Chapter 19: Data Flashcards
What is personal data?
- Personal data is data that can allow someone to identify a specific individual.
- Organisations have to respect personal data when making decisions
What are the consequences of breaching data protection acts?
- Criminal offences
- Prosecution
- Fine
- Jail time
- Reputational damage
One example of how data can be used by competitors, is data is against consumer protection laws?
- Sharing data among competitors to fix prices
What is sensitive personal data?
Data that if it is disclosed without the individual’s consent can cause an individual a high level of stress or damage.
Subject to much stricter legislation than normal personal data
List seven examples of sensitive personal data
- Racial or ethnic origin
- Political opinions
- Religious or other similar beliefs
- Membership in a trade union
- Physical or mental health
- Sexual life
- Convictions, proceedings and criminal acts
What are the three main characteristics of big data?
- Very large data sets
- Data brought together by different sources
- Data that can be analysed very quickly
- Anonymise to protect individuals relates to
- Data is relevant and not excessive
- Transparent when they collect data and explain to what use the data will be put - get consent to use data
- Held securely
What is data governance
A term used to describe the overall management of the availability, usability, integrity and security of the data in an organisation
What is the aim of the data governance policy?
A document that sets out the proper data guidelines
- Specific role and responsibility of individuals in an organisation wrt to data
- How data is captured, analysed and processed
- issues wrt data security
- Controls in place to ensure proper data standards apply
- How the adequacy of these controls will be measured
- Ensure legal requirements are met
What risks can an organisation face if the governance policy is not adhered to (4)
1 Legal and regulatory non-compliance
2 Inability to rely on data for decision making
3 Reputational damage
4 Ensuring additional costs such as fines
What risks are associated with the use of data? (9)
QUERIED
- Data contains errors or omissions leading to incorrect conclusions being made
- Insufficient credible data to provide credible results
- There might be enough data to provide credible results but not enough to give a credible estimate of an adverse circumstance, i.e., what occurs in the tails of the distribution
- External sources might not be relevant or appropriate for this circumstance
- Historical data might not be a good representation of the future experience
- Past abnormal event
- Significant random fluctuations
- Future trends not being reflected in past data
- Change in the way data was recorded
- Changes in homogenous groups
- Past data is not up to date - Difficulties in creating homogenous groups due to
- Groups being too small to be credible
- Merging with other groups can cause it not to be homogenous anymore - Data might not be in the appropriate form
- Data collected for a specific purpose, so it is not appropriate for this purpose
- Lack of confidence in the available data reduce confidence in final conclusion
Q - Quantity (credibility) U - Up-to-date E - Errors R - Relevance (heterogeneity) I - Incomplete E - Exceptionals D - Detail and format
List six areas of actuarial work where data would be required
- Setting provisions
- Pricing/setting contributions
- Investment management
- Risk management
- Management information / financial control
- Accounts/ statutory or supervisory reporting
- Experience statistics/analyses
- Marketing
What makes data poor quality?
- Errors
- Insufficient quantity
- Insufficient detail
- Lack of relevance
How can errors in data be curbed?
- Asking clear and unambiguous questions on the proposal form
- Carrying out data checks - ensuring now blanks or impossible values
- Reconciliation with previous years
- Cross-checking with other sources
- Consistency checks
- Random spot checks
How can an insufficient quantity of data be curbed?
Obtain external data - care to ensure it is relevant
How can an insufficient detail of data be curbed?
Ensure all required fields are captured
How can a lack of relevance in data be curbed?
Collated into homogenous groups (while ensuring data is still credible)
Adjustments to data (especially external data) to fit a specific purpose
What allowances can be made to poor quality data
- Include risk margins or contingency loadings - based on actuarial judgement
- Disclosed to client
- Warning on the extent to which data can be relied on
What needs to be considered when designing an insurance proposal form
General
- Proposal form must generate relevant information as it provides information for a range of purposes
- Questions should be designed and unambiguous so that the information given is easy to verify and interpret
- Easy to process by admin and underwriting teams, e.g., occur in the same order as they are loaded onto the system
- Any medical or occupational underwriting should be added
- Ask for information pertaining to each rating factor in personal lines GI
- Information will be sued to cross-check against claims information
Underwriters
- Sufficient information needs to be gathered for an accurate rating
- The form should identify risks for which further information will be required, e.g., further information on an illness suffered
Policyholders
- The questions should be in line with policyholder expectations and market practice
- Form shouldn’t be too excessive so that it is unattractive to policyholders
Regulator
* Form shouldn’t breach any regulatory guidelines, e.g., asking for genetic testing
What assertions can be examined relating to data
- Liability or asset exists on a given date
- Liability held / asset is owned on a given date
- Even recorded, the time the event occurs and the associated income or expenditure are allocated to the correct accounting periods
- Data is complete, no unrecorded liabilities, assets or events
- Appropriate value of an asset or liability has been recorded
How can data be checked?
Reconciliations
- Total members/policies
- Total contributions/benefits/premiums
- Beginning + On - Off = End
- Check movements inaccordance to accounting data
Cross checks
- Against other sources of data and any discrepancies should be investigates
- Check against accounting and custodian data
- Is data valid?
- How does it compare to previous years’ data
Consistency checks
- Contributions/Pension benefits/Sum assured paid should be consistent with the number of active members/ pensioners/ premium
- Average sum assured and average premiums to be consistent, if not we expect data to be missing
- Investment income should be consistent to the value of assets
Spot checks
- Unusual values, e.g., adding to many 0s
- Full deed audit
What are the 8 sources data?
TRAINERS
T - Tables (mortality tables0
R - Reinsurers
A - Abroad
I - Industry (e.g., companies collecting data from members and sharing it with other members
N - National statistics
E - Experience with existing contracts (internal)
R - Regulatory reports and company accounts
S - Similar contracts (internal)
Why should companies have a single integrated data system? (5)
- Reduce the chance of data being corrupted between different sources
- Reduced risk of data being used inconsistently through time
- Better controls on who can enter/amend data
- No need to reconcile different sources
- Information is easier to access
Problems with industry data?
DR DONEQ
D - Detail insufficient (usually summarised)
R - Risk factors coded in a different way, e.g., use of age bands instead of individual ages
D - Differences between insurers in the industry (heterogenous)
O - Out of date (e.g., collected only every 5 years
N - Not everyone contributes (not a picture of the entire market)
E - Errors in data
Q - Quality only as good as that of contributors
Causes of heterogeneity in industry data - cause differences - cause different claim experiences? (8)
Differences in:
- Econ/Social status
- Type of policies sold
- Sales method
- Practices (e.g., underwriting or claim settlement)
- Coding of risk factors
- Stored differently
- Geographical errors
- Contract terms
