19.Data Flashcards
Personal data
Information allowing individual to be identified, either on its own or in combination with other info
Sensitive personal data
Info which disclosure to others without consent can cause high level of distress/damage
Circumstances under which sensitive personal data can be processed
Explicit consent given
Required by law for employment purposes
Protect vital interests of individual/individual/another person
Needed for administration of justice/legal proceedings
Characteristics of big data
- Large data sets
- Brought together from different sources
- Can be analysed quickly
Big data consideration
- May be exessive/irrelvant
Data governance
Overall management of availability, usability, integrity and security of data employed in organisation
Data governance risks
- Legal and regulatory non-compliance
- Can’t rely on data to make decisions
- Reputational issues
- Additional costs from fines etc
Data risks
- Inaccuracte/incomplete
- Not sufficiently relevant for intended purpose
- Not reflect future experience
- Chosen data groups not optimal
- Not available in appropriate form for intended purpose
- Not credible due to insufficient volume, particularly due to estimation of extreme outsomes
Reasons why data may not reflect future
- Past abnormal events
- Once-off impacts
- Future trends not sufficiently reflected
- Changes in way past data was recorded
- Significant random fluctuations
- Changes in balance of any homogeneous groups
- Heterogeneity with group to which assumptions relate
- Not up to date
- Other changes e.g. medical, social and economic
Algorithmic decision making
Automated trading involving buying/selling of financial securities electronically to capitalize on price discrepancies for same stock/assets in different markets
Data requirements
Must be controlled through single, integrated system
Advantages of keeping data in a single system
- Reduced chance of corruption
- Reduced chance of inconsistent treatment of information
- Better control over who may change or enter info
- Easier access to info
- No need for reconciliation between systems
Sources of data
Public data
- Publsihed accounts
- Overseas data
- National statistics
- Industry data
Internal data
Reinsurer
Industry-wide collection schemesReasons why data from industry collection schemes may not be comparable
- Operate in different geo/socio-economic sectors of the market
- Non-identical policies sold
- Non-identical sales methods
- Different practices e.g. underwriting
- Differences in nature of data stored
- Differences in coding used to code for risk factors
Other problems with data from industry wide collection schemes
- Data may be less detailed/flexible
- Data may be out of date
- Data quality may be poor
- Not all companies contribute, therefore not representative of whole market
Checks on data
• Past data can help verify current data
• Accounting data is useful to help verify income and outgo + value of assets
• Data on individual assets could be checked and verify:
- Existence of assets
- Allowed to be held for valuation purposes
- If valuation is restricted by legislation/regulation
Assertions to check quality of data
- Reconciliations of member/policy #s
- Reconciliations of benefits + premiums
- Reconciliation of beneficial owner and custodian records where assets are owned by 3rd party.
- Records picked at random spot checks
- Consistency between contribution and benefit levels with accounts
- Consistency between average sum assured + premium for each class, and when compared with previous investigations
- Consistency of asset income data and accounts
- Consistency between start and end period shareholdings
- Full deed audit for certain assets e.g. property
- Validity of dates
- Movement of data against accounts
Lack of ideal data
- Insufficient volume to provide credible result
- Data may not be captured at a sufficiently detailed level
- Actuary may only have summarised data …
- … this is not suitable for all valuation purposes
Sources of poor quality data
- Poor management control of data/verification process
* Poor data system design
Mechanisms that can be used to ensure good quality data
- Proposal form
- Claim form
- Input of data onto system
- Other
Proposal form
Must be designed to:
Collect data at appropriate level, incl data not currently used but may be needed in future
Clear and unambiguous to give correct information
Have inputs be as quantitative as possible
Claim form
Must be clear and unambiguous and must link to proposal form so cross-checking can be done
Input of data onto system
Inputs must be in same order as in proposal form so person inputting info doesn’t need to interpret info
Staff inputting info must be well trained
Financial incentives for accuracy
System must have validation checks, e.g. checks on
- blank entry fields
- sensible entry values e.g. sensible bounds on ages and sum assureds
Insurer may send policyholder key info for verification
Other features of good data system
System must be capable of storing info, so that historical data can be used for future pricing exercises
System must be robust and flexibles
Secure- many can view but not many can amend
At regular intervals, checks of movement analyses must be carried out and checks of changes in policy details, e.g. how sum assured is changing from year to year
Use of proposal form to assess claims
- Cross-check against claims info at time of claim to check validity of form
- Can also check endorsements (changes to policy)
Good quality data
- Accurate
- Complete
- Up-to-date
- At sufficient level as required
- Consistent with past data
