What does internal data include?
Includes:
Policy data (from proposal forms)
Premium data
Claims data
One form of external data is industry wide data. Explain this? benefits and cons
in the UK, the ABI collects and collates a wide variety of insurance data.
Mainly benefical for insurance companies 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.
pros: increases data quantity, allows benchmarking
cons: may reduce data relevance, consistency, and quality due heterongenity.
-The data will be much less detailed and less flexible than those available internally
-External data is often much more out of date than internal data.
-not all companies contribute
-The data quality will depend on the quality of the data systems of all its contributors so one mistake of one company will invalidate the whole data, more companies contributing means more likley to havew mistakes
Heterogneity issue with using industry wide data?
-companies operate in different geographical or socio-economic sections of the market
-the policies sold by different companies are not identical
-the companies will have different practices; for example, underwriting, claim settlement and outstanding claim reserving policies
-the nature of the data stored by different companies will not always be the same
-the coding used for the risk factors may vary from company to company.
What are the uses of the policy/claims data?
administration accounting
statutory returns
investment strategy and performance analysis financial control and management information risk management
reserving (including unexpired risk assessment) experience statistics
premium rating and product costing marketing
capital modelling catastrophe modelling.
Why data should be controlled by one single system?
-reduced chance of existing data being corrupted
-reduced chance of inconsistent treatment of information
-better level of control those who may enter or amend data
-easier to access info
- wont need to spend time reconciling data from different systems
Level of data required?
high level overview –> agregated and published data for accounts
strategic / operational decisions –> more detailed amangement data for profitbility by class
pricing and data –> individual risk data
Restrictions on the use of data are customer information and data protection/security
customer information
-need for idenfication
-support with cross selling/ customer lifetime analysis
-may require combining data across diff product systems
data protection / secuirty
-legal requirments such as what data may be held, how may it be used.
-breaches will lead to criminal offense and rep damage
consent for peronal data
-third parties must follow guidelines, delet data after use
-insurers must maintain secure system with password and safe storage and transmission of data. use only for appropraite purposes.
Users of data and their data needs.
User Main Use of Data
-Senior management–>Strategic decisions, business planning
Accounting–>Premium collection, claims payments, financial reporting
Underwriting–>Pricing, selection, portfolio monitoring, identifying improvements
Claims–>Timely and accurate claim processing & settlement
Marketing–>Performance assessment, targeting opportunities
Investment team–>Monitoring asset performance, supporting investment strategy
Actuarial–>Pricing, reserving, solvency & capital modelling, reinsurance strategy, management information
IT (Computing)–>Build, maintain and control data capture systems
Outwards reinsurance–>Tracking reinsurance use and adequacy, performance analysis
Risk management–>Identifying concentrations, risk controls, exposure monitoring
Catastrophe modelling–>Quantifying catastrophe risks and accumulation
why data matters for pricing
Pricing actuaries depend on accurate data from all other areas nad likely to be involved in the techincal easepcts
Poor data handling in one function can cascade into:
Incorrect premiums
Incorrect reserving
Incorrect capital requirements
Effective pricing requires collaboration and consistent data flows across the busine
The availability of data of good quality and quantity will vary greatly between organisations, within organisations and between classes of business
Between organisations
-Size & age of company
-Quality and compatibility of data systems
-Strength of management & staff on collecting and maintain data
-Nature of business (direct vs reinsurer)
Within organizations
-Depends on distribution method of business
between different COB
-Class of business
What are the reasons for the variations by organisations?
impact of size and age of company on data quantity
-Large insurers → more data → can rely on own experience
-Small / new insurers → limited history → rely on industry data
-Long-tailed classes take years before enough credible data exists
impact of size and age of company on data quality
-Large firms often have better systems, but may be outdated so difficult to amend
-New firms may have modern, flexible systems but lack history
- Building a new system is costly, slow, and requires parallel running. new system may be a big project as expensive and time consuming.
-larger companys from merging /acquistion may see harder legacy system difficutlties when intergtaing two or more data systems with different data items.
-Legacy systems issues strongly affect analysis:
Mergers → different structures & data items
Hard/impossible to transfer all historic data
Often two systems must run:
-Better system for new business
-Original systems for existing business
Full actuarial usefulness may take years until enough data exists
Implications for actuaries:
-Allow for approximations
-Allocate more time to reconcile/clean data
-Integrity of systems - To ensure quality:
Data should be entered once only
Data should be Entered accurately
Data should be Backed up and protected from corruption
Procedural + system-based controls essential
Management and staff
-Poor controls or awareness → low-quality input
-Budget constraints may cause under-investment in systems
-Actuarial involvement in design improves relevance & accuracy
-Good systems take time before sufficient history is built
RI vs Direct insurer
Direct insurer - Detailed, individual risk data, Data easier to validate, timely data
RI - Often receives aggregated/bordereau data, Accuracy harder to verify, Data may be delayed/out of date
Extra complication for excess-of-loss:
-Cedant may fail to report claim that may breach retention → remains IBNR to reinsurer
-Often require reporting when claim exceeds 50% of retention
insurers can distribute business in 3 distribution ways
brokers
agents
directly with customers
differences in brokers/agents can arise due to Role in sales admin & claims, Remuneration
and speed of prcoessing
Role in sales admin & claims
-Delegated underwriting/claims authority
-Insurer may only receive summarised (bordereau) data
-Authority levels vary → inconsistent data
different rumeration
-Impacts brokers’ motivation to provide timely / detailed data
-Bordereau formats may be inconsistent
Speed of processing
-Paper-based → manual input → delays + errors
-Only large losses entered individually → smaller claims grouped in bulk
-Electronic feeds improve quality and timeliness
Generally which distribution channel has better data quality?
Direct - Higher quality and more detailed.nsurer captures all information directly, often electronically
whereas brokers/agent Lower quality, less detail, slower. Bulk data, delegated authority, manual processes, different data standards
Data quality depends of class of business which is dependent on frequency, length of tail and subjectivity.
Frequency
-Higher frequency → More data. better credibility
Length of tail - Long-tailed classes take years before adequate data develops
Slow notification & development → delays analysis and pricing
Statisical VS judgemental
- better when stastical factors used in underwriting . motor insurance has stored ratign factors. more credible
-judgmental underwriting less data quality. for specialty e..g marine varies by risk
data sstems better for when modern or legacy
modern, integrated is better for data quality rather than legacy.
primary objective of an insurer is that an computer system can be used effectively. give exmamples
Well-designed data capture forms → clear, objective questions
Staff training → correct input and awareness of importance
Parallel running of new vs old systems → confirm reliability
Ongoing performance monitoring → fix errors/improve process
Main stages of a good information system to ensure that good quality data is captured and stored
Identify users’ requirements
Design proposal & claim forms
Ensure capability to record premium & claim features
consideration of premium/claims data to be collected
Provide adequate staff training
WHat claims and premium need to be recored?
Premium
-Needed for pricing, reserving, monitoring.
-Record accurately:
Amounts (written/signed gross & net of RI)
Timings (due and paid dates)
Adjustments (endorsements, NCD, reinstatement premiums)
Commission (percentages, intermediaries)
Other deductions (discounts)
Cross-selling indicators (to evaluate loss-leader strategies)
System must store both original & adjusted premiums and be able to track overdue payments.
Claims
Claim definition (when opened/closed → affects frequency trends)
Outstanding amount (estimate + update history)
Set at notification or later
Retain previous estimates for development triangles
Multiple payments
Track date, amount, type, currency
Record recoveries (salvage, subrogation, RI) as negative payments
Reopened claims
Keep original claim reference
Store original closure date → monitor reopening rates
Claims handling expenses
Either separate or combined — must be consistent
Reinsurance recoveries
Link to claim where possible
Distinguish paid vs outstanding recoveries
Record reinstatement premiums
Class-level adjustments
Must allow for:
Pure IBNR (events occurred but not reported)
IBNER (reported but inadequate estimates)
Consistency over time is essential for actuarial trend analysis.
USers requirments in more detial
-Understand what each department needs (pricing, underwriting, claims, finance…)
-Conflicts may exist → need compromise
-System must be: Compatible across organisation, Integrated between functions
design proposal and claims form in more detail?
-Primary source of risk information
-Questions should be: Relevant, unambiguous, objective
Avoid excessive questions → reduces data quality
Store info and its changes → maintain full history
-Link to claims via policy reference number
- Keep history so claims data and exposure match for analysis
claims -Main source of claim cause/details. Must link correctly to the policy. Clear questions to allow automated verification of cover
why does it matter to have a good system?
High-quality information systems:
-Enable accurate pricing and reserving
-Support efficient operations (claims, underwriting, finance)
-Provide reliable MI for business decisions
Poor systems → inaccurate data → pricing errors & solvency risks
core info to record for each policy/claim?
Risk definition & cover details (class, options, sums insured, excess)
Claim details (cause code, type)
Status (policy: in-force/expired/cancelled | claim: open/closed/reopened)
Control dates (policy start/end; claim notification/payment)
Money & currencies (premium, exposure, claims, recoveries)
Administrative details (narrative if needed)
Sources of errors
Wrong claim number, wrong policy number, wrong risk details, wrong claim date - wrong year means distorts freuency and development patterns , wrong payment dates, wrong claim types, wrong cause of claim
These can lead to incorrect prcing/reserving
-
Chapter 1 - insurance companies21
-
Chapter 2 - insurance products background34
-
Chapter 3 - Insurance Products - Types33
-
Chapter 5 - RI Products -background12
-
Chapter 6 - RI products (Types)21
-
Chapter 7 - General insurance markets17
-
Chapter 8 - External Environment36
-
Chapter 9 - Risk & Uncertainty7
-
Chapter 10 - Data38
-
Chapter 11 - aggregate claims distribution models8
-
Chapter 12 - Rating Methods and bases48
-
Chapter 13 - further consideration when rating28
-
Chapter 14 - rating using frequency severity and burning cost approaches36
-
Chapter 15 - rating using original risk curves38
-
Chpater 16 - Generalised linear modelling29
-
Chapter 17 - use of multivariate models in pricing14
-
Chapter 18 - credibility theory15
-
Chapter 19 - Actuarial Investigations45
-
Chapter 20 - Reinsurance pricing39
-
Chapter 21 - use of cat models9
-
Exam Quesitons16
