|
Insurance Abstract
The invention relates to a system and method for evaluating valuations
groups of insurance policies. The steps involve a) retrieving at
least one characteristic for each policy from the plurality of characteristics
for each policy in the group of policies; b) obtaining at least
one derived characteristics for each policy in the group of policies
from plurality of characteristics for each policy in the group of
policies; c) calculating a group expected value for each of the
at least one characteristic and each of the at least one derived
characteristic; d) receiving from the input device, a set of tolerances
for each of the at least one characteristics and each of the at
least one derived characteristic; e) minimizing a linear objective
function with a set of policy weights wherein a sum of an at least
one weighted characteristic, obtained by multiplying the policy
weight with one each one of the at least one characteristic and
each one of the at least one derived characteristic, is equal to
or within the received tolerance of the group expected value for
each of the one or more characteristic and each of the one or more
derived characteristic; f) selecting policies with a non-zero policy
weight; g) calculating at least one risk valuation result using
the selected policies; and h) outputting the result of the at least
one risk valuation result to the output device. The results of the
at least one risk valuation result using the selected policies substantially
correspond to the results of calculating the at least one risk valuation
result on the group of policies.
Insurance Claims
1. A system for evaluating risk scenarios relating to a group of
insurance policies comprising a processor in communication with
a database containing a plurality of characteristics for each policy
in the group of policies relating to value and risk, an input device;
and an output device; and code implemented in the system for instructing
the processor to: a) retrieve at least one characteristic for each
policy from the plurality of characteristics for each policy in
the group of policies; b) obtain at least one derived characteristic
for each policy in the group of policies from the plurality of characteristics
for each policy in the group of policies; c) calculate a group expected
value for each of the at least one characteristic and each of the
at least one derived characteristic; d) receive from the input device,
a set of tolerances for each of the at least one characteristics
and each of the at least one derived characteristic; e) minimize
a linear objective function with a set of policy weights wherein
a sum of an at least one weighted characteristic, obtained by multiplying
the policy weight with each one of the at least one characteristic
and each one of the at least one derived characteristic, is equal
to or within the received tolerance of the group expected value
for each of the at least one characteristic and each of the at least
one derived characteristic; f) select policies with a non-zero policy
weight; g) calculate at least one risk valuation result using the
selected policies; and h) output the result of the at least one
risk valuation result to the output device; wherein the system outputs
results of the at least one risk valuation result using the selected
policies that substantially correspond to the results of calculating
the at least one valuation result on the group of policies.
2. The system of claim 1 wherein the code instructs the processor
to minimize a linear objective function with a set of policy weights
wherein a sum of the at least one weighted characteristic, obtained
by multiplying the policy weight with each one of the at least one
characteristic and each one of the at least one derived characteristic,
is equal to or within the received tolerance of the group expected
value for each one of the at least one characteristic and each one
of the at least one derived characteristic: i) forming a matrix
containing the at least one characteristic and at least one derived
characteristics for all policies in the group of policies; ii) forming
a first vector containing the group expected value for each of the
at least one characteristic and each of the at least one derived
characteristic; iii) forming a second vector containing policy weights
for each of the policies in the group of policies; iv) minimizing
the linear objective function to obtain the policy weights such
that the matrix combined with the policy weights is within the tolerances
of the first vector.
3. The system of claim 1 wherein the code instructs the processor
to select policies with a policy weight that is not zero or near-zero.
4. A method of efficiently calculating scenarios for a collection
of policies comprising the steps of: a) retrieving at least one
characteristic for each policy from the plurality of characteristics
for each policy in the group of policies; b) obtaining at least
one derived characteristics for each policy in the group of policies
from the plurality of characteristics for each policy in the group
of policies; c) calculating a group expected value for each of the
at least one characteristic and each of the at least one derived
characteristic; d) receiving from the input device, a set of tolerances
for each of the at least one characteristics and each of the at
least one derived characteristic; e) minimizing a linear objective
function with a set of policy weights wherein a sum of an at least
one weighted characteristic, obtained by multiplying the policy
weight with each one of the at least one characteristic and each
one of the at least one derived characteristic, is equal to or within
the received tolerance of the group expected value for each of the
one or more characteristic and each of the one or more derived characteristic;
f) selecting policies with a non-zero policy weight; g) calculating
at least one risk valuation result using the selected policies;
and h) outputting the result of the at least one risk valuation
result to the output device; wherein the results of the at least
one risk valuation result using the selected policies substantially
correspond to the results of calculating the at least one risk valuation
result on the group of policies.
5. The method of claim 4 wherein the step of minimizing a linear
objective function with a set of policy weights wherein a sum of
the at least one weighted characteristic, obtained by multiplying
the policy weight with each one of the at least one characteristic
and each one of the at least one derived characteristic, is equal
to or within the received tolerance of the group expected value
for each of the one or more characteristic and each of the one or
more derived characteristic further comprises the steps of: i) forming
a matrix containing the at least one characteristic and at least
one derived characteristics for all policies in the group of policies;
ii) forming a first vector containing the group expected value for
each of the at least one characteristic and each of the at least
one derived characteristic; iii) forming a second vector containing
policy weights for each of the policies in the group of policies;
iv) minimizing the linear objective function to obtain the policy
weights such that the matrix combined with the policy weights is
within the tolerances of the first vector.
6. The method of claim 4 where the selecting of policies comprises
selecting policies with a policy weight that is not zero or near-zero.
Insurance Description
FIELD OF THE INVENTION
[0001] This invention relates to a system, apparatus and method
for issuing insurance policies by more efficiently and cost effectively
evaluating the value of financial insurance products. In particular
this invention relates to efficiently determining the value of numerous
financial policies.
BACKGROUND OF THE INVENTION
[0002] Insurance contracts are used by individuals and organizations
to manage risks. As people interact and make decisions, they must
evaluate risks and make choices. In the face of financially severe
but unlikely events, people may make decisions to act in a risk
adverse manner to avoid the possibility of such outcomes. Such decisions
may negatively affect business activity and the economy when beneficial
but risky activities are not undertaken. With insurance, a person
can shift risk and may therefore evaluate available options differently.
Beneficial but risky activities may be more likely to be taken,
positively benefiting business activity and the economy. The availability
of insurance policies can therefore benefit those participating
in the economy as well as the economy as a whole.
[0003] Insurance companies often sell financial guarantees embedded
in life insurance products to customers. Generally, the focus is
on selling products to people with money who want to plan for their
retirement. Many of these products offer customers, the investors
or policyholders, investment returns and in addition contain embedded
financial guarantees. A simple product of this design is a Guaranteed
Minimum Accumulation Benefit, or GMAB, where a policyholder invests
money in a mutual fund or similar vehicle and is at least guaranteed
to get their principal back after eight years regardless of actual
fund performance. With a GMAB, the policyholder has the potential
upside if markets increase over the eight years, and if the markets
have fallen, the policyholder will at least get their money back.
[0004] Companies selling these financial guarantees must periodically
value and report on the risk of the financial guarantees. In addition,
regulatory requirements often require companies to report their
risk exposure and require the companies to have sufficient reserve
assets and capital on hand to support the risk profile associated
with the financial guarantees they have sold. Valuing financial
guarantees embedded in life insurance products for financial, risk
management and regulatory reporting, is a computationally challenging
prospect for insurance companies. Companies often use substantial
computer power and internal and external resources to perform the
necessary calculations to value and report on such products like
variable annuities, segregated funds or unit linked like contracts.
[0005] There are at least several reasons why it is generally time
consuming and difficult to calculate the value of such complex insurance
products. Typically, these products have long maturities, with a
single policy having a life span of over 30 years. In addition,
the valuation of the product is path dependent, which means their
value is driven not only by the final state conditions but also
on the path taken to reach the final state. Further, the industry
practice is to use monthly cash flows over 30 years with up to 5000
scenarios use seriatim calculations as a guiding valuation principle.
Calculations on a seriatim basis means calculating the result on
a policy by policy basis, in other words, the calculation is completed
for every policy on a quarterly basis and perhaps more frequently
in the case of multi-jurisdictional financial and regulatory reporting
requirements. As an example, for a single policy, with 5000 scenarios
and 360 time steps, about 1,800,000 cash flows have to be modelled,
discounted and then summed back to time zero to create a net present
value vector with a corresponding net present value result for each
scenario.
[0006] In addition, regulatory reporting requirements may require
that a conditional tail expectation be used to determine the appropriate
reserves and capital requirement for the business activity.
[0007] If there is a hedge program in place, additional simulations
may be required to reflect the hedging activity over each time step
in each scenario. Such calculations require calculating the liability
value and sensitivities, and the payoffs from the hedge portfolio
at each point to create a hedging cash flow matrix with the same
dimensions as the hedge item or naked liability cash flow matrix
to create an overall net cash flow matrix. These nominal cash flows
can then be discounted and summed back to time zero to produce a
vector of net present values, of length equal to the number of scenarios
used in the valuation process, which are in turn used to calculate
an appropriate conditional tail expectation, under Canadian financial
reporting and, with some modification, for United States regulatory
reporting.
[0008] A conditional tail expectation is a sample average, or measure
of central tendency on pre-selected group of ranked sample observations.
CTE0 is defined as the sample average. CTE95 is defined to be the
average of the worst 5% of sample observations.
[0009] The computations needed to calculate these values over all
policies and over all the scenarios requires substantial time and
resources. For some companies, it may take hundreds of hours using
hundreds of computers to calculate the necessary quarterly financial
valuations.
[0010] As described above, the calculations are performed on a
seriatim basis. For each policy in the company's portfolio, each
cash flow is modelled and relevant information regarding the policy
is collected. For a company that has millions of policies, billions
of calculations are required to produce summary valuation results.
Regulators often require that the valuations be performed on every
policy to ensure that sufficient capital is available and a low
estimate not used. If fewer policies are used, regulators may require
that companies demonstrate to the regulator that their model contains
all the important risk characteristics of the whole population of
policies and will not produce, intentionally or unintentionally,
less conservative capital figures. It is believed that most insurance
companies rely on seriatim calculations, which get more numerous
as they sell more policies. Therefore, the time and resources required
to calculation the valuations gets larger over time. Such constraints
place effective limits on the number of policies that can be effective
sold and managed in absence of additional computing resources. Companies
may spend ever increasing amounts of money on these computing resources
to value these products, including costs associated with internal
and external resources, such as employees, consultants, hardware,
redundancy and security.
[0011] These costs ultimately are built into the policy premiums
and over time paid for by the policyholders, increasing the cost
of the policies and making them less affordable and therefore less
available to the consumer. This has a deleterious effect on risk
avoidance, and thus risky activities that may be beneficial to industry
and the economy are less likely to be taken, detracting from business
activity and the economy.
[0012] One technique to lessen the number of required valuation
calculations is called grouping. Grouping usually involves creating
a list of quantitative characteristics and dividing each of these
quantitative characteristic in to a series of relevant ranges, known
as buckets. Each policy can then be mapped to an intersection of
these ranges based on the selected quantitative characteristics
to create a cell or group of similar policies. A weighing mechanism
can then be used to create a `representative` or pseudo policy if
more than one policy if found in a cell. Such a weighing mechanism
may be a midpoint, sum, or dollar weighted average. The more characteristics
that are used and the more buckets that are employed in each range,
the larger the number of representative policies that will be found
in the final grouping.
[0013] Typically, less than 10 basic quantitative characteristics
are used and typically less than 15 buckets are used for each quantitative
characteristic. Using this approach, one typically ends up with
about 15-30% of the original policy count in grouped policy selection
process. With the grouped policies, one can then perform the seriatim
valuation calculations described above on the group policies identified
in the grouping process instead of on every policy.
[0014] Grouping has several disadvantages. Firstly, there is the
selection of the correct set of quantitative characteristics. Selecting
a poor set of quantitative characteristics may result in poor results.
Secondly, the choice of the buckets can also affect the results.
If the selection of buckets is done poorly, one may have cells with
a no policies, whereas other cells may have thousands of policies.
The choice of weighing mechanism to obtain the representative policy
within each cell can also affect quality of the results. Perhaps
the most significant disadvantage is the lack of certainty that
the grouped selection will reproduce the quantitative characteristics
of the original population, let alone time zero seriatim values
or risk factor sensitivities. It is difficult to provide estimates
of the accuracy of the valuation results using a grouping technique.
Generally, regulators are hesitant to approve the use of such a
technique without guarantees the results derived from grouping correspond
to the actual results from the full population of policies.
[0015] If the amount of time and resources required to calculate
the valuations is fixed, then additional policies can not be issued
without adding additional resources to complete the calculations
on time. In addition, if the calculations for various scenarios
can be made more efficient, more scenarios can be calculated in
the same amount of time resulting in more accurate valuation results
for quarterly and annual reporting and for any hedging programs.
BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In drawings which illustrate by way of example only a preferred
embodiment of the invention,
[0017] FIG. 1 is a schematic representation of an apparatus for
implementing an embodiment of the invention.
[0018] FIG. 2 is a flow chart showing the operations performed
by the apparatus of FIG. 1.
[0019] FIG. 3 is a flow chart showing a method of implementation
of an embodiment of the invention.
[0020] FIGS. 4a and 4b are first and second portions of a chart
containing example policy information.
[0021] FIG. 5 is a chart showing group expected values for quantitative
characteristics.
[0022] FIG. 6 is a chart comparing the solution to the expected
group values.
[0023] FIG. 7 is a chart containing the selected policies and corresponding
policy weight coefficients.
DETAILED DESCRIPTION OF THE INVENTION
[0024] The preferred system and method can be used to determine
the valuation of a group of policies to allow issuance of further
policies and to perform calculations more efficiently and at less
cost than conventional techniques.
[0025] With reference to FIG. 1, a system for implementing an embodiment
of the method of the invention performs a series of operations.
[0026] In block 100 illustrated in FIG. 2, information is obtained
regarding the collection of policies of interest. Information about
the policies may include value, expiry, age of a policy holder,
gender of a policy holder, and/or any other information relevant
to a determination of the value or intrinsic risk of policy. The
database may contain a large number of policies and a large amount
of information on the policies. The information in the database
may optionally be transferred to the processor or memory as a first
step or access to said information may be deferred until specific
information is required in the later steps performed in the method.
[0027] With reference to block 105, derived data is calculated
for each policy based on information extracted from each policy
and from external sources of information. The derived data may be
derived from a valuation model, or from perturbation or pre-selected
risk scenarios. Derived data may include intermediate and final
valuation calculation figures used in financial reporting or in
risk management processes. Examples of the derived information are
delta, vega, gamma, rho, theta, and other sensitivities of a policy
to valuation model inputs. Delta, vega, gamma, rho and theta are
examples of `The Greeks`, well known in mathematical finance and
used to represent a specific measure of market based risk. For example,
delta usually represents a measure of the sensitivity to changes
in the price of the underlying option or policy liability and gamma
is a measure of the rate of change of the delta. The derived data
may also include the net present value of the policy under different
scenarios or across different time steps along one or more scenarios.
Each intrinsic or derived value becomes a quantitative characteristic
of a policy.
[0028] The intrinsic and derived data from the policies may preferably
be managed by the processor as a matrix. In the matrix, the data
for each policy forms the columns. Each row of the matrix contains
the data for all policies for a particular quantitative characteristic.
[0029] As indicated in block 110, an expected group value is calculated
for each quantitative characteristic by adding the numeric quantity
associated with each policy for a given quantitative characteristic.
Each policy is given an equal weight of `one` as an initial policy
weight to obtain the expected group value. The expected group values
are preferably managed by the processor as a vector with a length
equal to the number of quantitative characteristics.
[0030] With the matrix of data for the policies and a vector of
expected group values as the constraint, as indicated at block 115,
the processor minimizes a linear objective function formed by combining
the policy weights and policy coefficients subject to the constraints.
The policy weights were fixed at `one` to obtain the expected group
value. If the policy weights are fixed at `one` and if the policy
coefficients of the objective function are also set to `one`, then
all constraints would be satisfied, and the objective function would
equal the total number of policies. To make effective use of the
system, tolerances are specified for each quantitative characteristic.
The tolerances may be obtained from a user using the system or from
a configuration file associated with system. The tolerances set
bounds on each expected group value. The minimizing policy weights,
when applied to the policies, may result in group values that differ
from the expected group values by the amounts specified in the tolerances.
[0031] To aid in determining the minimized policy weights, the
processor may use a linear algebra optimization technique such as
linear programming. The technique may be represented in the following
form: Minimize Z=c.sup.Tx
[0032] such that: Ax.about.b x.about.0.sub.n
[0033] where [0034] Z represents an objective function; [0035]
c represents a vector of linear coefficients for the policy coefficients,
and c.sup.T indicates the transpose of c; [0036] A represents the
matrix of quantitative characteristics for all the polices; [0037]
b represents a vector of expected group values; [0038] x represents
the policy weights; 0.sub.n represents a vector of zeros, indicating
the contraints on the system.
[0039] The minimizing policy weights determined by the processor
specify the influence to be associated with each policy in the result.
A number of the policy weights are likely to be zero or very close
to zero, indicating that the policy associated with that coefficient
is not a member of the selected policies. A policy weight which
is not zero or very close to zero indicates that a policy is part
of the selected policies. A group of selected policies is obtained
as indicated at block 120.
[0040] As indicated in block 125, the group of selected policies
and associated weights may be used by the system to calculate various
values and sensitivities. Calculations performed on the selected
scenarios match, within the tolerances specified, corresponding
calculations performed on all the policies for all the quantitative
characteristics.
[0041] Computation time and resources are reduced because the calculations
need only be done on the select group of policies rather than on
all the policies to obtain relevant risk and valuation results.
By reducing the time and resources required to perform the calculations,
an organization that issues policies can issue additional policies,
improve the accuracy of current valuation statistics, and allow
for risk management studies to be completed and other statistics
of interest to be calculated within the same time period using the
same resources.
[0042] By using tolerances, the required degree of accuracy of
the scenario calculations can be balanced by the amount of calculations
to be done. Generally, reducing the tolerances will result in a
large group of selected policies. In addition, the results of a
scenario calculation will be known to be within the specified tolerances
of the result that would have been obtained by running the scenario
on the all the policies.
[0043] The number of selected policies identified at block 120
is generally substantially smaller than the total number of policies
and is typically about seven selected policies for every 100,000
policies. By performing scenario calculations on the selected policies,
substantial time and resources can be saved or reallocated as compared
to performing scenario calculations on all the policies or on the
grouped policies. Less policies results in fewer computers being
needed to perform the required calculations. In addition, the computing
time can be used to calculate additional risk measures, help calculate
the required information to support a daily hedging operation, and
otherwise improve the accuracy of existing valuation statistics.
[0044] Selected policies match the initial quantitative characteristics
at time zero, including the derived characteristics. In contrast,
for grouping, pseudo policies will be created to match quantitative
characteristics of each cell, but when combined, may not match the
same seriatim statistics, or when used in the valuation process,
match the reserves, capital, and sensitivity figures derived from
seriatim calculations.
[0045] Using grouping, may result in absolute valuation differences
greater than 5% versus seriatim valuations, as compared to typically
0.05% for valuation differences based on selected policies.
EXAMPLE 1
[0046] In the following example, the technique described above
will be applied to a set of policies. FIGS. 4a and 4b represent
100 example policies containing both intrinsic data, such as sex,
age, gender, contract maturity, account value, and benefit value,
and derived data, such as persistency at Time T and discounted value.
For the purposes of this example, persistency is a function of gender
and age of a policyholder using the following formula: Persistency=0.7+(gender)/10-age/1000
[0047] For the purposes of this example, the net present value
of a policy for a scenario can be calculated based on the persistency,
the benefit value, account value and the risk free rate. NPV=Persistency.times.max(benefitBase-accountValue,0).times.e.sup.-(riskF-
reeRate)T
[0048] In this example, a number of scenarios are included and
each has a different account value return ranging from 7% return
in scenario 1, to -7% return in scenario 10.
[0049] FIG. 5 indicates the expected group values for this example,
including the sum of age, gender, contract maturity, account value,
benefit value, persistency, and the ten scenarios across the 100
policies in our example. In this example, age, gender, contract
maturity, account value, benefit value, persistency, and the ten
scenarios are the quantitative characteristics.
[0050] FIG. 5 also indicates tolerances specified for each of the
quantitative characteristics provided in this case by the user.
[0051] With this information, the minimization problem may be solved.
As indicated in FIG. 6, the optimized result is within the provided
tolerances of all the quantitative characteristics and only uses
five of the policies as the final weights indicated. Although not
shown, the five policies may be used in further scenario calculations
to represent the characteristics of the all the policies.
System
[0052] With reference to FIG. 1, a system implementing the invention
may consist of a computer processor 5 that processes and acts upon
instructions contained in code. The processor 5 may consist of a
single processor, a series of processors within a single computer,
or a series of computers in mutual communication containing one
or more processors. As used herein, a processor includes or has
access to the memory 25 to perform the operations. In a preferred
system, the memory 25 may consist of primary, secondary and tertiary
memory.
[0053] The system further consists of a database or data store
10 containing information about one or more policies in communication
with the processor 5. The processor 5 can retrieve information from
the database 10.
[0054] An input device 15 in communication with the processor 5
enables the processor 5 to obtain either information from a user
of the system or information created prior to the system's operation,
such as a configuration file or input date file. An output device
20 is also in communication with the processor 5 and capable of
displaying or storing information from the processor 5. The output
device 20 may be a display device such as a computer monitor or
screen, a printer, or may produce text messages, email or other
electronic output.
[0055] One application for the invention is in supporting intra-day
hedging activities for complex risks like variable annuities, unit
linked or equity indexed annuity risks. It is difficult to calculate
the relevant information to successfully manage the risks for such
vehicles on an intra-day basis because of the calculation time associated
with seriatim scenario based calculations and because the risks
and values are sensitive to interest rate, volatility and equity
market movements that occur throughout the day. Often, an overnight
run is used to collect the necessary information but long run times
reduce the quality and breadth of information to manage such complex
risks. By selecting a small group of policies, more simulations
can be completed in the available time and additional quantitative
characteristics can be included to help understand and mange changing
risk profiles. Typically, calculations on the selected policies
take minutes to perform. With this additional information, more
accurate estimates can be obtained and better risk limiting measures
taken in a variable annuity hedging program.
[0056] In some regulatory environments, such as those in place
in the Canadian and United States marketplaces, very substantial
calculations must be performed for regulatory reporting on naked
variable annuity risks. For example, the calculation of naked capital
figures requires using 5000 scenarios in Canada, the use of pads
or conservative parameter estimates and a conditional tail expectation
with the worst 5% of all the outcomes, known as a CTE95. Regulators
generally prefer seriatim calculations which would require billions
of calculations to be performed in each quarter. Calculations using
only selected policies significantly reduce the number of calculations
that must be calculated.
[0057] A similar application can be found for the regulatory financial
reporting of hedged variable annuity risks in the United States
and Canada. In addition to the reporting requirements for naked
variable annuity risks, reporting requirements for hedged variable
annuity risks requires simulating the hedging strategy through time
and include the payoffs of the hedge portfolio when calculating
reserves or capital. This means that the value and sensitivity of
the financial guarantees embedded in the variable annuity contracts
must be found at each and every time step and path. Payoffs from
the hedge portfolio must be calculated and collected along with
the naked liability cash flows. This is an enormous calculational
burden that generally can not be done on a seriatim basis because
the valuation calculations are generally of a stochastic on stochastic
nature. By using the selected policies, such calculations are more
feasible and a process for selecting a small group of policies can
be articulated to relevant regulators. Quantitative characteristics
used for selecting relevant liability policies can include expected
cash flows through time, time zero values and sensitivities, and
individual value and sensitivity figures under specific market scenarios.
By using only selected policies, the complex calculations can be
completed at the end of each quarter on a timely basis, and important
time zero and other step information can be matched and reflected
in the selection process thereby producing more accurate regulatory
reporting results and perhaps enhanced capital relief.
[0058] Various embodiments of the present invention having been
thus described in detail by way of example, it will be apparent
to those skilled in the art that variations and modifications may
be made without departing from the invention.
|