Data - Chapter 19

Question 1

Define and give examples of personal data

Answer 1

• Definition: personal data is information which would allow an individual to be identified, either on its own or when combined with other information
• Examples: name, address, email, occupation, DOB, health status, race/ethnicity

Question 2

What can reduce the legal burden of personal data?

Answer 2

• Anonymisation: anonymised data, which removes the ability to identify an individual, can reduce the legal burden of a firm, who will have considerably less obligations

Question 3

Give 2 examples of how competition legislation limits how data can be used

Answer 3

• Anti-competitive agreements – e.g. data being shared among small number of companies to fix prices in market
• Abuse of dominant market position e.g. imposing unfair trading terms such as exclusivity – imposing restrictions on the use of a product originator’s data to approve new products

Question 4

Define and give examples of sensitive personal data

Answer 4

• Definition: information which is more private to the individual and is generally subject to much stricter regulation than other personal data
• Examples: race, political opinions, religion, membership of trade union, physical/mental health condition, sexual life, convictions

Question 5

Under what conditions may personal data be used?

Answer 5

PEAR
o Protect vital interests of individual or other person
o Explicit consent given
o Administration of justice/legal proceedings
o Required by law for employment

Question 6

Define data governance

Answer 6

• Definition: the overall management of the availability, usability, integrity and security of data

Question 7

What could non-compliance with a data governance framework entail?

Answer 7

PAIR
o Poor quality data – inability to rely on data for decision making
o Additional costs e.g. fines, legal costs
o Intervention by the regulator
o Reputational issues – loss of current + potential
future customers

Question 8

What should a data governance policy specify?

Answer 8

CARps
o Controls to be put in place to ensure required data standards are applied
o Adequacy of controls will be monitored on an ongoing basis
o Roles and responsibilities of individuals with regards to data
o Capture, analyse and PROCESS data
o Data SECURITY and privacy

Question 9

Define and give examples of big data

Answer 9

• Definition: big data comprises very large data sets, often brought together from different sources and which can be analysed very quickly
• Examples: motor insurance – driving data obtained from use of telematics or health insurance – analysis of data on diet from analysing purchases used on loyalty card/wearable fitness gear

Question 10

Considerations when dealing with big data

Answer 10

o Key feature of big data is to use all the data, which raises questions of whether it is excessive
o Variety of data sources often used prompts questions of whether personal information is actually relevant
o Organisations need to be clear with what they expect to learn/be able to achieve by processing the data, as well as the data being relevant/not excessive
o Organisations need to be transparent when they collect data, explain how it will be used
o Complexity of big data analytics is not an acceptable excuse for failing to obtain consent where it is required

Question 11

What are the risks of using data?

Answer 11

• FONT SHOP
• Form – available data may not be in a form that is appropriate for the required purpose
• Omissions/Errors – the available data may contain errors or omissions, which could lead to erroneous results or conclusions
• Not credible – data not credible due to being of insufficient volume
• Tail evaluation – insufficient data available to provide a credible estimate of a risk in very adverse circumstances
• Suitable for purpose – data not sufficiently relevant for intended purpose (may have been collected for another purpose)
• Heterogeneity – actuary divides group data into broadly homogeneous groups, which may be too small for credible analysis. If there is sufficient data in each group to be credible, may not be sufficiently homogeneous
• Output is suspect
• Poor projection – historical data may not be good reflection of future experience, owing to:
o Past abnormal events
o Significant random fluctuations
o Future trends not being reflected sufficiently in past data
o Changes in the way in which past data was recorded
o Changes in balance of any homogeneous groups underlying the data
o Heterogeneity within group to which assumptions are to relate
o Past data not sufficiently up to date

Question 12

Where is data used?

Answer 12

• A MAP FAIR PASS
• Algorithmic decision making
• Marketing
• Analyses of experience
• Premium rating/product costing/contribution calculations
• Financial control and management information
• Administration
• Investment
• Risk Management
• Provisioning/reserving
• Accounting
• Statutory returns
• Statistics of experience e.g. what are loss ratios

Question 13

Define and give advantages of algorithmic trading

Answer 13

• Definition: Automated investment trading that involves buying or selling financial securities electronically to capitalise on price discrepancies for the same stock or asset in different markets

• Advantages:
o Increased speed and efficiency of trading
o Can result in lower dealing costs
o Facilitate execution of complex trading strategies that previously would not have been possible

Question 14

What are the risks associated with algorithmic trading?

Answer 14

PEA
o Possible impact on financial system – 5-6% plunge and rebound in US equity indices within space of few minutes. Increasing integration between markets and asset classes means meltdown in one market impacts other markets/asset classes
o Error in algorithm/data leading to potential losses on each trade – issue when large number of trades completed very quickly
o Algorithm may not operate properly under adverse conditions e.g turbulent markets

Question 15

Sources for internal data?

Answer 15

Proposal form, claims forms, or data from similar products (issue may be different target market therefore different needs)

Question 16

Sources for external data?

Answer 16

• Reinsurers
o They provide cover to many insurers in the market and can thereforeprovide useful data to insurers wishing to launch/sell a product to new target market
o Aggregate data collected could be shared with insurer to assist with pricing and reserving, especially when own data and experience is limited
• Publicly available data – data available from published company accounts, regulatory returns, industry data or data from other countries
• National statistics – data collected in relation to population of a country and often includes data on inflation, employment and population characteristics
• Industry-wide data collection schemes
o Organisations collect data from their member offices and make available summaries of the data collected
o This data is suitable for setting bases or used in product pricing but not for valuing an individual policy
o Useful for small insurers or insurers writing new classes of business, as own data may be insufficient

Question 17

Why would a company use external data?

Answer 17

• Sufficient detail (insufficient detail in own data)
o You might want to use a new rating factor but you don’t have any data so you use the reinsurer’s data
• Quality
o It could be better quality than your own
• Competitive advantage
o What’s happening with competitors
• Quantity
o You might not have enough data due to:
 New type of contract never written before
 New entity
 Niche product – aiming at small market segment
 Small entity (i.t.o entire entity, the division writing the contract)

Question 18

What are issues with using external data?

Answer 18

• SHOP Q
• Summarised – data received is summarised
o No validation (yourself or unclear on how much was validated)
o Heterogeneity distortions
o Guarantees undervalued e.g. in GI, have value added services – this is difficult to see in summarised external data
• Heterogeneous
o Difficult to make homogenous
o If one company doesn’t collect in one field, can’t use any of the data
• Older
o Since data collected by 3rd party and cleaned
• Participation
o Not all orgs. Contribute and those that do not necessarily representative of market as a whole
o How good the data is depends on purpose
o E.g. if using data to find competitive advantage then ideally, would want data from biggest competitors

• Quality
o Depends on the quality of data systems of its contributors
o If data input is off poor quality, then output will be of poor quality
o If one mistake in data, can’t use any of it

Question 19

Data supplied from different organisations may not be precisely comparable because..

Answer 19

• COIND
• Coding used for the risk factors may vary from organisation to organisation
• Operate in different geographical or socio-economic sections of market
• Identical issues:
o Policies sold by different companies are not identical
o Sales methods are not identical
• Nature of data stored by different companies will not be the same
• Different practices between companies e.g. underwriting, claim settlement

Question 20

Problems of data quality and quantity can be a result of…

Answer 20

o Poor management and control of data recording or its verification processes
o Poor design of data systems

Question 21

Data typically needed to be captured in policy/claim records are…

Answer 21

• Details of risk or risks covered
• Details of cover (level of excess, maximum payout)
• Details of claim
• Status of present record (if claim is open, settled or has been reopened)
• Control dates (start, end dates of each record dates of claims etc.)
• Relevant amounts (sums insured, premiums, claims payments etc.)

Question 22

What is good quality data?

Answer 22

•	SUCCA
•	Sufficiently detailed
o	Homogeneity – every factor used to divide risks into homogenous groups identified
o	Purpose – significance of homogenous data depends on the purpose i.e. level of detail required depends on purpose e.g. if pension provided, age of individual is financially significant but if its payable to spouse on death of individual, spouse’s existence age not significant now
•	Up to date - relevance
•	Consistent 
o	Across assets and liabilities
o	With previous work
•	Complete
o	No omissions
o	Captured everything required
•	Accurate

Question 23

Where do things go wrong?

Answer 23

• Collection Stage
o Information is not requested
 NB info is not on proposal form sent out
o Respondent does not understand
 If on proposal form but they didn’t understand the question
 Wording of the form is important
o Information not given
 Respondent knows information but withholds it
• Capturing Stage
o Source and systems are not consistent
 Collection and storage inconsistent – M & F on form but 0 & 1 in system
 Language differences major problem
o Poor or inconsistent coding
o Poor training of data capturers
o Mismatched incentives
 Need to incentivise not only about quantity captured but quality too
o Multiple systems
o No checking processes
o No validation processes
 E.g. no checks to see if “Age” is a number
o No feedback from actuarial checks
 If find error, then need to report to people who handle the data. Otherwise it can’t be corrected

Question 24

What to check?

Answer 24

SUBSIST PANIC (compare to summary)