Week 10

Mortgage-backed securities
analysis

Presented by
Dr James Cummings
Discipline of Finance

The University of Sydney Page 1

 Learning Objectives

After studying this topic, you will understand

– the cash flow yield methodology for analysing residential mortgage-backed
securities
– the limitations of the cash flow yield methodology
– how the effective duration and convexity are calculated for the cash flow
yield methodology
– one measure for estimating prepayment sensitivity
– why the Monte Carlo simulation methodology is used to value residential
mortgage-backed securities
– how interest-rate paths are simulated in a Monte Carlo simulation
methodology
– how the Monte Carlo simulation methodology can be used to determine the
theoretical value of a residential mortgage-backed security

The University of Sydney Page 2

 Learning Objectives

After studying this topic, you will understand

– how the option-adjusted spread, effective duration, and effective convexity
are computed using the Monte Carlo simulation methodology
– the complexities of modelling collateralised mortgage obligations
– the limitations of option-adjusted spread
– modelling risk and how it can be stress tested
– how the total return is calculated for a residential mortgage-backed security
– the difficulties of applying the total return framework to residential
mortgage-backed securities

The University of Sydney Page 3

 Static cash flow yield methodology

– The static cash flow yield methodology is the simplest to use, although we
shall see that it offers little insight into the relative value of an RMBS.
– It begins with the computation of the cash flow yield measure that we
described for pass-throughs.
– To illustrate the cash flow yield, we will use a CMO structure.
– We can compute cash flow yields according to various PSA prepayment
assumptions for the four tranches assuming different purchase prices.
– The greater the discount assumed to be paid for the tranche, the more a
tranche will benefit from faster prepayments.
– The converse is true for a tranche for which a premium is paid: The
faster the prepayments, the lower the cash flow yield will be.

The University of Sydney Page 4

 Price cash flow yield table for the four tranches in FJF-06

The University of Sydney Page 5

 Price cash flow yield table for the four tranches in FJF-06

The University of Sydney Page 6

 Price cash flow yield table for the four tranches in FJF-06

The University of Sydney Page 7

 Price cash flow yield table for the four tranches in FJF-06

The University of Sydney Page 8

 Static cash flow yield methodology

– Vector analysis
– One practice that market participants use to overcome the drawback of
the PSA benchmark is to assume that the PSA speed can change over
time.
• This technique is referred to as vector analysis.
• Vector analysis is useful for CMO tranches that are dramatically
affected by the initial slowing down of prepayments and then
speeding up of prepayments, or vice versa.

The University of Sydney Page 9

 Static cash flow yield methodology

– Limitations of the cash flow yield

– The shortcomings present in the yield to maturity are equally present in
the cash flow yield measure:
1) the projected cash flows are assumed to be reinvested at the cash
flow yield
2) the RMBS is assumed to be held until the final payout based on
some prepayment assumption
– The cash flow yield is dependent on realisation of the projected cash
flow according to some prepayment rate.
• If actual prepayments vary from the prepayment rate assumed, the
cash flow yield will not be realised.

The University of Sydney Page 10

 Static cash flow yield methodology

– Yield spread to Treasuries

– At the time of purchase, it is not possible to determine an exact yield for
an RMBS.
• The yield will depend on the actual prepayment experience of the
mortgages in the pool.
• Nevertheless, the convention in all fixed-income markets is to
measure the yield on a non-Treasury security relative to that of a
‘comparable’ Treasury security.
– The repayment of principal over time makes it inappropriate to
compare the yield of an RMBS to a Treasury of a stated maturity.
• Instead, market participants have used two measures: duration and
average life.

The University of Sydney Page 11

 Static cash flow yield methodology

– Static spread
– The practice of spreading the yield to the average life on the
interpolated Treasury yield curve is improper for an amortising bond,
even in the absence of interest-rate volatility.
• What should be done is to calculate what is called the static
spread.
– This is the yield spread in a static scenario (i.e., no volatility of
interest rates) of the bond over the entire theoretical Treasury
spot rate curve.

The University of Sydney Page 12

 Static cash flow yield methodology

– Static spread
– There are two ways to compute the static spread for RMBS.
• One way is to use today’s yield curve to discount future cash flows
and keep the mortgage refinancing rate fixed at today’s mortgage
rate.
– Because the refinancing rate is fixed, the investor can specify a
reasonable prepayment rate for the life of the security.
• The second way allows the mortgage rate to go up the curve as
implied by the forward interest rates.
– This procedure is called the zero-volatility OAS or the Z-
spread.
– In this case, a prepayment model is needed to determine the
vector of future prepayment rates implied by the vector of
future refinancing rates.

The University of Sydney Page 13

 Static cash flow yield methodology

– Effective duration
– Modified duration is a measure of the sensitivity of a bond’s price to
interest-rate changes, assuming that the expected cash flow does not
change with interest rates.
• Modified duration is not an appropriate measure for mortgage-
backed securities because prepayments cause the projected cash
flow to change as interest rates change.
• When interest rates fall, prepayments are expected to rise.
• As a result, when interest rates fall, duration may decrease rather
than increase.
• This property is referred to as negative convexity.

The University of Sydney Page 14

 Static cash flow yield methodology

– Effective duration
– Negative convexity has the same impact on the price performance of an
RMBS as it does on the performance of a callable bond.
• When interest rates decline, a bond with an embedded call option,
which is what an RMBS is, will not perform as well as an option-free
bond.
• Although modified duration is an inappropriate measure of interest-
rate sensitivity, there is a way to allow for changing prepayment
rates on cash flow as interest rates change.
• This is achieved by calculating the effective duration, which allows
for changing cash flow when interest rates change.
– To illustrate calculation of effective duration for CMO classes, consider
FJF-06 once again.

The University of Sydney Page 15

 Static cash flow yield methodology

– Effective duration
– To illustrate the calculation, consider tranche C.
– The data for calculating modified duration using the approximation
formula is
P− 102.1875
= (at 165 PSA), P+ 98.4063 (at 165 PSA), P0 =100.2813,
∆y =0.0025
• Substituting into the duration formula yields
102.1875 − 98.4063
modified duration
= = 7.54
2 × (100.2813) × (0.0025)

The University of Sydney Page 16

 Static cash flow yield methodology

– Effective duration
– The effective duration for the same bond class is calculated as follows:
P− 101.9063
= (at 200 PSA), P+ 98.3438 (at 150 PSA), P0 =100.2813,
∆y =0.0025
• Substituting into the formula gives

101.9063 − 98.3438
effective duration
= = 7.11
2 × (100.2813) × (0.0025)
• For all four tranches and the collateral, the effective duration is less
than the modified duration.

The University of Sydney Page 17

 Static cash flow yield methodology

– Effective duration
– The divergence between modified duration and effective duration is
more dramatic for bond classes trading at a substantial discount from
par or at a substantial premium over par.

The University of Sydney Page 18

 Static cash flow yield methodology

– Effective convexity
– To illustrate the convexity formula, consider once again tranche C in FJF-
06.
• The standard convexity is approximated as follows:

98.4063 + 102.1875 − 2 × (100.2813)

2
= 24.930
(100.2813) × (0.0025)
• The effective convexity is
98.3438 + 101.9063 − 2 × (100.2813)
2
= −249.299
(100.2813) × (0.0025)

The University of Sydney Page 19

 Static cash flow yield methodology

– Effective convexity
– The standard convexity indicates that the four tranches have positive
convexity, whereas the effective convexity indicates they have negative
convexity.
• The difference is even more dramatic for bonds not trading near
par.

The University of Sydney Page 20

 Static cash flow yield methodology

– Prepayment sensitivity measure

– The value of an RMBS will depend on prepayments.
• To assess prepayment sensitivity, market participants have used the
following measure: the basis point change in the price of an RMBS
for a 1% increase in prepayments.
• Specifically, prepayment sensitivity is defined as

prepayment sensitivity = ( Ps – P0 ) × 100

– where

P0 = initial price ( per $100 par value ) at assumed prepayment speed

Ps = price ( per $100 par value ) assuming a 1% increase in prepayment speed

The University of Sydney Page 21

 Static cash flow yield methodology

– Prepayment sensitivity measure

– A security that is adversely affected by an increase in prepayment
speeds will have a negative prepayment sensitivity, while a security that
benefits from an increase in prepayment speed will have a positive
prepayment sensitivity.

The University of Sydney Page 22

 Monte Carlo simulation methodology

– For some fixed-income securities and derivative instruments, the periodic

cash flows are path dependent.
– This means that the cash flows received in one period are determined
not only by the current and future interest-rate levels but also by the
path that interest rates took to get to the current level.
– In the case of mortgage pass-through securities, prepayments are path
dependent because this month’s prepayment rate depends on whether there
have been prior opportunities to refinance since the underlying mortgages
were originated.
– Unlike mortgage loans, the decision by a corporate issuer whether to
refund an issue when the current rate is below the issue’s coupon rate is
not dependent on how rates evolved over time to the current level.

The University of Sydney Page 23

 Monte Carlo simulation methodology

– Pools of pass-throughs are used as collateral for the creation of

collateralised mortgage obligations (CMOs).
– For CMOs there are typically two sources of path dependency in a
CMO tranche’s cash flows.
1) The collateral prepayments are path dependent.
2) The cash flow to be received in the current month by a CMO
tranche depends on the outstanding balances of the other
tranches in the deal.
– Thus, we need the history of prepayments to calculate these balances.

The University of Sydney Page 24

 Monte Carlo simulation methodology

– Because of the path dependency of an RMBS’s cash flow, the Monte Carlo
simulation method is used for these securities rather than the static method.
– Conceptually, the valuation of pass-throughs using the Monte Carlo method
is simple.
– In practice, it is very complex.
– The simulation involves generating a set of cash flows based on
simulated future mortgage refinancing rates, which in turn imply
simulated prepayment rates.
– Valuation modelling for CMOs is similar to valuation modelling for pass-
throughs, although the difficulties are amplified because the issuer has sliced
and diced both the prepayment risk and the interest-rate risk into smaller
pieces called tranches.

The University of Sydney Page 25

 Monte Carlo simulation methodology

– Using simulation to generate interest-rate paths and cash flows

– The typical model that Wall Street firms and commercial vendors use to
generate random interest-rate paths takes as inputs today’s term
structure of interest rates and a volatility assumption.
• The term structure of interest rates is the theoretical spot rate curve
implied by today’s Treasury securities.
• The volatility assumption determines the dispersion of future interest
rates in the simulation.
• The simulations should be normalised so that the average simulated
price of a zero-coupon Treasury bond equals today’s actual price.

The University of Sydney Page 26

 Monte Carlo simulation methodology

– Using simulation to generate interest-rate paths and cash flows

– The simulation works by generating many scenarios of future interest-
rate paths.
• In each month of the scenario, a monthly interest rate and a
mortgage refinancing rate are generated.
• The monthly interest rates are used to discount the projected cash
flows in the scenario.
• The mortgage refinancing rate is needed to determine the cash flow
because it represents the opportunity cost that the mortgagor is
facing at that time.
– Prepayments are projected by feeding the refinancing rate and loan
characteristics, such as age, into a prepayment model.
• Given the projected prepayments (voluntary and involuntary), the
cash flow along an interest-rate path can be determined.

The University of Sydney Page 27

 Monte Carlo simulation methodology

– Using simulation to generate interest-rate paths and cash flows

– To make this more concrete, consider a newly issued mortgage pass-
through security with a maturity of 360 months.
• We simulate N interest-rate path scenarios.
– Each scenario consists of a path of 360 simulated one-month
future interest rates.
– Just how many paths should be generated is explained later.
• We simulate the paths of mortgage refinancing rates corresponding
to the scenarios.
• Assuming these mortgage refinancing rates, we predict the cash flow
for each scenario path.

The University of Sydney Page 28

 Simulated paths of 1-month future interest rates

The University of Sydney Page 29

 Simulated paths of mortgage refinancing rates

The University of Sydney Page 30

 Simulated cash flow on each of the interest-rate paths

The University of Sydney Page 31

 Monte Carlo simulation methodology

– Calculating the present value for a scenario interest-rate path

– Given the cash flow on an interest-rate path, its present value can be
calculated.
• The discount rate is the simulated spot rate for each month plus an
appropriate spread.
• The simulated spot rate for month T on path n is

{[1 + f1 (n)][1 + f 2 (n)][1 + fT (n)]}1/T − 1

zT (n) =
– where
zT (n) = simulated spot rate for month T on path n
f j (n) = simulated future 1-month rate for month j on path n
– The interest-rate path for the simulated future one-month rates can be
converted to the interest-rate path for the simulated monthly spot rates.

The University of Sydney Page 32

 Simulated paths of monthly spot rates

The University of Sydney Page 33

 Monte Carlo simulation methodology

– Calculating the present value for a scenario interest-rate path

– The present value of the cash flow for month T on interest-rate path n
discounted at the simulated spot rate for month T plus some spread is

CT (n)
PV[CT (n)] =
[1 + zT (n) + K ]T
• where
CT (n) = cash flow for month T on path n
zT (n) = spot rate for month T on path n
K = appropriate risk-adjusted spread
– The present value for path n is the sum of the present value of the cash
flow for each month on path n.

The University of Sydney Page 34

 Monte Carlo simulation methodology

– Determining the theoretical value

– The present value of a given interest-rate path can be thought of as the
theoretical value of a pass-through if that path was realised.
• The theoretical value of the pass-through can be determined by
calculating the average of the theoretical value of all the interest-
rate paths.
– Looking at the distribution of the path values
– The theoretical value generated by the Monte Carlo simulation method
is the average of the path values.
• There is valuable information in the distribution of the path values.
• If there is substantial dispersion of the path values, the investor is
warned about the potential variability of the model’s value.

The University of Sydney Page 35

 Monte Carlo simulation methodology

– Simulated average life

– The average life reported in a Monte Carlo analysis is the average of
the average lives along the interest-rate paths.
• The greater the range and standard deviation of the average life,
the more uncertainty there is about the security’s average life.
– Option-adjusted spread
– The option-adjusted spread (OAS) is a measure of the yield spread that
can be used to convert dollar differences between value and price.
• It represents a spread over the issuer’s spot rate curve or
benchmark.
– In the Monte Carlo model, the OAS is the spread K that—when added
to all the spot rates on all interest-rate paths—will make the average
present value of the paths equal to the observed market price (plus
accrued interest).

The University of Sydney Page 36

 Monte Carlo simulation methodology

– Option cost
– The implied cost of the option embedded in any RMBS can be obtained
by calculating the difference between the OAS at the assumed volatility
of interest rates and the static spread:
option cost static spread − option-adjusted spread
=

– Effective duration and convexity

– These measures can be calculated using the Monte Carlo method.
• The bond’s OAS is found using the current term structure of interest
rates.
• The bond is repriced holding OAS constant but shifting the term
structure.
– Two shifts are used to get the prices needed: In the first, yields
are increased; in the second, yields are decreased.

The University of Sydney Page 37

 Monte Carlo simulation methodology

– Selecting the number of interest-rate paths

– A typical OAS run will be done for 512 to 1,024 interest-rate paths.
– The number of interest-rate paths determines how ‘good’ the estimate is,
not relative to the truth but relative to the model.
• The more paths, the more average spread tends to settle down.
• It is a statistical sampling problem.
– Most Monte Carlo simulation models employ some form of variance
reduction to cut down on the number of sample paths necessary to get
a good statistical sample.
• Variance reduction techniques allow us to obtain price estimates
within a tick.

The University of Sydney Page 38

 Monte Carlo simulation methodology

– Drawbacks of the OAS methodology

– The OAS methodology provides a measure for relative valuation of
MBS products that is vastly superior to some of the ad hoc measures that
were historically used by portfolio managers.
• Those traditional measures fail to recognise the embedded options
in these complex products.
• Nevertheless, the methodology is not without its drawbacks.
– The first drawback is the prepayment model uncertainty.
– The second drawback is that in pricing MBS, there are differing OAS
levels.

The University of Sydney Page 39

 Monte Carlo simulation methodology

– OAS constancy problem

– Davidson and Levin identify a problem with calculating price sensitivity
measures, such as interest-rate and prepayment sensitivity, using the
OAS methodology.
• In measuring effective duration and effective convexity, the
calculation assumed that the OAS is held constant, and the interest-
rate paths are shifted by a specified number of basis points.
– In the market, MBS products are priced at differing OAS levels.
– This is a serious blow to the idea of keeping OAS constant for measuring
price sensitivity.
• One idea is to adjust a prepayment model so that it equates OAS
levels.

The University of Sydney Page 40

 Total return analysis

– Neither the static cash flow methodology nor the Monte Carlo simulation
methodology will tell a money manager whether investment objectives can
be satisfied.
– The performance evaluation of an individual RMBS requires
specification of an investment horizon, whose length for most financial
institutions is dictated by the nature of its liabilities.
– The measure that should be used to assess the performance of a security or
a portfolio over some investment horizon is the total return.

The University of Sydney Page 41

 Total return analysis

– Horizon price for collateralised mortgage obligation tranches

– The most difficult part of estimating total return is projecting the price at
the horizon date.
• In the case of a CMO tranche, the price depends on the
characteristics of the tranche and the spread to Treasuries at the
termination date.
• The key determinants are the ‘quality’ of the tranche, its average
life (or duration), and its convexity.
– Quality refers to the type of CMO tranche.

The University of Sydney Page 42

 Total return analysis

– Option-adjusted spread total return

– The total return and OAS frameworks can be combined to determine
the projected price at the horizon date.
• At the end of the investment horizon, it is necessary to specify how
the OAS is expected to change.
• The horizon price can be ‘backed out’ of the Monte Carlo simulation
model.
– Assumptions about the OAS value at the investment horizon reflect the
expectations of the money manager.
• It is common to assume that the OAS at the horizon date will be the
same as the OAS at the time of purchase.
• A total return calculated using this assumption is referred to as a
constant-OAS total return.

The University of Sydney Page 43

 Required reading

– Fabozzi, F.J. and Fabozzi, F.A. (2021), Bond Markets, Analysis, and
Strategies, 10th edition, MIT Press
– Chapter 20

The University of Sydney Page 44

 Homework problems

– Fabozzi and Fabozzi (2021)

– Chapter 20
• Questions 1, 14, 23

The University of Sydney Page 45