PROFESSOR: The credit spread, also sometimes called

a yield spread, is the difference

in the yield-to-maturity on a risky bond
and on a Treasury bond of similar maturity.
The statistical approach to pricing credit risk
seeks to explain a portion of the credit spread
based on estimates of default and recovery rates.
Data on credit spreads by rating is readily available,
for instance, on the website, FRED,
which provides economic time series data from the Federal
Reserve Bank of St. Louis.
Such time series data are often presented
as an option-adjusted spread.
As I mentioned last week, that means that the reported yields
have been adjusted to remove the effects
of other embedded options, such as prepayment or call options.
For a given credit rating, the observed yield spread also
tends to vary with maturity.
Credit spreads tend to increase with maturity,
as one might expect, due to the greater
risk of an adverse event occurring over longer horizons.
However, that pattern sometimes reverses
at very long maturities, with credit spreads
starting to narrow.
That may partially reflect compositional differences
in firms borrowing at different maturities,
with only high quality borrowers able to issue
very long maturity debt.

These figures show the average credit spread on BB bonds

on the top panel and on AA bonds in the bottom panel
between 1997 and 2020.
The shaded areas are recessions.
The BB bonds are below investment grade
and have significantly higher spreads
than do the much safer AA bonds.
Not surprisingly, the spreads mirror the pattern
we saw for default rates, which spike
during economic downturns.
The increase in yields during recessions
translates into a decrease in prices,
again, consistent with the idea that risky debt has
a positive beta that implies that their prices will
be lower because of the market risk premium.

The difference in the yield to maturity between treasuries

and default of securities of the same maturity
can conceptually be decomposed into several components.
The first is based on the statistics we've just
been discussing, the default rate and the loss given
default. The product of the two gives
an annualized expected loss rate.

In addition, there's compensation

for the market or beta risk associated with these losses.
In the representation here, that's
incorporated using risk-neutral probabilities
and risk-neutral expectations in the calculation of the expected
loss rate.
The risk-neutral expected loss rate
is higher than the physical expected loss rate
because it includes the value of the associated market risk.
Empirical analyses of credit spreads
find that plausible estimates of risk-adjusted expected losses
account for only a fraction of the observed credit spread.
The remaining non-credit component
is attributable to other factors like differences and tax
treatment and the presence of other embedded options.
The residual, after accounting for all identifiable effects,
is usually described as a liquidity premium.
The liquidity premium is viewed as compensation
to investors for the relative illiquidity of corporate bonds
relative to treasuries.
Finally, something to remember in modeling credit
spreads and their implications for default and recovery rates
is that because the non-credit component of the rate spread
is often sizable, neglecting that fact
may result in overestimating the probability of default
and the loss rate.

A fairly simple valuation model for risky debt

that's consistent with the statistical approach to credit
risk modeling is developed in the next few slides.
The model takes, as inputs, a bond's fixed characteristics
that include its coupon rate and time to maturity
and assumes a face value of a dollar.
It also takes as inputs estimates of the default
rate, the recovery rate and the risk-adjusted expected return
on the bond.
The model outputs are the bond's predicted yield to maturity
per period and its price.
The model is implemented in the spreadsheet default.xls that's
available to you on the class page.
Note that the model abstracts from non-credit features
of the yield spread and so may not
produce price predictions that match observed market prices.
Matching market prices might require an adjustment
based on an estimate of the liquidity
premium and other factors.
In this model, T denotes the number of remaining periods
in a bond's life.
If coupons are paid annually, then T will represent years.
If coupons are semi-annual, T will be years times 2.
And the coupon rate, c, will be the annual coupon
rate divided by 2.
Similarly, if coupons are paid monthly,
T should be the number of years of remaining life times 12.
And the coupon rate, c, should be
the annual rate divided by 12.
All calculations are per dollar of face value.

With a constant default rate, d, assumed

per period, the probability of the firm
not defaulting up until time t is
1 minus d to the power of t minus 1.
The fixed recovery rate is denoted by a small g.
It's the share of principal and accrued interest that's
assumed to be recovered at the time of a default.
The expected rate of return, r, is the risk-free rate
plus a market risk premium.
For example, a BBB bond might be expected
to have a beta of about 0.1.
This expected return will be used
in the model for discounting expected cash flows.

In the model, the bond price is found

by calculating the present value of expected future cash
flows discounting at the expected return, r.

To understand the model, consider the expected payment

on a bond at a time i which is less than T.
It clearly depends on whether or not
a default has already occurred, because if the bond is already
in default, then the expected payment at that time is zero.
The probability that the firm has survived without default
until time i is 1 minus d to the i minus 1.
And in that case, it will make a payment during that period.
If there's no default at time i, the full coupon payment
will be made.
That happens with probability 1 minus d.
If there is a default, which happens with probability d,
a recovery will be made based on the recovery rate, g,
multiplied by the amount owed of 1 plus c.
The last term here represents the present value
of the return of principal at maturity
when there's no default over the entire life of the bond.

Now, once you've used this equation

to estimate the bond's price, we can figure out
its yield-to-maturity in the usual way
by equating that price to the present value of the promised
cash flows treating them as if they're certain.

If you play with the spreadsheet,

you'll see that for bonds with high default or low recovery
rates, the yield-to-maturity will be significantly higher
than the expected return.
That's because part of the yield spread
is compensation for expected losses.