Documente Academic
Documente Profesional
Documente Cultură
BIOGRAPHY: Mr. Feldblum works at the Allstate Research and Planning Center in Menlo Park, California. He was graduated from Harvard University in 1978 and spent the next two years at a visiting fellow at the Hebrew University in Jerusalem. He became a Fellow of the CAS in 1987, a CPCU in 1986, and an Associate of the SOA in 1986. He currently serves as President of the Casualty Actuaries of the Bay Area, Vice President of Research of the Northern California Chapter of the Society of CPCU, and as a member of the cAS Education and Testing Methods Task Force. Previous papers and discussions of his have appeared in Be&z's Review, the CPCU Journal, the Proceedings -of the Casualty Actuarial Society, and the Actuarial= In addition, he submitted one of the winning papers to the 1988 Insurance Data Management Association call paper program. ABSTRACT : Asset-liability investigated by and non-life immunization Property/Casualty matching, long known to life insurers, is currently being Several crucial differences between life casualty actuaries. modification of traditional insurance operations require when applied to and duration matching techniques insurers:
terms; nominal expressed in 1. Life insurance liabilities are Inflation Property/Casualty insurance liabilities are inflation sensitive. sensitive liabilities are similar to short duration assets with regard to the effects of interest rate fluctuations. bonds provide 2. With a normal, upward-sloping yield curve, short duration lower returns than long duration bonds provide. In other words, duration matching with fixed income securities would reduce investment returns for Property/Casualty insurers. with inflation, in 3. Common stock prices directly expected vary which vary inversely with expected contradistinction to bond prices, inflation. Thus, both coxrumx stocks and Property/Casualty insurance liabilities are inflation sensitive. 4. Property/Casualty insurers do not segment funds, expect a steady premium inflow, and do not face disintermediation problams. Losses and expenses are generally paid from insurance cash flows, not from liquidation of assets. 5. On statutory financial statements, fluctuation. But when marked-to-market, amortized bonds show little price long-term bonds are risky assets.
A superficial acquaintance with asset/liability matching theory would indicate a short duration fixed iip!t ,,I'? :z,et portfolio to back a Property/Casualty insurer's liabilities. .% CS..; ~:nr&al consideration of the factors listed above indicates a portfolta. 1.f o+ :.;::I oecnrities and long-term bonds. ** ---------l * I am grateful ',.n :B,~j.;~..~;~ :--!fkowitz and Richard Wall for extensive ,a1 ?.bis Paper. corrections to an excl..~.-. The rem,%.i.ning errors, of course, are my own. 117
3SSETLIABILIT!fHATCRINGPORPROPRRTY/(ZA5JWTy
INSURERS
matching has long been known to life Casualty stability actuaries as well. now wonder whether
insurers it can
changes, life
influence techniques
ctandard he applied
of immunization
and duration
matching
differently
to Property/Casualty
companies.
growing
matching? returns
If
a life
and stable,
investment pricing
are steady,
on internal rate8
change.
insufficient
policy
insureds
higher
investment losses
returns
cash neede.
scenarios
operating
Life
insurers
have responded
in three
ways:
inr....:xs
are pi-omoting
term policies
of permanent policies.
reduces
interest
rate
risk.
are
and
to the policyholder
(3)
insurer6
are
to mitigate
the
fluctuations
Characteristics
of Property/Casualty
fnaurers:
coverages,
insurers
date. life
coranercial Malpractice,
in
Products
Medical Peril
about four
after the
Moreover,
the investment to
on the assets
For
supporting these
can not
be shifted
reasons,
incurers
investment
portfolios.
canpaniee immunization
have
begun
to High
the rates
life
insurance loss
techniques.
and slow
patterns
strategy
cruciaL
%he characteristics
Property/Casualty
insurance
operations: 119
(1) Traditional nominal pension 65. date. ultimate inflation the market inflation terms.
life
insurance
in or
For instance,
fund to pay $500 a month for Casualty Inflation liability. sensitive. values rates fall. of loss obligations, the
the insured's
at age
are determined
between In
dates
influences liabilities
other
increase,
is
bonds decline.
The reverse
(2)
involves
holding
an
asset
portfolio
whose
means short
whose returns
inflation,
But with
normal,
upward-sloping assets,
returns
than do the
bonds, portfolio
of immunization
and financial if
analysts
ascribe only
long durations
to connnon from
seems true
resulting
and expected
dividends,
changes. or expected
in interest
depresses a rise
of a few weeks to a
in interest
120
insurance the
reserves
are similar.
market
fluctuations. risks
insurance
are contagion
interpretation
insurers face.
do not face the same disintermediation Regardless expect a steady of interest of rate premium Investment any given
Property/Casualty Moreover,
insurers
stream
Property/Casualty for
insurers
of
asset
risk
often
concentrate
on nominal
returns.
For
amortized for
except
Property/Casualty term
bonds are risky
measure,
long
assets.
insurers
balancing
act.
of
reserves income
directly
securities rates
on insurance
returns in
for
of business. risky
insurers
long-term
which are
when marked-to-
yet insurers
to the fluctuations
of the insurance
underwriting
is estate.
into
equity
holdings considerable
market values
investment
the Property/Casualty
industry,
of admitted
assets.
market values,
instead
of
would rules
more heavily
common stocks.
influence security selection as much as operating income does.' ___________________----------------------------------------------------------1 Federal income tax laws also influence Tax law financial portfolios. changes affect asset holdings in ways that asset/liability matching theory does not recognize. The 1996 federal income tax modifications provide several example5 : (A) The tax rate on long term capital gains used to be lower than the rate on net investment income. As a result, growth Current law taxes both equally. stocks have lost their tax advantage over income stocks and corporate bonds. The reduced tax exemptions for municipal bond and corporate dividend reduces the tax advantages of these securities over corporate bonds, federal bonds, and growth stocks.
income
(9)
(C) The strengthening of the Alternative Minimum Tax rules reduces the tax advantages of municipal bonds and income stocks over corporate bonds, federal bonds, and growth stocks. Asset selection depends on current tax law as much as it does on the liability portfolio. Nevertheless, taxes are ignored in this paper, because tax laws 122
TI.NCX4INAL'JERSUSINF'LATIONSENSITIVELIABlLITIES
matching
theory
is
grounded
on
two valued
characteristics liabilities
of and
insurance First.
policies: life
nominally insurance
traditional
liabilities
but are funded at least rate used for pricing is sufficiently little years, in their risk.
by investment
is "conservative'! returns
it
investment returns
- interest
have forced
Consequently,
interest
on life
profitability.
Second,
many
actuaries risks
that
life this
insurers
faced
strong it
Although underdogs,
insurers
accepted
further
discussion
to Property/Casualty reserves
and policy
do not accumulate.
insurer
liabilities
sensitive
insurer may issue a $100,000 policy for a level net premium of $1,000 per ----------------------------------------------------------'-------------------change frequently and unexpectedly. This paper concent r&.$6 on the continuing characteristics of the property/Casualty industry. However, actuaries must be cognizant of factors such as tax treatment and accounting regulations when providing financial investment recommendations. 123
annum.
Policy to
exclusions
(e.g.,
a sufficient
risk
eliminate portfolio
contagion of
factors
insurance insignificant
homogeneous
insured6
rate
risk
remains.
The $1,000 net annual premium may have been on policy reserves. If interest rates
loss.
the
converse
can also
occur: issued
a drop
in
interest
rates certain
may on
the annuity
Moreover,
bought a lo-year
10% annually
interest
rates
fell 1,
to 1988,
5% per
would
1987?
issued
on January
5% coupons. its
old bond, which has 10% coupons, Its market price discounted
of the
cash flows
at 5%, or
Price
$100
$100
0.952
$100
0.9522
$1,100
0.952=
$1,455.
In other words, the insurer pay $1,100 to the annuitant, insurer can lose its as easily liabilities
Could
sell
the bond for $1,455 on January the remaining from interest nominal terms, $355. rate
and pocket as it
Of course, changes.
can gain
in
reason is that
but its
investment
interest the
rate market
changes. values
When a change in new money of liabilities and The greater assets the
the greater
the interest
this
is a speculative rate
risk,
changes. rate
interest
asset/liability
A conservative
to match
more closely.
Hatching
Techniques:
Two common methods of matching and duration portfolio insurance securities matching. Exact to interest from its
assets
are cash flow matching creates an asset/liability forecasts fixed net income at the
estimates
such that
Exact cash flow matching may closely better yield. &.hrih liability :C:?erest
and costly.
One bond a
not simply
a possibility
of loss.
only and
of risk
reduction
expenses.
This is rarely
the case.
Duretion Nat-:
Duration
small interest
rate
changes. First,
on bond prices. rate. When interest When interest the rates bond's fall.
at the
returns. Second,
coupons when
lower
price
declines
interest
and rises
when interest
If
a fixed
j.nccme security
is used to fund
a nominal returns
then
the
depends on If a is
of $1,100 is due at the end of the first of coupons. change. The only change in the bond's used above,
In the illustration
the insurer
$355 price
increaee.
At a bond's
expiration
is received.
Interim rate
price do
changes
a See, for example, Martin L. Leibowits and Alfred Weinberger, %ptimal Cash Plow Matching: Minimum Risk Bona Portfolios for Fulfilling Prescribed Schedules of Liabilities" (New York: Salcam Brothers, Inc., n.d.). 126
influence
the final
wealth.
Suppose a 10 year bond were used to fund a nominal liability ln years. together If with interest rates
of $2,594 due in
remain at 10% per annum, the $100 annual coupons, of the bond, accumulate rates fall to $2,594 at the
accumulated
coupons plus
to $2,258.
The insurer
between the
nominal liability
There is one date when the change in the bond's market value change in its reinvestment returns , as illustrated in Figure
just 1.
balances
the
0 -200 -400
10 year 10% annual coupon bond; interest rate6 decline -------------------------------------------*---------------------------------Thus, if the liability inaurer
1066S,
payment date is one year from purchase it is ten years five from the
years
gains. If it is
If
approximately even.
from the
of purchase,
insurer
just-breaks
127
Hacaulay Duration:
and negative
effects duration
of a a of the bond.
change in Formally, is
is the "Macaulay"
or liabilities,
a 10 year 10% annual coupon bond issued on January 1, 1988. of the $100 coupon payments 1, 1997. figure each January values for yieLds 1, and the
consist
repayment
on January yield;
The present
2 illustrates
--_-___----------_-----------------------------------------------------------Figure Date 1,/l/88 l/1/89 l/l/90 1/l/91 l/1/92 l/1/93 l&J94 l/1/95 l/1/96
l/l Jw
2: Bond Duration Present Value at: 5% ) 95.2 90.7 86.4 82.3 70.4 74.6 71.1 67.1 64.5 675.3 $1,386.20 Principal Repayment 0 0 0 0 0 0 0 0 0 1,000 10% $ 90.9 82.6 75.1 68.3 62.1 56.4 51.3 46.1 42.4 424.1 $l,OOO.OO
Coupon Payment $100 100 100 100 100 100 100 100 100 100
Total
(10 year 10% annual coupon jond; current yield rates of 10% and 5% per annum.) ---e---s ------_-__.. c ,, .-__.. -. _-----------------------------------------------128
At a 10% current
yield,
is
(l"90.9
+ 2'82.6
+ . . . + 10'424.1)
= 6.76 yr6.
At a 5% yield,
the duration
is
(1'95.2
+ 2*90.7 + . . . + lO"675.3)
= 7.29 yrs.
liability
of $1,905 due in
of
two bond6
is
the
weighted
average
of
their
individual
and liabilities
by equating
(overall)
duration
by examining
or by computing fixed
of the durations
policies. the
income securities
whose overall
duration
matches
of the liabilities.
Duration interest
matching
eliminates
asset/liebility
mismatch
risk
only
for
small as
rate changes. 3.
change too,
shown in Figure
----e-m
-----------_-----------------------------------------------------------
3 More precisely, the weight6 are the sum6 of the discounted values of each bond's cash flows. Barring marketplace imperfections, these are the market prices of the bonds. 129
--_---__-____--_________________________-------------------------------------Figure Current Yield 3: Bond Duration versus Current Yields (10 year 10% annual coupon bond) Duration Current Yield 7% 6 5 Duration 7.07 yrs 7.17 yrs 1.27 yrs
10% 6.76 yrs 9 6.86 yrs 8 6.97 yrs -___-------__--__------------------------------------------------------------Even when the interest tnterest. its rate, rates asset they and liability portfolios
have the
same duration
at one
durations giving
at another.
Fortunately,
generally
the insurer
time to "rebalance"
asset portfolio.4
Duration
Hatching
for Property/Casualty
Insurers:
If duration
matching
insurers,
insurers by line
To find
of business.
company reports
or from statutory
Some analysts
Durations of whole life policies generally exceed the durations of longterm bonds. This complicates duration matching. the Moreover, rebalancing asset portfolio may be difficult when interest rates change sharply. To solve these problems, some financial analysts have suggested taking long positions in futures and buying call options to lengthen the duration of the asset portfolio. In practice, however, unless there are other reasons for investing in futures and options (such as hedging), the additional cost of these transactions outweighs the gains from more accurate duration matching. Few actuaries examine the costs of asset/liability matching. Fewer sti1.1 seek to balance these costs against the benefits of a more stable operating income and the reduced risk of insolvency. See below for further discussion of this. 130
Richard Woll has shown how to estimate business for from Annual Statement P lines data.6
by line
of
of business
payouts reserve
decay mode1.7
durations
by discounting
at an appropriate
calculation Part
of liability
"durations"
by line payout
of business. for
to forecast a paid
patterns
each accident
analysis},
the percentages
shown in Figure
Management Strategies example, Peter 0. Noris, "Asset/Liability & Casualty Companies," Morgan Stanley, May 1985.
6 Richard G. Wall, "Insurance Profits: Keeping Score," in Financial Analysis of Insurance Companies, CAS 1987 Discussion Paper Program, particularly pages 505-514 and 522-523. -J The exponential decay assumption simplifies the mathematics, but it is not crucial for the estimates of reserve duration. Any reasonable model, such as simply dividing the remaining loss payouts equally over the next five years, produces approximately the same durations. The exponential decay model is described by Charles A. McClenahan, !'A Mathematical Model for Loss Reserve Analysis," Proceedings -of the Casualty .-Actuarial Society, Lx11 (1975) pp. 134153.
cit., p. 511 (Exhibit IX, Page 1). Wall's exhibit shows The numbers in Fi.gn-: ! are the first differences of 131
____-__________--___---------------------------------------------------------Figure 4: General Liability Loss Reserve Payout Pattern Development Year Accident Year
1983
10
1984 1985
16.1% 15.3% 14.6% 11.9% 9.2% 7.0% 16.4 16.3 14.3 11.7 9.1 6.8 17.3 16.1 14.2 11.6 8.9 6.8
5.3% 4.2% 3.3% 2.6% 5.2 4.1 3.3 2.5 5.1 4.1 3.2 2.5
Notes : Since these are loss reserve payout patterns, the first development For example, year is the first calendar year subsequent to the accident year. development year #l for accident year 1985 is calendar year 1986.
The numbers shown are the percentages of required reserves for each accident year that are paid in each development year. For instance, 16.1% of required General Liability reserves for accident year 1983 were paid in 1984; 15.3% were paid in 1985; and so forth. The percentages for calendar years 1986 and onward are estimates based on historical loss payout patterns. -----------------------------------------------------------------------------The timing of loss payments within the tenth calendar each calendar year, year, as well as the pattern effect on the
loss payments are made at mid10 years year are paid out evenly
and that
years.held after
are still
We therefore
11 through
P The exponential decay model used by Richard Woll is more accurate. However, the loss payout patterns are then more difficult to calculate. The simplified method in the text is sufficient for the heuristic purposes of this paper.
Similarly, the average General Liability 1066 payment is not made midpoint of the policy year or of the succeeding years. Rather, the payment date is c!.ose to the end of the policy term during the initial and it gradually recedes towards the midpoint of the policy term development years continue. This phenomenon has little effect arguments in the text, and it unduly complicates the mathematics. 132
In
new issues
of
grade Since
Aaa corporate
bonds were
insurers'
the reserve
"duration"
is therefore
16.1*0.5*(1/1.12)"-5 + 15.3*1.5*(1/1.12)'.5 + 15.3 * (l/1.12)1-5 16.1 * (1,/1.12)-5 or 3.2 years. This figure is substantially
actuaries and analysts for liability reserve "durations."'1 _________-__^____---____________________-------------------------------------10 Richard Wall uses a 5.9% interest rate for 1983, based on short term Treaa1x-y bill rates. The risk free rate is appropriate for discounting loss For duration reserves, thereby separating insurance and investment returns. matching, however, we must use the same interest rate for the liabilities as we use for new investable assets.
11 For 1983, Richard Woll calculated the General Liability loss reserve Schedule P loss reserve duration as 2.1 duration as 3.2 years, the overall years, and the all lines loss reserve duration as 1.95 years; see Wall, E duration of 2.5 year6 for a, p. 523. Peter Noris obtained a 1983 liability Automobile Physical Damage, an insurance portfolio of Automobile Liability, Workers' Compensation, Multi-Peril, and General Liability, weighted in the same proportions as the overall industry portfolio (see Noris, op. cit., pp, 8 and 26). This is longer than Wall's 2.1 years for all Schedule P lines combined, since Noris assumes slower payout patterns. For instance, his General Liability payout pattern assumes that only 32% of losses are paid in the first three years. Wall estimates that 4b% of initial General Liability loss reserves are paid within the first three subsequent years. This discrepancy results from the type of loss payout pattern used by each analyst. Noris uses the loss payout pattern for losses incurred during a specific accident year. In agreement with Noris, Woll estimates that 31% of General Liability losses are paid In the first three years. But for liability durations, Wall uses the payout pattern for loss reserves, not for incurred losses A The +.wo types of loss payment patterns are quite different. Many losses are paid during the year of occurrence, and do not appear as reserve liabilities on the year end accounting statement. Those losses that remain outstanding as of December 31 as generally the slower settling ones. Thus, the loss reserve payout pattern iS usually slower than the -incurred loss payout pattern. ----Thus, 133
conclusion
is
that
to
duration
match
a General
Liability an average
Interest
Rate Changes:
is misleading. changes.
Asset/liability
the risk
of
effects
of interest
changes on insurance
--_-__-___-_____-___----------------------------------------------------------
Richard Woll shows an average payment date for all lines combined of 1.72 years for incurred losses and 2.33 years for loss reserves (see Wall, c cit., p. 510, Table XIV). Wall shows the opposite and Medical Malpractice, For General Liability relationship: significantly longer average payment dates for incurred losses than for loss reserves. Perhaps this is due to differences in the loss dates in the two calculations: loss reserves include losses from older accident years, while incurred losses are from the current accident year. The implication in the past, would be that General Liability and Medical Malpractice claims were settled more quickly, but average payment dates have lengthened in recent years. However, Wall's Exhibit VIII on page 506 does not support this hypothesis: average payment dates have remained constant for these two lines of business between 1977 and 1985. For further discussion of this topic, see Wall, op. cit., p. 513.
The proper payout pattern depends on the type of matching. If one matches assets held as of December 31 with liability obligations aa of the same date, then one should use the loss reserve payout pattern. If one matches assets purchased with liabilities incurred, then one should use the loss incurred payout pattern. The discount rates used by Noris and Wall do not differ much, and do not account for the duration discrepancy. Noris uses the municipal bond rate to estimate an after-tax return. Woll uses an after-tax short-term Treasury bill rate to estimate the risk free return. These two rates are close enough that their difference has little effect on the liability "duration." 134
Liability
loss
reserve inflation
that
will
be
paid
$100,000.
suppose that
but are
Liability inflation
rate.
For simplicity,
liability,
the insurer
invests
$62,092 in five
yield
What happens when interest per annum shortly and it after remains inflation,
rates the at
change? is
accelerates
purchased,
the next
Bond yields
ultimate matures
loss is now $125,000 ( = $100,000 * (3.15/1.10)5 for $100,000 in five the risk years. rate Duration changes. matching
).
But
has not
mitigate
of interest
I2 The assumption that the bond yield equals the loss cost trend is made for simplicity only. The three assumptions that underlie the argument are: (1) Inflation affects General Liability date. (2) General Liability loss cost trends (3) Bond yields vary with inflation. loss payments through the settlement vary with inflation.
2.3 The term of a zero-coupon bond equals its duration. For the illustrative bonds in the text, both the duration and the term are five years. 135
Inflation
Sensitive
Cash
Pbw6:
the
loss fixed
reserve income in
matching life
expressed insurance
(a) balances
and investment
the insurance
The
cash
f 'lows If date
General losses
Liabil are
ity
losses, to
however, inflation
are
inflation the
sensitive. settlement
through
no lag between
and its
effects
on losses)
is equivalent
a duration
that
inflation
(e.g.,
common stocks
and real
In practice, settlement
are
not
fully
inflation
sensitive
through
Compensation Bodily
indemnity Injury
payments are largely wage loss a year which and medical or two before
a+ the accident may be determined settlement insurance iM!prac,:ica, Injury, -e--w... date.
Automobile
- often damages,"
payments
General
Products coverages,
Medical Bodily
Commercial Multi-Peril
and Automobile
important
losses
through
the
such as "pain of
money at
and
the
influenced date.
by
Medical
the accident
and settlement
usually
and expected
changes in the demand for are inflation reserves, are paid out sensitive; on the quickly,
assets.
sensitive.
however,
equally short durations." _-_--__-__-_---_-_-----------------------------------------------------------14 For a more complete discussion of the timing of inflation on insurance "The Effect of Inflation on Losses and Premiums losses, see Robert P. Butsic, for Property-Liability Insurers," Inflation Implications for Property-Casualty Insurance, 1981 Casualty Actuarial Society Discussion PaprProgram, p. 51.
When the insurance payments indemnify economic losses, such as work disability and medical bills in automobile personal injury claims, the "loss date," or "treatment date I1 should replace the accident date in Butsic's model. If the loss date is c&erminous with the accident date, this revision has only a minor effect. Sometimes, however, the lVloss date" or "treatment date" is closer to the payment or settlement date than to the accident date. For instance, suppose a motor vehicle accident victim in a no-fault compensation state suffers an injury requiring extended medical treatment. The medical bills may continue for years after the accident, and the insurer will reimburse the victim soon after the treatment. Butsic notes that both interest rates and loss cost trends vary closely with inflation. Butsic's reviewer takes him to task for this, claiming that interest rates and inflation are not as well correlated as Butsic implies. In the short term, this is correct, since numerous factors besides inflation affect interest rates. Over the long term, however, interest rates do vary closely with inflation. One may infer, as both this paper and But&c's paper expected interest rates are directly correlated with expected do, that inflation. 16 Steven D'Arcy makes the same argument in his excellent review Ferguson's "Duration" in the Proceedings -of the Casualty Actuarial 137 of Ronald Society,
paper If
has
a duration rates
similar increase,
to
that
of
interest
payment
Reinvestment similarly.
returns
from the
maturity is little
payments increase
change in market price. and liabilities and connnercial (GL losses) paper yields
paper) trends
the magnitudes
This
is
for
investments
in
commercial
Short-term costs
sections.
for equity
_______-____________----------------------------------------------------
--em--
Peter Noris also notes the inflation sensitivity of LXX1 (1984) pp. 8-25. insurance losses, but he believes that conservative reserving obviates this problem; see Noris, op. cit., pp. 43-45. However, conservative reserving and asset/liability matching are different types of solutions to the general valuation problem. The former says, "Make reserves sufficiently redundant so that they will be adequate even under adverse conditions." The latter says, "Keep reserves accurate, but choose matching assets 60 that changes in conditions affect the two sides equally." When conservative reserving is used to guard non-financial against uncertainties, it may be used even with asset/liability matching. For example, American courts have interpreted the pollution exclusion in General Liability policies in diverse ways. The insurer may set up a large bulk reserve for such "extra-contractual" liabilities that the courts discern in the policy and impose on the carrier, This is unrelated to the asset/liability matching problem. I6 The similarity in magnitude assumes only that the yields change by the the same amount, not that they be equal. General Liability loss cost trends in the 1970's and 1980's have exceeded commercial paper yields. If inflation accelerates, however, the changes in trends and yields may be about equal. 138
III.
EQUITY DURATIONS
section
distinguished
nominal
Securities insurance
and annuity
averages values
vary
insurance matching.
Macaulay durations
for asset/liability
This is the crux of this the effects in market interest of interest value rate. is
section. rate
durations
proportional
change rate
Long-term
to interest
changes
Figure five
5 illustrates
this
phenomenon. respectively.
Consider
with and
= $386, respectively.
rates
point
the day
of these bonds.
139
___-----______--____---------------------------------------------------------Effects of interest Figure 5: rate changes on market values Market value: Zero-coupon bonds at 10% at 9% f Change % I at 10% at 11% $593 352 Change d % -4.5% -8.8
+4.7% 1 $621 5 year term: $621 $650 $29 +9.3 1 386 36 386 10 year term: 422 --------___---_-__-_---------------------------------------------------------A zero coupon bond's and 10 years, duration equals its term.
$28
34
respectively. that
Ccamm Stocks:
How should
the
duration
of
common stocks
be measured?
Stated
differently,
17 Mathematically, the change in price equals the negative of the Macaulay duration times the market price times the change in the interest rate, or
Change in Price
= -1
Duration rates
Price
Change in Interest
Rate. in
from
Ten year bond: -1 * 10 * $386 l 0.01 Five year bond: -1 l 5 * $621 * 0.01
These figures differ from the actual market price changes because the change in the intereat rate is not "infinitesimally" small. For a n-ore extensive treatment of this subject, see G. 0. Bier-wag, George G. Kaufman, and Alden Toevs , "Duration: Its Development and Use in Bond Portfolio Management," Financial Analysts Journal, July-August 1983, especially pp. 17-18.
of
cormnon stock
duration
helps
quantify
the
effects
of
measurement of equity
of
duration
use6
the
"dividend as a are
valuation.
model views
bond that
grow
pays dividends
an infinite
term.
are discounted is
at K% per annum,
(current
dividend)
(1 + G) / (K - ,).I,
of a fixed logarithm
equals value
Using this
equation
the duration
- d (ln
(current
(K _1G)
rate
of 10% implies
dividend
t=t where "t" is the number of years since the purchase formula in the text. For further exposition, Finance, eighth edition formula, *KS represents growth rate.
current
dividend l (1 + IQ=
see J. Fred Weston and Thomas E. Copeland, Managerial (Chicago: The Dryden Press, 1986), pp. 609-X). In the the cost of capital and "G" represents the dividend
19 See, for example, Martin L. Leibwitz, et al., "A Total Differential Approach to Equity Duration," (New York: Salomon Brothers, Inc., n.d.), p. 3. Leibowitz similarly concludes that traditional measures of equity duration give meaninglessly high figures. 141
a duration duration
of
25 years
for
the
average bonds.
common stock.
This
exceeds
the
of even long-term
corporate
Market Price
Changes:
to
prices prices
rates strongly
change? in the
A long opposite
implies
of the change.
is not true
Tnterest ways.
rate
and inflation
rate
changes affect
in several
1. ---value should
In theory, by
the real
value If
of the firm's
major
inflation.
inflation
accelerate, its
should increase
inflation-adjusted
remains constant.
revenues
and interest
supply
costs is
If inflation
is V'demand-pull,Vt there
is "supply-push," households
Xoreover, further
rates
encourage
142
inflation
rates
increase
In such cases,
of the firm
3. Investment holdings
strategy:
When interest
rates
rise,
investors in"
often
shift
their
to long-term
bonds, to "lock
reduces their
market prices.
The first
effect
In other
words, prices
when inflation
decline
common stock
at first,
later.
this
inflation term
rates government
with
(a)
cormnon capital
bonds
appreciation. The correlations below used 1962-81 annual returns.Z0 ------_------_---------------------------------------------------------------20 Annual Singuefield, (Charlottesville, Ibbotson and appreciation
and monthly figures are from Roger C. Ibbotson and Rex A. Stocks, Bonds. Bill6 and Inflation: The Past and the Future _---Virginia: Thxanzl Analysts Research Foundation, 1982). Sinq-uefield do not provide a long-term corporate bond capital series.
Ibbotson and Sinquefeld's series begin at 1926. The 1929 depression, as well as the post World War II economic prosperity, caused major fluctuations in common stock prices, which overwhelmed the effects of interest rate changes. The 1962-1981 period is more representative of current conditions. This paper uses the Consumer Price Index (CPI) as a proxy for new money interest rates. An alternative series available from Ibbotson and Singuefield is the Treasury Bill total return series. However, this lags inflation by up to half a year, and so it is less useful for the correlations. These series are used only to illustrate the argument in the text. To accurately determine the correlations among new money interest rates, longterm bond prices, the following adjustments would be and conrnon stock prices, needed: (1) New money interest rates should issued high quality corporate bonds. be measured as coupon rates on newly
143
coefficient series
between
the
long
term government
bond capital
index is -50%.
of long-term
strongly
(and by implication,
If
the long
term government
bond capital
series negative
is
lagged
one
year,
is still
from the
decline results
the reduction
in the correlation
coefficient their
par values
coefficient
between index
the
common stock
capital
accumulation
is -19%.
of common stocks
inversely
However,
if
accumulation = +17%).
series
is lagged an
the
correlation
Presumably, equity
acceleration
of inflation
most firms,
and their
values
After several months, however, firms adjust their costs and prices, ----____-_______-------------------------------------------------------------(2) Smoothed monthly
or quarterly
rates
(3) One year lags should be replaced by lags of differing on the ease with which the firm can pass on cost increases. (4) Interest rate changes should be examined in conjunction values of the interest rates. See the discussion in the text for elaboration 144 of the third
vary directly
with recent
inflation
rates.
are sensitive
to inflation.
are
Why does the traditional traditional dividend change, growth calculation payments the rate dividend will in
duration
formula
give
The
dividend if
inflation
dislike
modifying rate,
the dividend
Once a firm changes the dividend model" rate. valuation of the firm will
though,
as the inflation
of
inflation
prices
of
a retail
increase
accelerates.
nominal rates.
value
in the
as
A municipal
changes to
department.
Unexpected of utilities."
should
be inversely
correlated
How, then,
should a Property/Casualty
insurer
divide
its
assets
among
comon
inflation affects insurance losses only through the accident date, not the settlement date, then trends in paid losses also lag economic It is not clear whether the lag is greater for the overall stock However, portfolio or for the industry-wide insurance portfolio. lag is large enough to affect the conclusions in the text. et al., "Total Differential," 145 a cit., make a similar
Leibowitz, argument.
stocks, bills
long-term
bonds,
and short-term
bills?
have a similar
a0.
relationship
to interest
changes as liability
Long-term
bonds have long durations changes. bills loss stock this risk. Common stocks do, allowing cost trends.
rate
short-term liability
connsensurate insurer
market
risks.
Short-term long-term
bills
Finally,
bonds entail
section costs,
examines
these
investment
expected
yields, is
cash flows,
matching strategy.
but it
-----------------------------------------------------------------------------Up until the 1970'6, institutional investors were not active traders of stocks and bonds in the secondary markets, so transaction costs were low. During the past decade, institutional investor trading has grown markedly, particularly in common stocks. 146
13
Iv.
matching securities.
theory
tells
insurer
For example,
bonds. shortening
managers
bond portfolio
in a reduction
there
in investment a steep,
is
typically
Expected Yields:
this
is
a major
concern.
A mismatched than
portfolio portfolio
a matched returns.
expected
Shortening investment
generally
differences offer
High quality
long-term
corporate bills
bonds and
Treasury
commercial with
paper
a Guaranteed l-year
an investment
types
of bonds:
?.5%, 2-year bonds yielding 1.X%, and 3-year bonds yielding _----------------------------------------------------------------------------* Noris, op. cit., p. 33.
147
that percent
"to
the
interest
rate
risks
associated annual
with
guaranteeing privileges
7.65 without
years
withdrawal
. . . II the insurer
can invest
only half
one third
interest
Transaction
Costs:
costs particularly
are equally if
important.
Long-term
bonds entail
the least
to maturity. investment
greater individual
trading
equities. of new
monitoring portfolio.
interest
Disintermediation:
What risk does asset/liability matching guard against? Answers like -----_------_-------_________________^__-------------------------------------IS
the
"the risk
James A. Tilley, "The Matching of Assets and Liabilities," Transactions of Society of Actuaries, XXXII, p. 293. Note also the fourth criticism 2 oonventional-bnization theory that Tilley cites: "Ieununisation theory in its conventional form places rigid constraints on investment operations, leaving investment officers with too little latitude for making policy decisions" (p, 264). Tilley believes, however, that the resolution of this problem is "largely a matter of education." 148
of interest rate
How do interest
changes affect
Property/Casualty
Life fall
answer:
rates yield in
decline, the
investment life
returns
guaranteed rise,
whole
policy.
rates
insured6
forcing
losses.
This
is misleading. that
interest chance
rates that
policies exceed
are so
And when new money interest insurers of clear. the receive extent investment of
rates
rise
return
regardless
disintermediation.
Consider annuitant
annuity
certain
with
loan
provision.
The promises
$1,000 with
the insurer
to pay $1,040 on December 31, but it the year, for a net profit of $10.
expects
may withdraw
of the annuity
to 6% interest
per annum.
149
rates
increase
from 5% to 8% on January 2, the annuitant the money at the higher annuitant not take the rate. loan,
would prefer
the contract of
its
exceeding
What about the capital loss simply allow occurs forces before
on January
l?
The capital
the annuitant
realization
of the capital
loss.
amortization
does not appear on the insurer's net income that statutory insurer results
income statement.
of statutory of the
accounting requires
and annuitants
investments
significant adverse effect on economic worth.26 ______________-________________________c-------------------------------------26 For participating policies, the issue is more complex. Suppose the premiums on a participating whole life policy assume 2% interest, new money interest rates are 9, and the policy loan interest rate is 6%. Suppose further that new money interest rates increase to 10% soon after the policy is issued. The insurer intends to return most of the difference between actual and assumed interest returns through policyholder dividends. If more than 4% of the interest rate differential is returned to an insured who has taken a policy loan of the entire cash value, then the insurer has a negative interest rate margin on the policy.
One solution to this problem is to vary the dividend payments by the extent of the policy loan. The full dividend is paid if the policy cash value remains intact, while a lesser dividend is paid to insureds with policy loans. Critics of this practice argue that obtaining a policy loan is a statutorily guaranteed right, no less than the beneficiary's right to receive the face value uyi)'i death of the insured. The insured's illness does not affect his policyholder dividends, even though it shortens his expected life and therefore reduces the insurer's mortality margin. So also a loan should not affect policyhol?er dividends. For further discussion of this, see R. C. 150
For
Property/Casualty
rates
insurers,
the
risk
is
different. return
interest
decline,
investment operating
return.
decline
income increases.
When
new
money
interest
rates
rise, return.
investment
return rise
below the new money investment the expected difference duration portfolio, rise. levels,
so net operating
income decreases.
the duration
the reduction
in net operating
income as interest
Interest
rate
in interest long-term
reduces
though it
and short-term
bonds ana
securities realize
the insurer
must sell
long-term
the capital
How likely is this to happen? Many Property/Casualty ------------------__---------------------------------------------------------Winters, "Philosophic Issues in Dividend Distributions," .cOCiety of Actuaries XXX, pp. 125-137. I51
insurers Transactions
do become of the --
is In
degree
are is
insolvencies the
caused
by that
words, that
"What they
probability an insurer
so dramatically
force
to sell
insolvent?"
plan
funding
vehicle
account
to back
its
from employment than The investment account block before manager their
than expected.
assets
account
of liabilities. direction
insurer
has a single
asset
account.
The insurer
holds income
bonds to maturity,
becomes insolvent
of other
such as inadequate
pricing
an insolvent insolvency
high
and the
risk
of
significant invest
long duration
152
Risks:
This
question
is
not its
related
to
matching. expose
of to
security systematic
presents
own risks.
market risks,
bonds, inflation
if
expose
an insurer
to
the
risk
of
The investment
these two
in the financial
Conclusion:
reserves similar
are
inflation
Commercial paper and Treasury but their and similar all but the expected yields are
investments
regulation managers.
too risky
investment of choice.
Cormnonstocks,
therefore,
s7 In truth, there was no 'm:::?':zt crash" in 1987, because the market returned about 5% over the year. 'The wression of a crash resulted from the sudden readjustment in October 1987. The 1987 experience differs from a true market collapse in that no econrwL.,: downturn followed October 1987.
1.53
rate
risk
for
stable
Long-term higher
a duration
income securities
actuary
analyst
must be aware of the types and liabilities. The traditional ensuring safety
and inflation
insurer's
assets
concern. returns,
of securities.
154