RECORD OF SOCIETY OF ACTUARIES

1992 VOL. 18 NO. 4A

VALUATION ACTUARY OVERVIEW --

THE NEW STATUTORY VALUATION OPINION

Moderator: ABRAHAM S. GOOTZEIT

Panelists: DOUGLAS C. DOLL
DEBORAH A. GERO
MARC F. PITONIAK
Recorder: ABRAHAM S. GOO'I'ZEIT

• Standard Valuation Law - Overview

- State adoptionstatus
- Complyingwith state variations
- Industry survey
• The asset adequacy analysis requirement
- Actuarial Standardsof Practice
-- Regulatoryabilityto oversee
- Opinion and memorandum
- Personal or company liability
• Practical considerations of cash-flow testing
- Preparation
- Dynamic assumptions
- C-1 and C-2 assumptions
- Modeling techniquesand methodologies
- Example
• What is an adequate reserve?
- How do you determine adequacy?
- How many scenarios are enough?
- Aggregation
-- Gross premium valuations

MR. ABRAHAM S. GOOTZEIT: Our panel includes Doug Doll, a consulting actuary
from the Atlanta office of TiUinghast; Debi Gero, vice president in charge of asset and
liability modeling at Conseco; and Marc Pitoniak, founder and president of ValAct, an
organization providing software and consulting services for asset and liability modeling.

Doug will discuss regulatory issues and the amendments to the standard valuation
law. Debi will then discuss practical considerations, such as accumulating data,
setting assumptions and creating models. Marc will discuss an example that illus-
trates a number of the issues and considerations involved in cash-flow testing.

MR. DOUGLAS C. DOLL: I'm going to cover some of the regulatory items with
regard to the Standard Valuation Law and then talk a little about actuarial liability.
There are 10 states that have adopted the new Standard Valuation Law effective for
1992 statements and five states for 1993. The 10 states for 1992 include some of
the big ones like California, Illinois and Texas. There will be an impact on 1992
statements on any company that's operating in several states.

Only one state has actually adopted the accompanying regulation, and that's Oregon,
but it's proposed in quite a few other states. Most of the regulations that are
proposed are basically the same as the standard adopted by the NAIC. California's

1459
 RECORD, VOLUME 18

regulation is somewhat interesting. The standard model regulation would allow a

three-year period in which to grade up extra reserves. In other words, in the first
year, if cash-flow testing says extra reserves are needed, you only have to put up
one-third. Next year you have to put up two-thirds of the extra, and so forth.
California deleted that provision. You have to put up the extra amount immediately.

There were a couple of changes made to the model regulation by the NAIC at its last
meeting in September. These aren't significant. They replaced the references to the
mandatory securities valuation reserve (MSVR) with the asset valuation reserve (AVR)
and interest maintenance reserve (IMR). You must use the proper allocation of assets
backing the IMR. You may include assets backing the AVR to the extent applied to
asset default, which was the same methodology that was intended to be used for the
MSVR.

They also changed the seven prescribed scenarios. The minimum interest rate is no
longer 4%, Now the minimum interest rate for a five-year Treasury note is ½ of the
starting rate. They also took out the maximum rate. The previous maximum rate
was 25%, so it's a ncnevent to take out the maximum.

They changed the headings for the reserve table in the actuarial opinion. That's no
big deal, but now the headings are intelligible.

Let's turn next to the Actuarial Standard of Practice (ASP) for the appointed actuary
opinion. I'm not associated with the Actuarial Standards Board (ASB). I've talked to
some people and have gotten some comments about what the next exposure draft is
going to have, but I'm not absolutely sure that I'm correct. It could change.

The ASP is going to be broken into two pieces. Section 7 opinions are for small
companies that do not need to do cash-flow testing. The Section 7 piece is not
going to be a standard of practice. It's going to be a compliance guideline. The
essence of that will be that the appointed actuary does not have to do cash-flow
testing if regulations specifically say you don't have to do cash-flow testing.

The Section 8 (asset-adequacy opinion) piece has had some changes made to it. The
changes were deemed significant enough that it's going to be reexposed rather than
be adopted as final this year, Both the Section 7 compliance guideline and the
revised Section 8 actuarial standard of practice are going to be reexposed and will
probably come out in early December 1992. Some of the interesting changes to the
Section 8 ASP are as follows. In the original draft, reserves have to satisfy substan-
tially-better-than-even scenarios. There's been a lot of question about what "substan-
tially-better-than-even" means. The new wording is going to be "moderately adverse
conditions." The ASB was actually considering using "best estimate." Because at
least you know what best estimate means. The next exposure draft is going to invite
comments. Should it be best estimate? Moderately adverse? Or what?

The original exposure draft said that with regard to actuarial guidelines and general
distributed interpretations of laws and regulations, the valuation actuary had to be
satisfied that those requirements were met in reserves. The new version takes away
that requirement and says that you need to be aware of actuarial guidelines and
generally published interpretations. You should disclose deviations in your actuarial

1460
 VALUATION ACTUARY OVERVIEW

memorandum, but there's no longer a requirement that you have to satisfy those in
your reserves.

There are some minor proposed changes in New York Regulation 126. They changed
the section on defaults. In the absence of credible experience, they've suggested
some default rates that you can use. They said it should not be less than 20% of
the AVR maximums. For example, for AAA bonds, that would be a default cost
assumption of 0.1%. They also changed the minimum interest rate on the prescribed
seven scenarios to 0.5% of the starting treasury rate.

The next topic is practice notes. It was announced at the Valuation Actuary Sympo-
sium, that there was a task force set up to develop practice notes. These practice
notes take various topics or issues that the valuation actuary has to deal with, such
as how to handle default rates, or how many scenarios should be tested, and develop
something to give guidance to the practicing actuary as to what would be a reason-
able thing to do. This is a task force through the Academy Committee on Life
Insurance. The task force chairperson is Donna Claire. There are about 10 topics.
These drafts are currently being edited by the Academy, with a view toward making
sure that these practice notes don't have rigid requirements in them, so that the
actuary who does something different than is in one of these practice notes is not
subject to criticism.

Many people are asking for guidance in certain areas. We have to draw that fine line
between doing something that will be of help to actuaries versus doing something
that the actuary will be criticized for if he or she doesn't follow. If problems develop
later, we don't want it contributing to possible actuarial liability.

I already mentioned what the new Actuarial Standard of Practice will say with regard
to state variations. Milliman & Robertson and -I]llinghast have proposed to the AAA
that they work on a research project to tabulate all the state variations on laws and
regulations and generally published interpretations of those laws and regulations.
We're working on compiling the state variations and actual laws and regulations, the
summary of actuarial guidelines, and a survey of state insurance departments'
interpretations of laws, regulations and guidelines. I hope that we'll have something
on that by year-end.

The next topic is reliance on investment professionals. I think many of us will at least
make certain asset assumptions and maybe projections of certain kinds of assets,
such as collateralized mortgage obligations (CMOs). The model regulation provides
wording for when reliance is used on asset projections and requires a statement from
the person relied upon. In the actuarial memorandum, the ASP would require a
statement of qualifications. This, by the way, is a little bit different than the original
exposure draft, which I think says something to the effect that the actuary had to
provide evidence that the person or firm relied on was qualified. Now it's a statement
of the qualifications.

My next topic is regulatory oversight. At the last meeting of the NAIC Actuarial Task
Force a couple of weeks ago, it was proposed that a group of insurance department
actuaries and other state financial examiners meet sometime next spring, after all the
statements are in, to review actuarial memorandums. The intent is that every state

1461
 RECORD, VOLUME 18

insurance department will submit actuarial memorandums that they want for review.
The purpose of this is to get some uniformity among states. Also, if individual
regulators have concerns or think that a memorandum is not up to snuff, it will give
them some comfort, before they challenge it, to get some other folks to agree with
them.

Finally, what is actuarial liability? It's like an illness. Who gets it? By definition,
actuaries get actuarial liability. How? It is usually contracted by an insolvent insur-
ance company for whom you've signed a recent opinion. How serious is it? Very,
very, very. It can consume all your time and energy. It can drain your assets and it
can ruin your career. Can it be cured? Once contracted, there's very little you can
do to cure it. It's a condition that responds well only to preventive measures. What
are those preventive measures? Do only work for which you're qualified. Know and
follow actuarial standards. Make sure your report is complete and clear; don't
sugarcoat the problem.

What do we have to do with regard to actuarial standards? Think of actuarial

standards as a coloring exercise. When you were a kid, you colored in coloring
books. You need to do two things. Don't make any mistakes; that is, color between
the lines and color all areas that should be colored. So in actuarial standards, do
what they say to do, and be very careful of going outside of what they say to do.

Be careful about three other things. Word your opinion carefully. Assume any
ambiguity in the standards or requirements will result against you. Last, make sure
your errors and omissions or directors and officers (D&O) liability coverage is current.
Performance is going to be measured against the skill and knowledge normally
possessed and used by members of the profession in good standing. Dangers that
we have to protect ourselves against include changes in actuarial standards over time.
Another possible danger is actuarial standards strictly applied. There we have to be
careful that when going outside the actuarial standards, there is a reasonable basis for
doing so.

MS. DEBORAH A. GERO: I'm going to be covering some basics of asset and liability
modeling. Many of you might find that this is more of the same old stuff, but I hope,
through our experiences of having modeled seven different companies, where we
acquired assets from various sources, together with the practices of our current
investment department, we can shed some light.

The first notion I would like to challenge is this perhaps misnomer called "valuation
actuary." The actuary is the one who gets to have all the fun and pull all the data
together and do the work, but really, the actuary is not acting alone. Doug talked
about reliance, and I think even the best actuaries would find it impossible to have all
the experience necessary to come up with the correct assumptions and input
required. We really have, at least in our case, to rely on the investment professionals
to come up with reasonable assumptions and methodologies on the asset side. We
also have to rely on our marketing people to give us feedback. I think it is a bit of the
old adage that the actuary can come up with a product that will never sell, but meets
all the profit criteria of the company. I think you can fall into the same trap when you
are trying to do cash-flow testing. You can come up with some wonderful results,
but your company would, for all intents and purposes, be out of business for new

1462
 VALUATION ACTUARY OVERVIEW

business because you would have a bad track record. You really need to be aware
of the various audiences that will be eagerly looking at your results and the types of
assumptions you make. You cut them off at the pass by going to them first.

The first subject I would like to cover is asset modeling, and the second subject will
be liability modeling.

ASSETS
Asset Modeling
I think that even in those companiesthat have the best asset/liabilitymanagement
systems put in place, the actuaries really have very littleknowledge of the specific
asset details,the nuances and characteristics,of those that are purchaseduntil after
they are purchased, and the actuary has to deal with them. The actuary needs to
investigate, when putting together cash-flow tests, those features that will have a
material impact on the resultsand variousscenarios.

I would liketo cover some of the aspects of existingassets, because as we all know,
there are Schedule B's and D's that containlong lists of assets, but unfortunately,
they are really a poor place for actuariesto get the kindof informationthey need.
Much of the information is also not found on administrativesystems. This, I think, is
especiallytrue, we found, when it comes to dealingwith mortgagesor collateralized
mortgage obligations(CMOs). You may have the basicpar and book value, but
much of the other required informationis missing.

I am briefly going to talk about mortgages, mortgage-backedsecurities,other CMOs,

bonds, floating rate securities,and interest rate swaps. I have not yet figured out
how to projectcommon stock with sufficientcredibility,so I am going to skip that
one.

There are many different features of mortgages and they typicaUy have various
prepayment previsions. You may have a period of time for which no prepayment is
possible. Floating-rate securities have reset dates, indexes, floors and ceilings, and
balloon provisions. You may also have some equity participation, which you might
want to consider in determining how good the collateral is. One of the ways we
have found to deal with all this (because our administrative systems don't have many
of these provisions on them and we have bought companies where their old adminis-
trative systems did not have this information on them) is to use the underwriting
guidelines that may have been applicable for given eras in the past. Mortgages issued
during a certain time period may typically, for a certain term of mortgage, have an
average time for which no prepayment is possible, end some standard indexes are
used. W_h the balloon provisions, you may also now want to look at how realistic it
is that those balloon payments will actually be made.

Mortgage-backed securities are assets that are extremely good for grouping together.
A govemment national mortgage association (Ginnie Mae 8) is a Ginnie Mae 8 is a
Ginnie Mae 8. There may be some specific geographic concentrations you may want
to take into consideration. You also would want to be careful about the original
length of the underlying coUateral and the issue dates.

1463
 RECORD, VOLUME 18

We commonly talk about call schedules with corporate bonds, but many bonds,
especially utility bonds have, in essence, two types of calls. There are calls that can
be made at any point and at anytime for a given prepayment price. There are also
calls that are called "refunds," which cannot be made if the issue is just refinancing at
a lower interest rate. The refunds are the most important for us to review when we
are dealing with callability of bonds in various interest rate scenarios.

The amount of callable bonds that are being issued has been drastically reduced. In
the 1970s and 1980s, 95% of bonds were callable. From 1989-90, about 30%
were callable. As interest rates have continued to decline, less bonds are being
issued callable. This is also attributable to the fact that the market is more sophisti-
cated in pricing the call option, t think the importance of this is that many of us,
when we put together our reinvestment assets, historically look at what spreads have
been available and go from there. Given that spread premium for calls may no longer
be available, you may want to look at some more recent data on recent purchases
that your company has made, instead of taking long-term historic averages. When a
bond is callable varies significantly by the bond and the term of the bond, and you
hope that information will be available to you.

Sinking-fund schedules are, again, not typical on new issues, but to the extent you
have any old issues hanging around, they may have sinking schedules. The sched-
ules are really only a starting point for your work. There are provisions that may, in
effect, render the sinking-fund schedules useless at this point. For example, some
companies have repurchased bonds and their sinking-fund schedule was only applica-
ble if a certain par amount of bond was outstanding, and they are below that level
now. Thus, it may affect how likely they are to sink the bond.

Some bonds, too, have put schedules. We are currently not dealing with putable
bonds. If you do not take advantage of them when you can, it will reduce the yield
in your projections. You may at least want to try and figure out how many of your
assets have puts. If you are severely penalizing yourself, it would be worth noting.

There was a workshop in 1991 in New Orleans and five or six of us in attendance
were greatly concerned about how to deal with modeling CMOs. Various answers
came from around the room as to how they were being modeled. One response was
to allocate them into capital and surplus, if you have that few. One company had
bought some and did not know what to do with them when it came time to model
them, so it sold them. Those of us who are forced to deal with them, because we
do have enough, have ended up spending quite a bit of time trying to model them
accurately.

There have been several methods that have emerged in our studies and in working
with some consultants. One way is to take a seriatim approach, where you look at
CMO-by-CMO. In doing this, you have to (using the fancy term on Wall Street)
"reverse engineer" a CMO deal. Now as actuaries, we just call it modeling. You
basically have to structure the deal and assign cash flows from the underlying
collateral to the specific tranches. You cannot just model the tranche you own
(unless you have A tranches), because your payments will be affected by the
tranches in the entire deal. You can get the tranche structure from the prospectus
that your investment department should have received before it bought the

1464
 VALUATION ACTUARY OVERVIEW

instrument. You also need the remaining collateral underlying the CMO to do your
work. That's available from Wall Street systems, either in software analytic packages
or on Blomberg on-line.

The group method has also been used. You look at all your CMOs, and you put
them into groups like Zs and planned amortization classes (PACs) and sequentials.
Look at the underlying collateral, the type of protection that each tranche would have,
and fit them into reasonably homogeneous groups. That takes the time, in my
opinion, of an investment professional. But again, it would require going through the
prospectus and really knowing what your CMOs are.

We have opted for a third option. We have gone to a software vendor who has
already done all this modeling for the CMOs for us. The software vendors model
most of the deals that are fairly straightforward. If you are buying quirky deals, they
may not be there. The purchases you made in the last few days of the month may
also not be included in the data base.

The software can greatly assist in the analytics, which from our experience, I think
regulators like to see. They keep talking about assigning ratings for CMOs, based on
volatility, ability to hedge them, and type. As they haven't done that yet, I think that
it gives great comfort, or at least insight, to a regulator if you can tell them how the
CMOs are broken out by tranche type.

You can use these methods in combination. To the extent that an outside software
vendor does not have all your CMOs modeled, you could group unmodeled ones with
modeled ones with similarattributes.

Another type of instrument you might have on your books might be an interest rate
swap. When I was naively lookingat this, I thought we bought a floating rate
security; we gave someone that payment, and they gave us a fixed payment back.
As I have come to find out, they are a little more complicated than that. I want to
kind of go throughthe mistakes that I naively made when I was first trying to model
it.

We thought we had to get the floating-rate provisions. You need to review counter-
party risk. If you have a swap, what's the likelihoodthat your counterparty will
default? So you have a floating-rate instrument, and you have promised that coupon
to someone else and they're promisingyou a payment back. Well on your Schedule
D is the floating-rateinstrument. You don't have that promise back, so you need to
get informationto do, in essence,a quality check on your counterparty. Another
thing you need to check on is if there is any mismatchin the amount and term of the
underlyingcollateral.

Here's a brief example. We have a floating-ratesecuritythat we bought, that pays

London Interbank Offered Rate (LIBOR)plus 0.25%, with a minimum 5% coupon.
The swap is that we pay out LIBOR and we receive back 6%. (Another item you
may have to contend with is that LIBORis not reallya commonly modeled index on
many projectionsystems.) In the interest rate environment we are projectinghere,
we are just going to assume it goes up 1% a year.

1465
 RECORD, VOLUME 18

Our first inclination, if we had not known about these swaps (which are really just a
footnote on the annual statement), would have been to assume the income from the
Schedule D floating-rate asset, i.e., the LIBOR payment to be received), subject to the
5% floor (not a schedule D item). Knowing about the swaps, the first inclination was
to use 6% as the interest income. But when we combined the swap with the
security, we found out that in low-interest-rate markets, we were able to keep the
difference between our floor and LIBOR. The net result actually has a bit of an
inverse floating property, which can be important if there is any type of fixed-rate
interest credits on the liability side.

ASSET ASSUMPTIONS
Defaults
Some of the basicsthat have been acceptableor standard for bond defaultsfrom our
dealings in the industryhave been in the Attman Study that came out in 1989. It
talked about default risk mortality rates and performance of corporate bonds. This
study essentially followed issues from their original bond rating. That study has since
been updated with further information on below-investment-grade bonds. New York
came out and said you can assume a default rate of specified percentages of the
AVR maximum.

The final check you should always use, instead of just deferring to an industry
standard, should be to review your own company's experience. Clearly, if your own
company's experience is worse in defaults, that would need to be reflected. Unfortu-
nately, in putting together company-default studies, there are several problems that
are faced. The law of large numbers may not apply. It is difficult to actually capture
much of the data on losses too. First, there are defaults that are taken in the form of
write-downs. Given the decrease in the interest rate environment, you could have
bought a bond at par, sold it at par, and you think you did not have a default. If
interest rates have changed substantially, there may, in fact, be an embedded credit
deterioration in the transaction that is not properly being captured. Nonetheless, to
the extent you have such information available, it would support using higher default
rates.

For mortgages, the defaults using the AVR formula will vary, based on company
experience. You can, de facto, get some of your company experience into the
mortgage default rate by using the AVR formula.

Some of the interesting things on defaults are really on the mortgage side, and 1am
not saying that because so many defaults have happened. Unfortunately, when it
comes to mortgages, we are not blessed with credit ratings. A company's mortgage
holdings largely reflect the foresight and practices of a few individuals in the mortgage
department. I do not think that those are standard across the industry. In coming up
with a mortgage assumption, too, there may be standard default rates out there in
the industry, but you may have an aged portfolio. You may have a portfolio where
you made the loan on the property, perhaps in the eady 1980s before the inflation
hit, so the fact that market values are depressed may still mean that you are in good
shape. The aging of a portfolio and the loan-to-value ratios are important to take into
account.

1466
 VALUATION ACTUARY OVERVIEW

There are published data taken in survey form by the ACLI that talk about differences
in delinquencies and foreclosures by the type of mortgage and geographic location.
There is something in the guidance notes that refers to some of the ACLI studies.

If you are fortunate enough to have your company participate in the ACLI survey, this
data may be readily available. If your company does not participate, you may have
problems, because it may not participate because it does not keep statistics that look
this way. But I'll mention an example here that I think was somewhat extreme, and
why averages may be inappropdate. In looking at the June 1992 survey results at
hotel/motel, the total in default and foreclosure was 15.89%. The Pacific area had a
23% default and foreclosure rate. The East/South Central U.S. was 2.81%. I think
it would be unduly harsh to assume an average, if you are blessed with very good
historic experience and have the loan-to-value ratios to back up a low-default
assumption.

Cals/Prepayment
Most projection systems handle calls by looking at when it is economically feasible for
the issuer to call the bond. That will vary by call premium, current refinancing options
and the cost of refinancing. Now, as I mentioned, we would again want to make a
difference between a refund provision, where you cannot prepay just to refinance to
take advantage of lower interest rates and a call that could be made at anytime.

Some companies actually buy callable instruments with an incredible, detailed knowl-
edge of the issuer's capital structure. If you own a bond that has a relatively low
interest rate in the issuer's financial structure, it would put you in a lower pdority to
refinance. Prepayments might be less than the pure economics of the issue, viewed
independently, would indicate.

Much of the same types of assumptions are made on mortgages. If you are dealing
with CMOs or pass-throughs, and you use an asset-side vendor, they will many times
have prepayment models already built. They will consider factors like burnout and
seasonality.

LtABIUTIES
UabWty Modeing
When we are putting together our models, despite the urge of the most detail-
oriented among us, it is usually not possible to do a seriatim projection. If we are
doing the New York 7 or an inverted yield curve and have a finite number of projec-
tions, it is very difficult to do, and as we look toward doing the greater stochastic
modeling, it becomes an impossibility. The liabilities can usually be grouped by some
finite charactedstics. But typically, we will look at plans based on their financial
significance and we will assign lower, less important plans to those, if we feel that
the profit margins are reasonably similar.

Another approach that can be taken is to model, say, 95% of your business; the
remaining 5% could be grossed up into the 95% modeled - this assumes that
unmodeled business has the same characteristics of everything you model.

If we are looking at lines of business in priority, interest-sensitive has caught every-

one's attention. We want to look toward broadening that to traditional lines, but even

1467
 RECORD, VOLUME 18

with traditional lines, we would probably want to differentiate between whole-life-type

plans and term insurance. Finally, there is health insurance, v_r_'_nin
health, disability
income, long-term care, and other longer liabilities would have a higher priority than
short-term medical coverages.

In looking at liability characteristics, we look at many of the same issues we look at

when we are trying to price a product. You would want to look at the issue age, sex
and the underwriting categories. You want to make sure that the model is not biased
or is properly biased by any pricing to a certain age you may have done. You may
also want to reflect benefit triggers at given ages. Many people talk about doing
annuity models with one issue age; if you have qualified business (at age 70½
minimum distribution start) it would probably be worthwhile to do a little refinement
on issue ages.

Other final points follow. The average face amount you get, if you were to throw all
data together, may not be representative of your underwriting practices. Issue mode
will affect the timing for lapses in your model. Date of issue is also important. In
terms of issue years, you can make some simplifying assumptions by grouping older
blocks of business together. They may represent a small portion of the in-force
business. If you have a lot of annuity business that is out of the surrender-charge
period, and nothing special happens at any given age, for all intents and purposes, it's
basically the same business.

Uability Assumptions
Expenses are not marginal. We need to make sure that when we do a projection, the
expenses that come out of the model and actually bear some resemblance to reality.

There are also dynamic assumptions that we need to make. We might have lapses
positive or negative. We have now had a decreasing interest rate environment where
we can track how policyholders react when interest rates go down. We need to be
very careful, when setting lapse assumptions, that we don't build in our standard
assumption rates resulting from the fact that there has been a decreasing interest rate
environment. The unfortunate wild card in the whole process at this point is what
happens when interest rates go up.

Another dynamic assumption is premium persistency, which may or may not be

important, depending on the new-money/old-money strategy. We would want to
approach policy-loan assumptions with a different amount of diligence, depending
upon whether the loans are fixed-rate loans or have a fixed spread, or are zero-cost
loans.

Credited interest is typically expressed as a function of target spread and a definition

of a competitor rate.

We need to look at mortality and morbidity and make sure those validate to actual.

I think that we need to do a lot of sensitivity testing. I know it's a struggle to get out
the seven New York scenarios, but that shouldn't be the end of it. As I mentioned,
we do not have any data on what happens when interest rates go up. That is a
good area in which to do extra testing. Given that the interest rate environment is in

1468
 VALUATION ACTUARY OVERVIEW

turmoil, it is important that we look at our assumptions and try to figure out whether
recent or prolonged history is really a decent indication of our future experience.

There are guidance notes that talk about how to do testing for "obligation" or the C-2
risk.

Some assumptions need to be made at the company level. Federal income tax is one
of them. What's your tax rate? Shareholder dividend practice is important for those
of you who have a practice of making them. One of the important items is that in
some of the scenarios where results are not good, results are good for a while. I
think for those who have problems, in say, increasing interest rate environments, how
good or bad those will be will depend largely where the interest rates came from
before the date of the study. For example, if we go into an increasing interest rate
environment today, it will take us awhile to get to where we were a year or two ago.
We may not even realize how bad a scenario will be until we get several years into it.
We need to look closely at how realistic it would be for us to continue paying
dividends in the eady years of the scenario that ultimately does not have desirable
results. If you have a stated objective to build surplus, then it is perfectly appropriate
to not pay dividends.

ADDITIONAL RESERVES
I have some finalthoughts on when we do establishadditionalreserves. Doug talked
a bit about the guidance notes that have been proposed, so I will not talk about that.
We do need to remember that it's not just policyholderreserves, but AVR is available
to offset defaults, and IMR is availableto be spread into the future, as well as to
offset any capital lossesfrom interest rates.

In the past, various individualshave had conversationswith New York. They have,
on occasion, used a standard that if you failtwo tests in a given direction (either
increasingor decreasing),you may be asked to do specialtests and set up reserves
with either a 30- or 40-basis-pointchange in that directionper annum. There are also
some questions as to how to judge adequacy of reserves. Do we look at the end of
the projectionperiod? That's typicallythe way it's been stated. Some guidance in
the actuarial standardsthat says you reallydo need to look if you have severe
deviations in your profits alongthe way. It's not just important as to what the
endpoint is, but if you were to go insolvent before you got there, it is not going to do
you or your policyholdersmuch good. There are also questions as to whether we
look at book value or market value. Some of the biggest issueswe have had include,
how good are the market-value numbers? There has been a lot of volatility in the
junk-bond market. Mortgages, real estate and privateplacements may not have
market values.

When you are balancingreserve adequacy, you also have to considerthe capital and
surplusstrength. It is reallynice to have reserves that are set up to be redundant.
We really likethose when we do testing. BUt we are given no credit for them when
we look at risk-based capital. We shouldlook at, not only how the resultscome out
of our projections,but why they are what they are. You may be able to use this
work in dealingwith ratingagencies, if you can prove that you have excess reserves
that provide extra margins that are not specificallyaccounted for in the risk-based
capital formulas.

1469
 RECORD, VOLUME 18

MR. MARC F. PITONIAK: I will be covering a case study of UL, with an emphasis on
the practical points the valuation actuary should consider when conducting cash-flow
testing. I want to point out that I will be discussing one approach to cash-flow
testing. Obviously, other approaches are possible.

The topics I will cover include basic Schedule D information, call and prepayments,
spread assumptions, reinvestment strategies, and reinvestment parameters. I will
briefly discuss the first two topics. Next I will discuss the dynamic elements that you
should consider when you do your modeling. I will list the sequence of steps your
model can take to do cash-flow testing. We'll spend the bulk of the time discussing
the case study.

I'd like to talk a little about segmentation of your existing and future assets. Now
when you file new products for approval, different state insurance departments are
requesting information on the kinds of assets you plan on purchasing with the
premium received and what your interest-crediting sl_ategy will be. Also, the NAIC
model regulation requires that the valuation actuary disclose each year any inconsis-
tencies in the asset-allocation methodology. This intimate tying of assets with their
liability counterparts is here to stay. An asset-segmentation methodology is becoming
both a regulatory and a practical requirement.

My focus in this case study will be on bonds. When I use the word profile in the
context of cash-flow modeling, I refer to the final set of assumptions and characteris-
tics that are used in your projections. There is a wide spectrum of approaches you
can use here. Schedule D usually provides a good starting point for me. My
preference is generally to project each asset in a company's Schedule D, and I begin
the process by requesting the data processing department to provide the most recent
Schedule D that can be dumped onto a PC in ASCII format. This is obviously only a
starting point, because you will need additional detail on calls, prepayments, sinking-
fund payments, and CMO detail.

Let's assume this information is already available. The approach I take is to use a
"social security" tag in each of the basic asset, call, and sinking-fund files that will link
the information of a given security. By "social security" tag I mean the CUSIP and lot
number of each security. I use this approach so that I am only restricted by the size
of the hard disk. For a given security in the basic asset file, I check the call flag to
see if it is true, if the bond is callable. From the call file I read the first and last call
premiums, the first and last call dates, and the frequency of the change in call
premium. Next, I check the sinking-fund flag in the basic asset file. If the security
has any sinking-fund provisions, then I read the sinking-fund file to see if there is any
current scheduled payments to be made. The sinking fund would contain the sinking-
fund dates and the amount to be paid. A given security could, of course, have more
than one sinking-fund date.

Next, you need some way to assign security-class characteristics to each of the
securities that are in the asset files. I include these security-class assumptions in
what I refer to here as the spread file. Examples of class assumptions are spread
over treasury, call barriers, call percentages, default, and investment expense,

1470
 VALUATION ACTUARY OVERVIEW

Next, you will have to decide what your reinvestment strategies are. You will have to
make key decisions, such as what your investment department will do in negative
cash-flow situations. Do you borrow or sell assets? Also, you presumably will have
different investment strategies to cover different shapes of the yield curve.

In the reinvestment parameter file, most of the fields would be similar to the spread
file. For each reinvestment strategy you will need additional information, such as the
amount of cash you want to invest in each security class.

Once you have made your asset assumptions, you can now shift your attention to
developing your liability assumptions. Basic information you need is as follows:
pricing spread, surrenders, crediting strategies, lapses, expenses, in-force data, loans,
and valuation. Since I will discuss dynamic assumptions later, I will restrict my
attention here to a practical approach you can employ to assimilate all of this
information.

The in-force file contains all of the information you would need, such as issue age,
issue date, etc, to calculate all basic values such as reserves, account values and
guaranteed maturity funds. Again, just as in the asset section, it is desirable to
download this information from the mainframe. You will want to model this raw data
into model cells by using a model-cell generator.

V_rrththe exception of the valuation menu, each record for each of these menus has a
field called the product code. In addition, each record in the in-force menu contains a
field called the valuation code. Now for a given policyholder cell in the in-force file,
you have a link established with the valuation menu so that you can calculate basic
values according to the proper valuation basis. In addition, for this same policyholder
cell, you have a link established with all of the other menus so that you can summa-
rize results at the product level.

Next we'll discuss C-1 and C-2 assumptions. As you probably realize, there are still
many areas in which the actuary will need guidance as he or she does an asset-
adequacy analysis. Donna Claire has recognized this need as the model regulation is
implemented and continues to evolve. To this end, Donna has formed a committee
called the Practice Notes Task Force that will address these areas. Two practice
notes of particular interest to us in this presentation is the guidance note on C-2
assumptions by Craig Likkel and the guidance note on defaults by Michael Zurcher.
Michael has also been very active in developing the future risk-based capital
requirements.

In his practice note draft, Mike states that current approaches that most valuation
actuaries employ to project the costs of defaults lack sophistication. Some of the
weaknesses are" (1) the practice of applying static default factors to each asset
quality class, (2) the lack of integration of default assumptions with both economic
conditions and the interest rate scenarios, and (3) not recognizing portfolio characteris-
tics, such as the number of assets, the size of individual issues and concentration of
assets with specific characteristics.

Now the asset valuation reserve (AVR) requirement that is in effect with the 1992
annual statement will require the appointed actuary to recognize the differences in

1471
 RECORD, VOLUME 18

defaults among the different asset-quality classes. Though the anticipated risk-based
capital (RBC)requirements are not part of the model regulation, the RBC C-1 factors
will recognize C-1 risk through quality, size and concentration factors. As you model
defaults assumptions for the asset-adequacy analysis, it would be appropriate for you
to also recognize the additional factors defined in the RBC C-1.

Ultimately, the valuation actuary will probably want and/or need to incorporate some
stochastic generation of default factors for the securities. I will mention that the
Committee on Valuation and Related Areas (COVARA), with the help of Peter Deakins
and Laird Zacheis, is currently working on developing such an approach.

Now we'll get into dynamism. There are parameters that should be dynamically
recognized in your cash-flow testing, including asset defaults and calls, crediting
strategies and lapses, disinvestments and reinvestrnents, and policy loans and
policyholder dividends.

Calls obviously should recognize the coupon value and the current market interest
rates. Two tests that you should model for are the spread between coupon rate and
current market rate and also the cost to the borrower of calling the bond. Unless you
are modeling mortgage securities, where there is always some financing that occurs
regardless of the interest environment, you generally should assume that the call will
not be made unless it makes economic sense to the borrower.

To assign a value to the asset you may be selling, you also need the capability to
calculate market value of assets. This means that you should have the capability to
determine both scheduled and nonscheduled cash flow and price compression of
bonds with embedded options. When you do a market-value calculation, you have, in
effect, one projection process nested within the primary projection.

For crediting, unless your business is captured, you will have to explicitly recognize
the difference between your crediting strategy and the competition's. Your crediting
will be a function of your competition, your portfolio rate and your pricing spread.

For lapses, it's important to recognize fund, premium and policy lapsation. The lapse
assumptions are a function of your credited rate and the competition's and the degree
of surrender-charge protection you have. Also for noninterest-sensitive lapsation, it
may be desirable to recognize a base lapse assumption with select and ultimate
characteristics.

Once you have made all of your asset and liability assumptions, a sequence of steps
that you can follow in the actual projection includes:

1. First process the assets. By process, I mean that for each security or modeled
security, you must first determine the spread over treasury of comparable
maturity, the default to be attributed to this security, and the investment
expense. Next, for the given security, you want to conduct tests, such as is
there a coupon payment or any calls or sinking-fund payments. What about
maturity? Answering these questions will allow you to determine both the
cash flow and investment income.

1472
 VALUATION ACTUARY OVERVIEW

2. Next, for each issue, calculate reserves, account values, guaranteed maturity
fund values, and credited amount, premium, coat of insurance, expense
charges, mortality, and lapse assumptions. Sum the results at the product
level.
3. Next, at the product level, determine your general expenses. These last two
steps (2 and 3) allow you to generate the insurance cash flow.
4. Next, generate both an income statement and contribution to surplus.
5. Finally, given your previously defined investment strategies, invest the total
cash available and repeat the process.

Now we're ready to start the case study. Let me first start defining some terms. I
will use the standard New York Regulation 126 seven scenarios in these projections.
Of course, projecting under the seven scenarios does not conatitute an adequate
spectrum of interest environments. You should test at least 100 additional scenarios
that can be randomly generated.

"Level" means that the interest rates for securities of different maturities stay con-
stant. In the up environment, uniform means interest rates increase 500 basis points
in the first 10 years. Up and down means interest rates increase 500 basis points in
the first five years and then decrease 500 basis points in years 6-10. Finally, spike
means interest rates increase immediately by 300 basis points. The down environ-
ment is described similarly, but with the interest rates headed in a different direction
from the up environment.

Pleaserefer to Chart 1. The issuer of the callable bond I modeled was not allowed to
call until 1994. The jump in the level scenario in year 1994 indicates that the
borrower has exercised his or her call option. For 1994, in the other scenarios, we
see that because interest rates increase beyond the critical call point, call activity
doesn't occur, except marginally, for the uniform scenario. Notice that in the year
2001, however, the original bond is called in the up-and-down scenario, because
interest rates have been decreased to their original levels.

The next observation is that in the up-and-down scenario, cash flow becomes very
negative. This is because the company's crediting strategy is not staying competitive
with its perceived competition, and as a result, there are considerable surrenders
occurring. It is not until 1999 that the company's crediting strategy, portfolio rate
and competition become more aligned.

Finally, note how level the cash flow is in the spike scenario. This is because the
spike scenario results in an interest rate that is not substantially different from the
portfolio rate, so that surrenders and premium lapsation don't occur to any great
extent.

Chart 2 is rather boring. In the down environment, calls occur in all scenarios. Also,
the reinveatment strategy of this study was to purchase noncallable bonds with
sinking-fund payments that commence five years from issue.

Because the original bond was called in 1994, all these scenarios show that in 1999-
2001, sinking-fund payments are being made from the bonds that were purchased in
1994.

1473
 RECORD, VOLUME 18

CHART 1
Universal Life Cash Flow
(Up Environment)

$4O
$35

$25

$20 .........

$10
$5
o $0
$-5 L................T......... .... ,....... ....
1992 1994 1996 1998 2001

Level ..... Uniform ..........

Upand Down ..........
Spike

Lever ..... Uniform .........

Up andDown ..........
Spike

1474
 VALUATION ACTUARY OVERVIEW

There are two reasons why there are not greater differences among these scenarios.
The first is that even the level scenario is enough to trigger substantial calls, and the
second reason is the floor I used for Treasuries negates the full effect that the other
scenarios would otherwise experience.

Chart 3 displays the statutory surplus build-up in an up environment. The level

scenario does better than the other scenarios. The up-and-down scenario does the
poorest. Recall that in the corresponding cash-flow graph, that was a significant
amount of negative cash flow. This graph illustrates the result of having to fund
excessive surrenders by selling assets at a market-value loss or borrowing at a higher
interest rate.

CHART 3
Universal Life Surplus
(Up Environment)

$35-

$301 J
0_
¢0
=
O $25-
r-

b, $20-
°
-= ...........
=
E- $15_
$10-- _ ":""" .......,"

$5 -_

$o _'_ ....
1992 1996' 1998' r 20'ol

-- Level ...... Uniform ...........

Up and Down ..........Spike

In Chart 4, we see that the level scenario again does the best. The other scenarios
are fairly close to the level because the floor assumption is in effect. Lower values
indicate the effect of a lower reinvestment rate.

Now, as you know, the AVR/IMR requirements are now part of the asset-adequacy
analysis. If anyone would like to see the comparable surplus graphs with the
AVR/IMR requirements in effect, please see me.

We should also look at the market value of surplus in Chart 5. The spike scenario is
no surprise, considering that interest rates immediately spiked 300 basis points. The
up-and-down scenario should start out slightly below the uniform and level scenarios.
Note that the company has the worst experience under the up-and-down scenario. It
isn't until 1999 that the company starts to recover, relative to the other scenarios.

1475
 RECORD, VOLUME 18

CHART 4
Universal Life Surplus
(Down Environment)

$35

=
O $25
r-

I_
o_ $20

$15
J
_'
o $1o T--

$0 ........................ _ .... _ ..................

, ..... , ..... ,
1992 1994 1996 1998 2001

Level ..... Uniform ..........

Up and Down ..........
Spike

CHART 5
Universal Life Surplus
(Up Environment)

$4O

"_ $35

¢__.$25 /
= $20 :"-...............

m $15
._ s_o
$5 ..... :';............ •............

=o19'92' 19'94 19'_ ' 1_8 ' 2o01

-- Level ..... Uniform ...........

Up and Down ..........
Spike

1476
 VALUATION ACTUARY OVERVIEW

In Chart 6 the level scenario ultimately prevails again, but the uniform scenario does
better for 1994-98. This is because the company is able to, for a while, maintain
larger spreads in its interest-crediting scheme. Ultimately, the effects of reinvesting at
a lower rate causes the uniform scenario to fall behind the level scenario.

CHART 6
Universal Life Surplus
(Down Environment)

$4O

$85

0
$80L
$25
r._
'-"z /s _ ,,.,,_:..""

$20-
O3

"_ $15 -'J'"

$10- .. '...... ,........

1992 1994 1996 1998 2001

-- Level ..... Uniform .........

Upand Down ..........
Spike

I will touch on the kinds of parameters upon which you should perform sensitivity
testing and illustrate some of the effects of the testing. It is expected, however, that
the valuation actuary performing sensitivity testing should go far beyond this illustra-
tion. For example, with respect to reinvestment strategies, you should consider
securities of different qualities, kinds and maturities.

The base scenario will be the one referred to as the level scenario in eadier charts.
Because the base scenario used callablebonds, I want to illustratethe effect that
noncallable bonds will have on surplus. Also, to see what shock a change in
reinveatment strategy will have, I will assume that the company wants to stay in a
cash position and will invest in one-year securities.

The crediting scenario will assume that the company follows the market with no
consideration of what the underlying portfolio rate is. The company is looking for just
its pricing spread. This study assumes the universal life (UL) policy was issued 10
years ago to someone of age 35.

1477
 RECORD, VOLUME 18

In Chart 7 the base, default and reinvestment scenarios all show quite a bit of call
activity in 1994. Because the company is investing in AAs, we see that even
doubling the default rate has no observableeffect on the cash flow. The call scenario
shows that the cash flow remainslevelthroughout the projectionperiod. The reinvest
scenario shows a level of cash to invest in every year after the originalcall activity.

CHART 7
Universal Life Cash Flow
(Asset Sensitivity)

=oao_ // \ ........

0 -_- -F...... I ............... r [ -I I ]

1992 1994 19196 1998 2001
! Base --_1"-- Defaults ]
i ................
Calls Reinvest i

In Chart 8, we really don't see much differencesin the cash flow, becauseof the
quality of the securitiespurchasedand the age of the policyholder.

In Chart 9, we see the effect of losingthe spreadthat is generallyassociatedwith

investing longer term. Sure the company has no cash-flow shortfalls,but the cumula-
tive effects of such a reinvestmentstrategy considerablyhurts the company's level of
surplus.

In Chart 10, we see the effect of changingthe creditingstrategy and surpassingthe

performance of the base scenario. Finally,the mortality scenarioillustratesthe
cumulative resultof increasingmortality by 50%.

MR. GOOTZEIT: The cash-flow testing surveythat was mentioned was publishedin
the June 1992 issueof The Actuary.

1478
 VALUATION ACTUARY OVERVIEW

CHART 8
Universal Life Cash Flow
(Liability Sensitivity)

¢ $35
_ $30
$25-
$20-
$15-

,.c

9 $5 .....
$10- = ........ ===
e_ _ .................::::::::::::::::::::::::::
$0- . .................:

$-51992 1_,_ 19'96 _ 19'98 ' =®1

Base ...... Crediting ..........

Premiums ..........
Mortality

CHART 9
Universal Life Surplus
(Asset Sensitivity)

$35
"10
" $30 ,

" t
ol
0
,-- $25

$20
crj

(n $15 _ ...............................

if)

$0 -_= --
1992 19'94 19'96 19'98 20'01

-- Base ...... Defaults ..........

Calls ..........Reinvest

1479
 RECORD, VOLUME 18

CHART 10
Universal Life Surplus
(Liability Sensitivity)

1992 1994 1996 1998 2001

Base ...... Crediting ..........

Premiums ...........
Mortality

1480