II.
THE
VALUATION
MODEL
where
a,
b,
and
c are
parameters,
and
Z1
is
a standard
Given
these
empirical
properties,
we
first
assume
that
the
dynamics
of
the
logarithm
of
the
credit
spread
X
are
given
by
the
stochastic
differential
equation:
insights
about
the
the
behavior
of
credit
spreads.
The
slope
coefficient
is
negative
in
all
the
regressions
-
and
significantly
negative
in
all
but
one
or
two.
This
indi-
cates
that
the
logarithm
of
the
credit
spread
is
mean-
reverting.
Annualizing
the
slope
coefficient
for
the
regression
implies
that
the
half-life
of
deviations
from
the
long-run
mean
value
ranges
from
about
0.7
to
1.0
years
for
the
industrial
bonds
to
about
1.5
to
4.0
years
for
the
utility
bonds.
Note
that
the
credit
spreads
of
the
lower-rated
bonds
display
less
mean
reversion.
Another
interesting
implication
of
the
regres-
sion
results
is
that
the
logarithm
of
the
credit
spread
is
more
volatile
for
the
higher-rated
bonds
than
for
the
lower-rated
bonds.
For
example,
the
standard
error
of
the
regression
for
the
Aaa-rated
industrial
bonds
is
0.4229,
while
the
same
measure
for
the
Baa-
rated
industrial
bonds
is
only
0.1560.
A
similar
pat-
tern
holds
for
the
utility
bonds.
Histograms
of
the
monthly
changes
in
X
indicate
that
the
distribution
of
changes
is
generally
well-approximated
by
the
nor-
mal
distributio