Data, assumptions and modelling

Question 1

What are the risks associated with data?

Answer 1

GPCQF

• Data grouping (mix of members?)
• Data credibility (not enough?)
• Data is not good quality (incorrect?)
• Data is not in the correct format (detail?)
• Data not good prediction of future (abormal fluctuations etc.)

Question 2

Main uses of data

Answer 2

• Calculating assumptions for setting premiums
• Analysis of surplus
• Risk factors for underwriting
• Investment management
• Experience analysis
• Setting provisions for valuations
• For accounting
• Statutory reportings

Question 3

Data checks before experience analysis performed

Answer 3

• Reconcile total number of policies (previous and movement data)
• Reconcile premiums received and claims paid ( accounting data/previous data and movement data)
• Check two data sets against one another e.g. claims data and accounting data
• Reconcile experience analysis against premium analysis
• Check for unusual values e.g. no weird dates
• Consistency in averages between data sets e.g. average premium received
• Random spot checks for individual policies

Question 4

Data checks when it comes to A and L

Answer 4

• Check that A/L exists
• Check A/L transaction is owned at given data
• Check A/L recorded at correct value
• Check A/L recorded at the correct accounting period
• Check that A/L no ommissions

Question 5

What do we do if there is insufficient data in a homogenous group?

Answer 5

• Group it with another group i.e. summarise data
• May lead to distortions
• We group data in order to make experience more stable due to unique features in the group
• This allows us to make better predictions about the group
• Ensure balance between credibility and homogeneity

Question 6

What should we consider when making assumptions?

Answer 6

CNR PAS

• Consistency with other assumptions
• Needs of clients
• Regulatory requirements
• Purpose of assumption
• Accuracy of assumption
• Significance of assumption

Question 7

What’s the process of setting incidence rates using experience data from another insurer?

Answer 7

1. Check the data
   - Underwriting practices may be different
   - Sales practices may be different
   - Homogenous groups may be different
   - Data may contain ommissions/ errors
2. Split data into homogenous groups
   - Can be done by policy/ other risk factors
3. Set crude incidence rates
   - look at actual claims in comparison with exposure data
   - smooth rates over different ages to allow for variation
   - consider any future trends e.g. medical advances
   - determine how trends can be allowed for in final rates

Question 8

Factors that influence the number of model points

Answer 8

CHISING MODELS

• computing power available
• sensitivity of model points
• the purpose of the model
• time constraints
• heterogeneity of the model

Question 9

Factors that influence whether to buy/develop/modify model

Answer 9

• cost of each option
• the in-house expertise available
• flexibility of the model
• number of times model will be used
• level of accuracy required

Question 10

What affects the basis you choose for liabilities provision?

Answer 10

Needs of client
- Purpose of valuation 
- Needs of client 
- Regulatory requirements 
Nature of assets 
- if liabilities are linked to investment 
- if the covenant has no value 
- market-consistent liability valuation

Question 11

What is the problem with historical data?

Answer 11

HHOMer FART

• Homogenous group balance change
• Heterogeneity within group has changed (affects assumptions)
• Outdated past data
• Medical changes or social change
• Fluctuations in the past that are significant
• Abnormal events in past
• Recording of data changes
• Trends in future not reflected in past data

Question 12

What is the problem with industry-wide data?

Answer 12

• Not detailed
• Not all industry players contribute
• Out of date
• Data quality depends on quality submitted by players

Question 13

What is algorithm trading and associated risks?

Answer 13

• Investment trading that can happen automatically
• Quality depends on robustness of trading rules which depends on data
• More efficient and faster
• Execute complex trading strategies
• Algorithm trading is when assets are bought on sold in aims to take advantage of price discrepancies

Algorithm trading risks:

• Data may contain errors/ trading rules may have errors thus making losses instead of profits
• Algorithm trading may not be able to operate in adverse circumstances
• Big impact on financial market

Question 14

What are the reasons for heterogeniety?

Answer 14

• Different geographical locations
• Policies sold by companies are not identical
• Sales methods are not identical
• Different underwriting or claim settlement methods
• Risk factors stored differently for each organisation
• Nature of stored data is different

Question 15

What is big data?

Answer 15

• Big data is a large volume of data that has been obtained from different sources
• Technology has allowed us to collect and analyse these big volumes of data
• The analysis can happen very quickly

However, it is very important that data regulations are complied with especially when personal data is involved