Chapter 11: Data Flashcards
Main purposes actuaries need data for
- premium rating
- reserving
- determining the level of capital to hold
Reasons for lack of adequate data in general insurance compared to other areas
- actuaries are realtive newcomers to general insurance, so there have been fewer years to establish appropriate data collection for actuarial applications
- range and scope of data needed is greater given the rapidly changing and competitive nature of general insurance and complex statistical models that are used to set accurate premiums
Main sources of data
- Internal data
- External data from industry sources
Industry-wide data collection schemes
Organisations that collect data from their member offices and make summaries of all the data available to their members
South African Insurance Association (SAIA)
Examples of industry-wide data
- catastrophe model datasets
- flood maps
- CRESTA zones
- credit ratings
- premium rates
CRESTA
Catastrophe Risk Evaluating and Standardising Target Accumulations
Used to help assess risks relating to natural hazards, particularly earthquakes, storms and floods. Areas are classified into zones accoding to the likelihood of catastrophes occurring in those zones
Benefits of industry-wide data collection schemes
- insurer participating can compare its experience with the indistry as a whole at overall level and at level of the categories into which the data is classified
- insurer wishing to expand may want to understand how the characteristics of the business it wants to attract differs from its current business
- provides a benchmark for insurers to assess their position compared to their competitors
- industry-based development factors may be valuable as benchmarks when reserving
Problems with using industry-wide data
- potential distortions within indistry wide data, particularly owning to heterogeneity
- much less detailed and flexible than internal data and more difficult to manipulate
- more out of date than internal data because it takes a while to collect, collate and distribute to the insurers
- data quality depends on the data quality of the data systems of all the individual contributors
- not all companies contribute unless it is compulsory to do so. Thus the indistry data may not be a true reflection of the industry’s experience as a whole
Possible reasons for heterogeneity in indistry-wide data
- companies operate in different geographical/socio-economic sections of the market
- the policies sold by different companies are not identical
- the companies have different practices, e.g. underwriting, claim settlement and outstanding claim reserving policies
- nature of the data stored by different companies will not always be the same
- coding used for risk factors may vary from company to company
Why do industry-wide data schemes exist?
Managers of insurance companies use the data to confirm or refute suspicions from their own data. Also, anybody managing any business should be aware of what is going on in the market place.
Main uses of policy and claims data by a general insurer
- administration
- preparing accounts
- preparing statutory returns
- analysing performance
- informing investment strategy
- financial control and management information
- risk management
- reserving (including unexpired risk assessment)
- experience statistics
- premium rating and product costing
- marketing
- capital modelling
Data protection
- subject to explicit law in many countries, usually to protect people’s personal information
- many commercial policyholders provide information that is commercially sensitive
- any concern that the insurer’s systems are not secure can be highly damaging to the insurer’s reputation and business volumes
Data protection laws
- may cover what info a company may hold and for what it may be used
- require that specified people are appointed to be responsible for certain aspects of data gathering, processing or use, or for the correctness of data held
The full development team for a computer system should include representatives from which departments?
- senior management
- accounting
- underwriting
- claims
- marketing
- investment
- actuarial/statistic
- computing
- reinsurance
Main uses of data for senior management
Making business decisions
Main uses of data for accounting department
- collecting premiums
- paying intermediaries, claimants, etc.
- preparing summaries
Main uses of data for underwriting department
- premium rating
- identifying improvements
- evidence of selection
- portfolio monitoring
Main uses of data for claims department
Processing and settling claims
Main uses of data for marketing department
- assessing marketing performance
- identifying opportunities
Main uses of data for investment department
Monitoring investment performance and opportunities
Main uses of data for actuarial department
- premium rating
- reserving
- assessing solvency
- assessing capital requirements
- assessing investment strategies
- assessing reinsurance strategies
- management information
Main uses of data for computing department
Writing and implementing the IT system
Main uses of data for reinsurance department
Monitoring reinsurance performance and adequacy
Factors that influence the quality and quantity of data
The availability of data that’s good quality and quantity will vary greatly:
between organisations, which will depend on:
- size and age of the company
- current data system in use, including the use of legacy systems
- the integrity of the data systems
- management and staff responsible for collecting and maintaining data
- nature of the organisation, e.g. direct insurer vs reinsurer
within organisations:
- depending on the distribution method of the business
- between different classes of business
Impact of size and age of company on data quality
Large companies will have much more data available than smaller ones. They are likely to make more use of their own data, rather than rely on indistry-wide data.
Newly established company may have insufficient historical data for planning and reserving purposes - may need to supplement with indistry sources
Impact of size and age of company on data quality
Large companies may have better data systems in place than small ones.
However, a large well-established company’s computer data system may be outdated and difficult to amend, while a new small company may have a modern system that can be readily adapted to change and allowing better quality data to be recorded.
Reinsurer vs direct insurer data
Non-proportional reinsurance
- receive aggregate data relating to all of its risks coming from a particular cedant
- impossible to check accuracy to any great extent without assessing the original records of cedant (expensive)
Treaty reinsurance business
- sometimes only grouped bordereau data available
- may not be able to check and/or analyse its experience data to as a high a level as would like
Excess of loss reinsurance
- cedant may fail to realise a claim may exceed the retention and may fail to notify reinsurer
- claims remain IBNR for a while
- notify sooner - help reinsurer estimate ultimate recoveries more accurately
In general, bulk data often provided long after the claim occurred = relevance of data low, reducing reinsurer’s ability to assess future income and outgo
Bordereau data
A detailed list of premiums, claims and other important statistics provided by ceding insurers to reinsurers, so that payments due under a reinsurance treaty (or delegated authority schemes in direct insurance) can be calculated
Main distribution channels for insurers
- through brokers (intermediaries)
- through agents (e.g. banks or building societies that sell a certain insurer’s buildings and contents insurance)
- directly to customers
What brokers and agents will differ in
- the role they play in the sales, administration and claims processes
- their level and form of remuneration
- the manner and speed with which they process policies/claims
Reasons for data varying between classes
Principally due to different nature of risks, which leads to the following:
- big variations in claim frequency between classes affects the quantity of claims data
- the length of the tail of some classes means that it takes considerable time to collect the necessary claims data
- subjectivity used in underwriting influences the ability to capture risk details
Stages required in the establishment of a good data system to ensure good quality data is captured and stored
- consideration of the users’ requirements
- carefult design of appropriate proposal and claims forms
- ensuring that features of premiums and claims can be recorded
- consideration of policy and claim information to be collected
- adequate training of staff
Proposal form
- prime information source
- questions should be well designed and unambiguous
- excessive number of questions could lead to poor quality data if individuals provide inaccurate responses to complete form quickly
- information should be kept so that cross-checking against claim information can happen
- when information changes, a record should be kept of the old information
Features of premiums that should be captured by the data system
- amounts
- timings
- adjustments to premiums, such as premium discounts and commissions paid
- cross-selling information
Endorsement
An amendment to a policy during the term of the policy. Sometimes called mid-term adjustments.
Features of claims that should be captured by the data system
- type/cause of claim (peril)
- description of the claim event
- claims paid to date
- estimated outstanding claim amount
- claims handling expenses
- reinsurance recoveries
System should allow for multiple claims on a single policy and for reopened claims to be captured.
System should also be capable of recording claim aspects that apply to a group of claims and not to individual claims - class-level adjustments
Changes to the rule for setting up a claim record could affect:
- the number of claims recorded
- the number of nil claims
Attainment of majority
Where a payment is made once the claimant reaches a certain age that is pre-sepcified by the courts
Case estimate
When a claim is notified to the insurer, if the full amount of the claim is not paid immediately then it is common practise to estimate the amount that will still be paid on that claim - referred to as the case estimate.
Differences in recording case estimates
- when the estimates are first set up
- the method used to determine the amount
- how often these amounts are revised
Types of claims payments
- indemnity oayments made to policyholders
- compensation payments made to third parties
- payments to claimants’ solicitors
- payments to loss adjusters and payments of interest
Reasons for claims being reopened
- it may be purely due to the closure definition used by the insurer (for example, where claims are closed when it is deemed unlikely for there to be future payments relating to a claim)
- a further liability, possibly of costs rather than indemnity, came to light
- the insurer has made some recovery against a third party involved
- error was made in closing the claim originally
Data requirements for each policy record
- unique policy identifier
- person number/code to link policyholder information (name, ID, sex, etc)
- risk definition and details of cover
- policyholder’s risk factors used to record premium rate
-status of current record (normally in-force, expired, cancelled) - control dates (policy inception date, cancellation date, date of endorsement, etc.)
- relevant amounts and currencies (exposure/sum insured, premiums, excess, etc.)
- payment dates where applicable e.g. date premium received
- admin details
Data requirements for each claim record
- unique claim identifier
- policy number/code to link to policy information
- details of claim
- status of present record (open, closed, reopened)
- control dates (date of incident/reporting)
- dates and amounts of payments (incl. claim payments and recoveries from reinsurance and salvage)
- payment type
- dates and estimates of amounts outstanding, incl. movement data as estimates change. May include estimates of dates of settlement
- currency of both payments and outstanding claim amounts
- rating factor detail
What might the risk definition include?
- class and subclass of business
- the details of cover (sum insured, excess, etc.)
What might the details of the claim include?
- type of claim (e.g. in motor, bodily injury/property claim)
- claim cause code (type of peril)
- seperate field for description of claim event
In holding a vast amount of data, it is necessary to strike a balance between:
- the capacity of the system
- the cost of data storage
- the amount of data stored
- the level of detail at which they are stored
Examples of sources of data error
- wrong claim number
- wrong policy number
- wrong risk detail
- wrong claim date
- wrong payment dates
- wrong claim type
Consequences of wrong claim number
Details of the claim could be allocated to the wrong record, and hence to the wrong claim risk group.
Could result in charging incorrect premium rates for the affected risk groups, incorrect capital requirement for a class, etc.
Consequences of wrong policy number
If claim record picks up its risk details from the wrong policy record, these are likely to be wrong. The claim details may also be allocated to the wrong class, year, etc.
Consequences of wrong risk details
This could happen if the current in-force details are entered, rather than those at the date of the claim.
This might happen if the policyholder has changes their car/address in the meantime. An error could also occur if the policy conditions have changed, but the new policy continues to be stored in the original rating group.
Consequences of wrong claim date
Could cause the claim details to be allocated to the wrong year, distorting both the apparent numbers of claims for those years and their development patterns.
Could also mean that the claim will relate to the wrong risk details.
Common cause of this is entering the date the claim was notified rather than original incident date
Consequences of wrong payment dates
Some claims are settled by several payments, made on different dates. If they’re not identified seperately, then development patterns may be distorted
Consequences of wrong claim type
An analysis of the different claim types for a particular class will be required, since they behave differently. If claim types are not identified correctly, this distorts the seperate development patterns and average values.
Examples of sources of data distortions that aren’t errors
- changes in claim handling processes
- case estimates
- processing delays
- large claims
- return premiums
- claims inflation
Consequences of erroneous claims data
- false accounting
- inappropriate reserving
- pricing wringly
- failure to make recoveries
- general management mistakes
Check digits
Policy numbers are often designed so that the last digit is a check digit. It’s defined by a mathematical formula based on the other digits so that the wrong entering of a policy number is likely to result in the rejection of the transaction being processed, rather than being processed to the wrong policy
Effect of inadequate data
- If reserves calculated are incorrect - distorts reported results and tax payments
- If premium rates calculates are incorrect - lead to unprofitable/uncompetitive rates/anti-selection
How can data capturing errors be avoided?
- check digits
- minimum and maximum values
- data field integrity checks
- mandatory fields
- error reports
- culture and training
Risk classification
The process of grouping data according to certain factors to obtain a homogenous experience within each group with respect to the factor being analysed
Purpose of risk calssification (and reducing heterogeneity)
- make the experience in each group more stable
- ensure risks in each group have similar characteristics
- use the data for projection purposes
- heterogeneity in groups distorts the results
