Topic 5 - Actuarial Modelling and Data Analysis Flashcards
What are Actuaries employed to do?
Solve financial problems
Give two examples of financial problems?
The premium to charge for a new product
How much money needed to fund a pension in
retirement
Actuaries use a series of tools that can be broken into:
Data
Assumptions
Models
What type of data do Actuaries use on a regular basis?
Raw data
Give two examples of raw data
Details of policyholders for an insurance company
Details of members for a pension scheme
What are some of the drawbacks with using raw data
Raw data can be difficult to manage & draw conclusions from
How do Actuaries ‘tidy up’ raw data?
Carry out types of data analysis to help tidy up data:
Descriptive –summary stats and simpler format
Inferential –using sample as proxy for whole population
Predictive –using past data to make future predictions
What steps do Actuaries take when working with data?
Actuaries use a number of steps when working with data:
- Develop objectives to be met by data analysis
- Identify the data items required for analysis
- Collection of the data from relevant sources
- Process and formatting data (access database or spreadsheet)
- Cleaning the data (looking for errors omissions)
- Exploratory data analysis
- Model the data
- Communicating the results (source, limitations)
- Monitoring the process (iterative process)
What is a model?
An imitation of a real-world system or process
How are models useful to Actuaries in life or pensions sector?
- Actuaries in life or pensions work over long timeframes
- Models allow us to study them in compressed time
- Allows actuaries to investigate or test outcomes without carrying out action or having to wait until they play out
Outline the iterative process used by Actuaries when building models
- Define objectives
- Plan modelling process and how model will be validated
- Collect and analyse data
- Define parameters and assumptions
- Build or write code for model
- Test the model and test reasonableness of output
- Review sensitivity of results to changes in assumptions
- Analyse output from model
- Adhere to any professional guidance
- Communicate and document results of the model
What type of model(s) will we use during this module FIN1013 Actuarial Maths 1?
Deterministic models
What is a deterministic model?
Key variable will take a definite value and is known in advance, producing a single outcome.
What is a stochastic model?
Key variable may take a range of values
(random)
An assumption is made about the statistical distribution of the key variable range
What are the benefits of using models?
- Models allow us to study complex problems in
compressed time - Allow us to study complex systems with stochastic
elements - Allow us to carry out multiple runs to see best fit
What are the limitations of using models?
- Model development requires time, money and expertise
- Models need to be maintained and updated
- For stochastic models need large number of runs
- Models more useful for comparing inputs than max outputs
- Can be lulled into false sense of security
- Models rely on quality of data
- Users need to understand model and limitations
- Hard to include all future events
- Can be difficult to interpret some of the outputs of model
What factors must be considered when considering the suitability of a model
- Objectives of the modelling exercise
• What are you trying to achieve? - Validity of the model, data and assumptions
- Impact and extent of correlation in random variables and results
• Matching of assets and liabilities principle
• if they move in same direction - Positively correlated - Relevance of previous models
• Can you update a previous model
• can this be used to validate - Credibility of data input and results output
• Test reasonableness of outputs
• Carry out sensitivity tests - Dangers of spurious accuracy
• Saying value of pension scheme is £9,876,543.21 say £9.9m - Ease of how model and results can be communicated
• Meet client needs and communicate with them in mind - Regulatory requirements
• Technical Actuarial Standards (TAS’s)
