 2007–2008 when liquidity injections are the counterpart of LSAPs
global financial crisis (GFC), suggest that there may be of Treasuries. Risk assets are defined here to include
additional information in the position of the hump in equities and corporate bonds, as well as direct invest-
curvature along the maturity axis as well as from the ments in capital projects. Greater investment spending
slope of the yield curve, itself. Looked at another way, boosts economic growth through the usual multiplier
it is the size and pattern of notional term premia across effects. A connected transmission channel involves credit
tenors that really matters, not a single maturity spread. providers, such as banks. When perceived credit risks are
high, they, too, hold government bonds as safe assets. As
MONETARY TRANSMISSION CHANNELS their risk appetite increases, possibly as the result of easier
funding conditions, they will substitute riskier corpo-
There are three broad monetary transmission rate loans for government bonds. The resulting boost
channels that involve the yield curve: to credit growth also raises future economic activity.

1. Short-term policy effect DATA DESCRIPTION

2. Longer-term inf lation effect
3. Risk-taking effect Persistent questions over the reliability of existing
empirical term premia estimates lead us to use data on

Spring 2018 The Journal of Fixed Income    25

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

spot yields directly. We take generic spot yield data for perfect correlation. The full sample and the GFC results
U.S. Treasury notes and bonds of 1-, 2-, 3-, 5-, 7-, 10- appear to attach greater significance to shorter-term
and 20-year constant maturities, all sourced from the spreads, with maturity combinations at longer maturities
U.S. Treasury, estimated using a spline-based model and being the least effective. In the 1947–1984 sample period,
published in the Federal Reserve’s H15 release. These the mid-duration 7-year less 1-year spread appears to
monthly data, chosen for their long history, are calculated work best, whereas in the 1985–2017 period it is the
as the average of nominal daily yields and they begin in longer 20-year less 1-year spread. However, it seems
April 1953, although 2-year notes start only from June clear that no one maturity combination stands out as
1976. To further increase the sample size, we also take ideal and the loadings attached to the different maturity
zero-coupon U.S. Treasury yields covering the earlier13 combinations show a remarkable variation over time.
period, 1946–1953, from Shiller [1990] and splice these This inconsistency is also apparent from principal
on to the official U.S. Treasury series. The data sample component analysis. Disaggregating the data sample by
extends through to July 2017. The U.S. Treasury, in decade, shows that although the absolute size of the three
addition, publishes yield data for 30-year bonds starting principal component loadings are remarkably constant,
from February 1977. However, the 30-year bond is the peak curvature loading frequently switches between
typically less liquid than some of the earlier maturi- maturities. Through the 1946–1949 immediate post-
ties, such as the 10-year note, and between March 2002 WWII years, the 1-year bond enjoyed the largest loading;
and January 2006 new issuance of 30-year bonds halted during the 1950s and 1960s, as economies strengthened,
altogether.14 This creates potential estimation problems the third principal component was most heavily loaded
because the 30-year is only available for roughly half the on to the five-year maturity; in the 1970s credit boom
entire data period. In addition, the typical yield curve this had moved out to the 7-year maturity, but through
f lattens noticeably beyond the 15–20 year maturities, the 1980s, the loading on the 3-year bond was heaviest;
which makes calculating a consistent statistic difficult. the 1990s saw a return to the 5-year maturity; it rose to
Consequently, we consider maturities that terminate 7 years in the 2000s, but then dropped back to the 3-year
with the 20-year Treasury note. Equally, we ignore maturity after 2010, in the wake of the Great Recession.
bills and instruments with maturities shorter than 1 Hanson, Lucca, and Wright [2017] similarly conclude
year. These are either dominated by near-term policy that the principal components of the yield curve are
decisions rather than interest rate expectations, or, in unstable over time.
many cases, they incorporate credit risk. D-star, the
position of the curvature hump along the maturity axis, CONCLUSION
is calculated using a method described more fully in
Appendix B. Our results suggest that the rigid use of a spe-
cific yield-curve spread, such as the 10-year less 1-year
EMPIRICAL RESULTS Treasury yield slope denies the richness of implied
term premia data. In short, this traditional and widely
The data in Exhibit 2 show the loadings and R 2 used single spread is a poor representative of monetary
statistics from a series of single equation models that seek transmission and is likely to be an ambiguous predictor
to explain the U.S. ISM Purchasing Managers’ Index of the business cycle. Taken by itself, it may not tell
12 months forward using various maturity combina- us much. The inconsistency in the efficacy of these
tions. We report four samples: the full period 1947–2017; different maturity combinations should not be a surprise
two subsamples, 1947–1984 and 1985–2017, with the given the evidence of changing curvature and the time-
latter characterized as the period following Federal variation in preferred habitats. Yet they give support to
Reserve Chairman Volcker’s tight money experiment; our contention that Treasury yield spreads largely ref lect
and a more recent period, 2005–2017, that focuses on underlying bond term premia. These facts may usefully
the GFC and the Great Recession. push researchers into finding better ways to capture the
The results show generally high statistical sig- information implicit in term premia data, not least given
nificance and consistently high correlation between the current interest among policymakers in the risk-
the various maturity combinations, although far from taking monetary transmission channel. We showed in

26    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

Exhibit 2
Maturity Combinations and Loadings, 1947–2017 (monthly)

Notes: Each maturity combination (e.g., y10 – y1) is defined as the spread between two spot yields. The results derive from the estimation of single equation
models including an intercept (not shown). The loadings refer to the estimate of the regression slope coefficient. R 2 denotes R 2. Average correlation measures
the “average” cross-correlation between the maturity combinations. The data show high, but not perfect, correlation. The parameter estimates demonstrate
high instability over time, and the importance of difference maturity combinations in predicting the business cycle changes.
*, **, and *** refer to statistical significance at the 10%, 5%, and 1% levels, respectively, using adjusted standard errors.

the opening section of this article the value of incorpo- ever, by definition, a leftward-shifting hump ahead of
rating other yield-curve parameters into the analysis, an upcoming business recession could induce the short
such as the size and position of the curvature hump maturity combinations, for example, y 2 – y1, to reg-
along the maturity axis. This latter measure should posi- ister “false positive” signals by steepening more, even
tively correlate with a steepening curve when curve though the longer-term spreads, for example, y10 – y1,
steepness ref lects an improving business outlook. How- themselves f latten.

Spring 2018 The Journal of Fixed Income    27

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

A p p e n d i x  A
PROBIT AND OLS REGRESSION RESULTS

Yield-Curve Slope (10 year – 1 year)

28    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

 Position of Hump (D-star)

Spring 2018 The Journal of Fixed Income    29

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

 Curvature (1-5-10 year “butterfly” spread)

30    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

A p p e n d i x  B Therefore, dividing both sides by (b 2 + 1) defines an
expression for distance at each tenor:
MEASURING THE POSITION
OF THE YIELD-CURVE HUMP ( ym 0 − a − bm0 )2
dm2 0 =
(b 2 + 1)
This section describes a general, non-parametric mea-
sure for the lateral position (m 0 ) of the hump in the term Or, taking the square root and re-expressing in absolute
structure that is independent of the mathematical formu- terms
lation used to generate the yield curve. It uses a standard
geometric measure of the perpendicular distance (dm0 ) from ym 0 − a − bm0
a point (m 0, ym0 ) to a line, as defined by dm 0 =
(b 2 + 1)
ym = a + b. m
At each point in time (t), there exist distance measures
where m denotes maturity and y m is the spot yield for for each maturity m = 0, …, n. The distance-weighted average
that maturity. Let the perpendicular intersect the line at of these maturities15 describes the position of the hump (Dt*):
another point (p, q). Then, the distance dm0 between these n
points is ∫ m ⋅ dm ,t n

Dt∗ = = ∫w ⋅m
m=0
n m ,t

dm 0 = ( p − m 0 ) + ( y m 0 − q )
2 2 ∫ m=0
dm ,t m=0

where wm,t = dm,t/∫dm,t is the weighting factor for each tenor, m.

This expression describes the maturity of the average
curvature and defines D-star (Dt*). The expression is posi-
tive by definition because maturity (m) exceeds zero and the
weights used to calculate the average comprise individual
distances (dm,t ) measured in absolute terms. It can be cal-
culated for different shaped interest rate term structures,
including multiple-humps, convex humps, inverted and
partially inverted structures. D-star can be defined in the
case of an inverted term structure because absolute distance
measures are used. Equally, the aggregation of these distance
The values of m 0, ym0, a and b are known. Squaring both sides measures over the cross-section of maturities means that
and multiplying by (b 2 + 1) to eliminate the two unknown D-star can be calculated for term structures with multiple
values p and q gives humps. It may or may not coincide with the point of peak
curvature along the maturity axis, which is defined as the
(b 2 + 1) ( p − m0 )2 + ( ym 0 − q )2  horizon (m) with the greatest distance measure (dm,t ). When
curvature is symmetric with a single hump, these measures
= ( p − m0 )2 − 2b( ym 0 − q )( p − m0 ) + b 2 ( ym 0 − q )2 should align, but when curvature is positively (negatively)
+ b 2 ( p − m0 )2 + 2b( ym 0 − q )( p − m0 ) + ( ym 0 − q )2 skewed the average maturity reading will lie above (below)
the peak curvature.
(b 2 + 1)dm2 0 = [b( p − m0 ) + ( ym 0 − q )]2 + [( p − m0 ) − b( ym 0 − q )]2 In practice, like the similar measures of level (e.g., y1,t ),
slope (e.g., y10,t –y1,t ) and curvature (e.g., y5,t–(y10,t + y1,t)/2),
And since if follows that q = a + bp and b = (p - m 0 )/ the value of D-star depends upon which bond maturities are
(ym0 - q), this can be rewritten as included in the calculation. Correspondingly, it is inappro-
priate to compare a D-star value calculated over a maturity
= ( ym 0 − a − bm0 )2 + 0 range of 20 years with one measured over 10 years. Another
issue is continuity. Bonds are often issued at specific maturi-
ties (and generic U.S. Treasury yields are published at 1, 2,
3, 5, 7, 10, 20, and 30 years), which means that this statistic

Spring 2018 The Journal of Fixed Income    31

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

will take discrete and possibly non-unique values at each Accord, the Fed was de facto under U.S. Treasury control
sampling point. Therefore, in order to estimate D-star, we and expected to maintain a 2½% Treasury yield ceiling.
approximate it using the discrete time horizons m = 1, …, n, 14
Traditionally, U.S. Treasury bond maturities from 11
for each monthly data point as to 29 years are considered the least liquid.
15
Market participants might describe this as the
∑
n
m ⋅ dm ,t “duration-weighted average butterf ly spread across tenors.”
∗
D ≈ m =1

∑
t n
m =1
dm ,t
REFERENCES
where m = 1, …., n, dm,t ≥ 0, for all m. Adrian, T., R. Crump, B. Mills, and E. Moench. “Treasury
Term Premia: 1961–Present.” Liberty Street Economics, Federal
ENDNOTES Reserve Bank of New York, May 2014.

The author would like to thank Pavol Povala for helpful Borio, C., and H. Zhu. “Capital Regulation, Risk-Taking
comments. and Monetary Policy: A Missing Link in the Transmis-
1
U.S. Treasury notes are issues with maturities of 2, 3, sion Mechanism?” Journal of Financial Stability, Vol. 8, No. 4
5, 7, and 10 years, while Treasury bonds have maturities of 20 (2012), pp. 236-251.
and 30 years: the only difference between notes and bonds is
the length until maturity. Treasury bills are short-term bonds Bruno, V., and H.S. Shin. “Capital Flows and the Risk-
that mature within 1 year or less from their time of issuance. Taking Channel of Monetary Policy.” Journal of Monetary
2
Or yield-to-maturity. Economics, Vol. 71, No. C (2015), pp. 119-132.
3
Bond analysis involves several risk dimensions including
illiquidity, default and duration. Traditionally, default-free, Campbell, J.Y. “Some Lessons from the Yield Curve.” Journal
liquid government bonds have a single risk or term premia, of Economic Perspectives, Vol. 9, No. 3 (1995), pp. 125-152.
based on their period to redemption.
4
The quality of these credits is evaluated by independent Cochrane, J.H., and M. Piazzesi. “Bond Risk Premia.”
agencies, such as Moody’s and Standard & Poor’s, who assign The American Economic Review, Vol. 95, No. 1 (2005).
ratings to each bond (e.g., AAA, B, etc).
5
According to Adrian et al. [2014], movements in term Estrella, A., and G. Hardouvelis. “The Term Structure as a
premia account for 69.2% of the variation in 10-year spot Predictor of Real Economic Activity.” The Journal of Finance,
yields, compared with13.1% of the variation in 1-year spot 46 (1991), pp. 555-576.
yields.
6
Root mean-squared error. Fama, E.F. “Term Premiums and Default Premiums in
7
From a constant probability assumption. Money Markets.” Journal of Financial Economics, Vol. 17, No 1
8
Meaning concave to the maturity axis. (1986), pp. 175-196.
9
Because curvature is a persistent yield-curve feature
(Gurkaynak, Sack, and Wright [2006]) another explanation Gorton, G., S. Lewellen, and A. Metrick. “The Safe Asset
comes from the implied greater option value inherent in Share.” AER P&P, 2012.
longer-dated bonds given yield volatility.
10
According to the IMF [2012], a safe asset is a financial Gurkaynak, R.S., B. Sack, and J.H. Wright. “The U.S. Trea-
instrument that provides: (1) low market and credit risks; sury Yield Curve: 1961 to the Present.” U.S. Federal Reserve
(2) high market liquidity; (3) limited inf lation risks; (4) low Discussion Paper, 2006–28, 2006.
exchange rate risks; and (5) limited idiosyncratic risks. It will
likely correlate negatively with investors’ risk appetite. Hanson, S.G., D.O. Lucca, and J.H. Wright. Interest Rate
11
We notionally assume 12 years required duration for Conundrums in the Twenty-First Century, Federal Reserve Bank
long-term funds; 6 years for foreign investors, and 3 years for of New York Staff Report #810, March 2017.
banks and P&C insurers.
12
Adrian et al. [2014]. International Monetary Fund (IMF). Global Financial Stability
13
The Third Liberty Bond Act, passed at the end of Report, Washington D.C., April 2012.
World War I, set a yield ceiling of 4¼% on all new issues
of longer than seven years maturity, and until the 1951

32    What Does the Yield-Curve Slope R eally Tell Us? Spring 2018

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018

Krishnamurthy, A., and A. Vissing-Jorgensen. “The
Effects of Quantitative Easing on Interest Rates: Channels
and Implications for Policy.” Brookings Papers, Fall 2011,
pp. 215-287.

Leibowitz, M.L., A. Bova, and S. Kogelman. “Long-Term

Bond Returns under Duration Targeting.” Financial Analysts
Journal, Vol. 70, No. 1 (2014).

Litterman, R., and J.A. Scheinkman. “Common Factors

Affecting Bond Returns.” The Journal of Fixed Income, Vol. 1,
No. 1 (1991), pp. 54-61.

Mishkin, F.S. “What Does the Term Structure Tell Us about

Future Inf lation?” Journal of Monetary Economics, Vol. 25,
No. 1 ( January 1990), pp. 77-95.

Moench, E. “Term Structure Surprises: The Predictive Con-

tent of Curvature, Level and Slope.” Journal of Applied Econo-
metrics, Vol. 27 (2012).

Shiller, R. “The Term Structure of Interest Rates. “In

Handbook of Monetary Economics, Volume 1, edited by B.M.
Friedman and F.H. Hahn. Elsevier, 1990.

To order reprints of this article, please contact David Rowe at

drowe@ iijournals.com or 212-224-3045.

Spring 2018 The Journal of Fixed Income    33

Downloaded from http://jfi.iijournals.com/ by guest on August 4, 2018