Decomposing expected return

One of the benefits of dividing expected returns into various components is that if a portfolio manager’s
view changes, he is able to pinpoint exactly how this change will affect his expected return. Moreover,
this decomposition allows us to divide the expected return on a bond position into a component given
an unchanged yield curve (a useful base case) and a component resulting from a portfolio manager’s
yield view relative to an unchanged yield curve. The decomposition should thus help investors to better
understand their own investment positions and any implicit bets that are reflected in their position.

Following is how the expected bond return can be decomposed into various components:

The yield income is simply the income that an investor receives from coupon payments relative to the
bond’s price. Hence, it is equal to a bond’s current yield times the strategy horizon (h). So we have

The return on reinvested income is the return that can be earned by reinvesting coupon payments at a
bond’s yield to maturity. Therefore, we have

where yn is the bond’s yield to maturity with the bond maturing in n years from now. This relationship
assumes that we invest in a bond right after its coupon has been paid, so that the first coupon a
bondholder receives would be paid a year after he invests in the bond.
Finish

Rolldown Return

If the yield curve was certain to remain unchanged over the investment horizon, then an investor’s total
expected return would be the yield income plus the return on reinvested income plus the rolldown
return.

The rolldown return is simply equal to the bond’s percentage price change resulting from an unchanged
yield curve over the strategy horizon h. Thus the bond has to be revalued at the end of the strategy
horizon assuming that yields don’t change. Then the rolldown return is:

where bond pricen-h is the bond’s price at the end of the strategy horizon h assuming an unchanged
yield curve and n-h years remaining until the bond’s maturity.

Rolldown Return

Objective: After this lesson, the learner will be able to describe rolldown returns and how to calculate
them.

Rolldown return = (bond pricen-h – bond pricen) / bond pricen

For more on the rolldown return and results from the capital gains / losses that are the result of the
bond “rolling down” the yield curve over the investment horizon toward a shorter remaining maturity,
see Module 1.

Value of convexity

Objective: After this lesson, the learner will be able to describe the value of convexity and how to
compute it.

The value of convexity comes into play if the yield curve is not constant but changes over time in level
and shape. The higher the interest rate volatility, the higher the value of convexity will be for an option-
free bond. (Note: For this analysis we will only look at option-free bonds, such as U.S. Treasury bonds.)
An investor with a long position in a bond always benefits from convexity. If a bond has positive
convexity (cx), then the bond price will increase more if interest rates decrease and it will decrease less
if interest rates increase than for a theoretical bond with zero convexity.

This relationship can be seen in the following graph.

Therefore, whatever the future path of interest rates is, the expected return of a bond with positive
convexity is going to be higher than for a zero-convexity bond. This return premium can be computed as
the value of convexity.

It is computed as follows:

Yield Volatility

Objective: After this lesson, the learner will be able to describe how yield volatilities are used and the
advantages of implied volatilities.

As yield volatilities (particularly implied volatilities obtained from options prices) are typically quoted as
relative volatilities (Vol(Δy/y)) where y is the yield and Δy is a yield change, we have to multiply the
volatility by the corresponding yield level in order to obtain the type of volatility required.

Historical volatilities are simply computed as the standard deviation of a historical interest rate series. In
contrast, implied volatilities are obtained from an option-pricing model. (More on this in Module 2.)

To come up with an implied volatility, an option embedded in a bond has to be assumed to be trading at
its fair price and an option pricing model has to be assumed to be the model that would generate the
fair price. In this case, the implied volatility is the volatility that would result in the option-pricing model
producing the fair price of the option, if the implied volatility was used as an input.
As mentioned above, the volatility estimate obtained is generally quoted in relative terms.

Expected Return from Interest Rate View

Objective: After this lesson, the learner will be able to describe expected return from interest rate view
and explain the return composition.

Finally, the expected return component resulting from the investor’s interest rate view is added on to
the relationship. This term would be zero if the investor expected the yield curve to remain unchanged.

This expected return component is computed as follows:

where Dn-h is the modified duration of a bond at the end of the strategy horizon—that is, a bond with
n-h years remaining to maturity and E(yn-h) is the bond’s expected n-h-year yield to maturity h years
from now (as opposed to the current n-h-year yield to maturity that we denote by yn-h).

We need to stress that the above return decomposition is only approximate. For example, only duration
and convexity are used to summarize the price-yield relationship and all higher-order terms are ignored.
While the expected return decomposition approximation is generally relatively precise for zero-coupon
bonds, for coupon-paying bonds, the return decomposition is more approximate, particularly for longer-
term bonds. In addition to ignoring higher-order terms in the price-yield relationship, a different coupon
reinvestment rate is assumed for different bonds as the yield to maturity measures used implicitly
assumed that all intermediate cash flows of the bond are reinvested at the same rate, namely the yield
to maturity.

There is one more potential component in the expected return decomposition that we have ignored so
far, namely local richness / cheapness effects as well as potential financing advantages. Local richness /
cheapness effects are deviations of individual maturity sectors from the fitted yield curve obtained by
using some kind of curve estimation technique. As yield curve estimation techniques are designed to
produce relatively smooth curves, there will be slight deviations along the curve. In addition to that,
there will be financing advantages to certain maturity sectors in the repo market. In most cases, these
two effects tend to be relatively small, and often the two effects at least partially offset each other.
Therefore, we have not included these effects in the above expected return decomposition.
Estimating the Inputs used to Decompose Expected Returns

Objective: After this lesson, the learner will be able to describe the differences regarding yield income,
rolldown return, value convexity, and interest rate view.

There are differences regarding how easily each of the individual components of the expected return
can be measured.

Although the quantitative approach may sound more objective, the choice of quantitative model itself
remains largely subjective, especially because a multitude of quantitative models are available that can
be used.

Decomposing Expected Returns: A Practical Example

Let’s take a look at a practical example to illustrate how the total expected return of a trading strategy
can be computed and how it can be broken up into different components.

Suppose an investor has a barbell–bullet position in bonds with remaining maturities of 2 years, 5 years,
and 10 years. The investor is long and short an initial value of $1,000,000 and his strategy horizon is one
year.

Make the following inputs in the accompanying spreadsheet to reflect this scenario:
The investor has computed the current spot rate curve, expected future spot rates, and he has obtained
the corresponding yield volatilities.

As you can see in the following graph, these inputs correspond to an expected steepening in the spot
rate curve:

Note: Remember the maturity convention we are using in the spreadsheet. Yields are aligned in
“number of years from now until maturity.” Therefore, a bond with a remaining maturity of one year at
the end of the strategy horizon one year from now will be shown as having two years from now
remaining until maturity.

Analysis of Decomposing Expected Returns Example

The total expected return (or expected profit and loss) can be decomposed as follows:
As we have weights of –0.5, 1, and –0.5 in Bond 1, 2, and 3 respectively, the total expected return of the
position computed from the individual return components is:

Total expected return = –0.5 × 3.20% + 1 × 3.05% – 0.5 × 0.46% = 1.22%

In contrast, the total expected return computed from revaluing each of the bonds at the end of the
strategy’s horizon based on the expected interest rate scenario and adding the future value of the coupon
payments (Vn-h) is:

where, for example, V1n-h is the value of the position in Bond 1 at the strategy horizon that is the sum of
the value of the bond position plus the future value of the reinvested coupon.

As you can see, the two expected return values differ slightly. However, this difference is expected
because the return decomposition used is only an approximation.

Walking Through Return Components

Let’s now use this example to go through each of the components of the total expected return and
illustrate how it is computed. As an example, we are going to focus on Bond 2.

1. Yield income: The yield income for Bond 2 is just its current yield times the strategy horizon in
years, that is, its coupon divided by the bond price times the strategy horizon. So we have

2. Return on reinvested income: This is the return that can be earned by reinvesting the coupon
payments at the bond’s yield to maturity. Therefore, for Bond 2 we have

where yn is the bond’s yield to maturity. This term is zero for Bond 2 because the strategy
horizon is only one year. Therefore, we get the first coupon payment only at the end of the strategy
horizon, too late to earn any return over the strategy horizon.

3. Rolldown return: As outlined earlier, the rolldown return equals the capital gains / losses that
are the result of the bond “rolling down” the yield curve over the investment horizon toward a
shorter remaining maturity. It is simply computed as the bond’s percentage price change
resulting from an unchanged yield curve over the strategy horizon h. So we have
 4. Rolling yield: The rolling yield is simply the sum of the yield income, the return on reinvested
income, and the rolldown return. So we have Rolling yield = 3.89% + 0.00% – 0.89% = 3.00%

5. Value of convexity: For option-free bonds, the curvature in the price / yield relationship of a
bond is always positive—that is, the relationship is convex. Therefore, if interest rates change,
an investor with a long position in a bond benefits from convexity (cx). The value of convexity
for Bond 2 is computed as follows:

Note that the yield volatility (17.22%), as given in the bond value table earlier, is obtained by
linear interpolation from the volatility term structure specified by the user. It is the volatility
that corresponds to a duration of 3.78 years, which is the duration of Bond 2 at the end of the
strategy horizon (as shown in the bond value table earlier).

6. Duration impact from view (or expected return from view): This is the impact of the expected
spot rate curve change on the return of the bond position. It is computed as follows:

This information is shown graphically for all three bonds below:

• Bond positions / trading strategies should be evaluated by examining their total expected return
rather than only yield income.

• Decomposing total expected returns allows investors to differentiate between expected returns
given static market conditions and expected returns resulting from their own yield view.

• The expected return decomposition further helps investors understand their bond position and
any implicit bets associated with it.

• However, the expected return decomposition is only an approximation and some components are
more difficult to estimate than others.

Question 1: Suppose you have a long position in an option-free bond. As the yield volatility across the
maturity spectrum increases, how does the expected return on your bond position change? The expected
return on the bond position... A is the correct answer. An investor with a long position in a bond always
benefits from convexity. If a bond has positive convexity, then the bond price will increase more if interest
rates decrease and it will decrease less if interest rates increase than for a theoretical bond with zero
convexity.

Question 2: Suppose the current spot rate curve is flat and equal to 4 percent. If you are holding a zero-
coupon bond with five years remaining to maturity and you expect no yield changes over the following
year, which of the following is a true statement? The total expected return over the following year is ...
Although you don’t expect the yield curve to change, the value of convexity is still positive because there
is uncertainty associated with future yield movements. Therefore, the total expected return is larger than
the yield income.

Question 3: Which of the following is the best answer to why the earlier expected return decomposition
approximation is less precise for coupon bonds than for zero-coupon bonds? The approximation is less
precise for coupon bonds because ... The return decomposition approximation is less precise for coupon
bonds because in order to come up with a yield summary measure for coupon bonds, a particular
reinvestment rate has to be assumed for all coupons. The most commonly used yield income measure,
yield to maturity, implicitly assumes that all coupons are reinvested at the same rate, namely the yield to
maturity. This assumption is unrealistic except when the yield curve is flat.

The price–yield relationship is approximated by using only duration. Including convexity increases the
precision of the price–yield relationship.

Richness/cheapness effects are deviations of individual maturity sections from the fitted yield curve.
Whether the richness/cheapness effects are larger for coupon bonds or zero- coupon bonds depends on
the estimation techniques used to set the yield curve. They could be larger or smaller.

Quesiton 4: Suppose one-year and two-year spot rates are currently 6 percent and 6.5 percent
respectively. Based on that information, compute the one-year forward rate one year from now.

What is the total expected return of the barbell–bullet strategy?

The total expected return is closest to ...

Therefore, the total expected return is: –0.5 × 3.10% + 1 × 3.41% – 0.5 × 3.84% = –0.06%

The total expected return of the bullet–barbell strategy equals the total expected return on the
intermediate-term bond less the equally weighted average total expected return of the short-term and
long-term bond.

Question 5

Suppose you hold a long position in a bond with one year remaining to maturity. Your holding horizon of
the bond is one year. The current spot rate curve is flat and equal to 5 percent. All other inputs are as in
Question 4. What is the total expected return on the bond with one year remaining to maturity? The total
expected return of the bond is ...

In this case, the one-year bond is held until maturity so the yield income from the bond is equivalent to
the total expected return. There is no uncertainty associated with the bond return because yields are
locked in once the bond is purchased and then held to maturity.