Chapter 18, 19 and 20: Modelling, Assumptions and Data Flashcards
The conditions of Popi
APPFIOSD
Accountability -The processor should be compliant
Processing limitation - fair and lawful processing
Purpose specification - PI only kept if there is a purpose for it
Further processing limitation - Further processing in line with initial purpose
Information quality - Accuracy and completeness holders responsibility
Openness - Documentation kept on all processing
Security - safe storage, processing only by aurthorised person
Data subject participation - Subject may request update or deletion of data
Data risks FHAG V
Form of data inappropriate Historic data not relevant Accuracy of data is poor Grouping of data may be inefficient Volume may lack (Especially to model extreme cases)
Uses of data PREMISES FAAr
- Premium rating, product costing, determining contributions
- Risk management
- Experience stats
- Marketing
- Investment
- Statutory returns
- Experience analytics
- Setting provisions
- Financial control, management info
- Admin
- Accounting
Sources of data TRAINERS c
Tables eg actuarial mortality tables Reinsurers Abroad (data from overseas contracts) Industry data National statistics Experience investigations on the existing contract Regulatory reports and company accounts Similar contracts Current economic conditions
Causes of poor data quality RUS RAD
Recording or verification
Underwriting Policy / Rating factors
System design
Insufficient:
- Relevance
- Amount
- Detail
Quality and quantity of data influenced by SEC
Size of business
Experience in line of business
Characteristics of the product (long tailed or short tailed)
Ways to Improve data quality TeCSReFS P
Technological resources to collect relevant real-time data Checks on data collected Systems to integrate newly collected and historical data Retain Record new data Robust Form improvement: -Proposal form RECTUREQ -Claim form Secure and sufficient storage Processing capacity
Ways to improve the claim and proposal form RECTURE Q
Relevant + reliable info collected Ease of analysis Cross-checking with claims form when processing claims Tick Boxes Underwriting info - LI Rating Factors - GI Ease of recording into the system Quantitative as far as possible
Assertions to assess in data TACA /
CATA OPI SpotS
That A or L exists or is held
Appropriate value of A and L
Complete data – no unrecorded A or L
Accounting period of events correlate
Opening + movements = closing Policy data vs Accounting data Investment data vs Accounting data Spot Checks for unusual values Summary statistics
Ways to check assertions: Checks to perform SCARFACE MVR
/
Ways to check assertions Reconcile, Consistency, DRaM
Spot checks Random check for weird values
Contribution(salary related)/benefit(in payment) consistent with accounts
Asset income consistent with accounts
Reconciliation of asset value between company and third party
Full deed audit on certain assets (like property)
Average sum assured and premium compared to previous investigation
Consistency of shareholdings (start and end of period)
Equate/reconcile member numbers and changes in them
Movement data against appropriate records( accounting data)
Validation of dates
Reconciliation of premiums and benefit amounts and changes in them
Reconcile NBA
Number of member and policies - previous and movement data
Benefit amounts and premiums and possible changes - previous + movement
Assets value correctly displayed - market value, and admissible assets
Consistency SACIS
SA and Premium in – class of business and previous
Changes in variable benefits and contributions consistent between policy and accounts
Investment income implied vs. accounting data
Shareholding
DRaM
Deed audit
Random spot check
Movement vs accounting data
For past data consider how to deal with:
CRC CHEAT PRUDT
Changes in recording of data
Random fluctuations
Changes in experience over time - Ideas!
Changes in balance of homogenous groups (Business Mix) Heterogeneity in groups Errors in Data Abnormal fluctuations Standard Tables: -Relevance and adjustments -National vs. Industry data
Changes in: Product design Rates changed (Mortality, withdrawal, investment) Underwriting practices Distribution channels Target market
Concerns with summarised data MiRC/ RelCanS
Miss significant differences between benefits
Reliability reduced
Cannot value client specific options or guarantees
Merits of industry wide data CA FLOQ HP
Comparison
Assess possible new business opportunities
Formats can differ Less flexible Out of date Quality will vary Heterogenity Not for provisions
Heterogeneity in data industry-wide data caused by difference in
PReCReS SNaP
Practices – Underwriting, i.e. Reinsurance Coding of risk factors Regulation Sales methods
Socio-economic
Nature of data sections
Policies sold
Different options when in need of a model:
NEC
New model built from scratch (outsourced or in house)
Existing model modified
Commercially purchased model
Factors to consider when choosing a model:
C U FLIP
Cost Usage frequency Flexibility Level of needed accuracy In house expertise available Purpose of the model
Requirements of a good model- VIVA RAPER DAD
Valid – relevant to purpose and economic principles assumed
Implement in various cases
Verify reasonableness of outputs independently
Appropriate inputs and outputs - data and parameters
Rigorous – Realistic under several circumstances
Avoid over complexity - costly, non-flexible
Parameters - allow for significant features in business and economic principles and assets and liabilities
Easily communicate and appreciate workings and results
Reflects risk profile of business modelled
Documented – Key assumptions stated
Adjustable Develop-able, Refine-able and Testable
Dynamic – Interaction between parameters and relevant variables that affect cash flows
Deterministic model development
SCCACCARD
- Specify purpose
- Collect, sort and modify data
- Choose the form of the model, identifying parameters and variables
- Ascribe the parameter values
- Construct the model based on the expected cash flows
- Check the goodness of fit is acceptable
- Attempt to fit a different model if the model does not fit well
- Run model using estimates of the values of future variables
- Do sensitivity testing on the parameter values
Stochastic model development
SCCSACCRP
- Specify the purpose of the investigation
- Collect group and modify data
- Choose a sensible density function for each of the variables to be modeled
- Specify correlation between variables
- Ascribe values to the variables not being modeled
- Construct a model based on the expected cash flows
- Check the goodness of fit
- Run the model many times each time using a different sample
- Produce a summary of the results that shows the distribution of modeled results
Deterministic vs. Stochastic models differences CUBE PO
Capital intensivity/Cost Understabability Build and run ease Economic scenario tested Parameters Output
Merits of Deterministic model LECE COS No
Less Capital intensive
Explainable - since no distributions are applied
Clarity on scenario tested
Easy and quick to design
Carefully consider which scenarios will be tested
Only point estimates produced
Some scenarios may be missed
Not good for valuing options and guarantees
Merits of a stochastic model WAQA SLICAH:
Wider range of scenarios tested
Assess financial guarantees/assumptions tested
Quality result / Quantify risks better - Statistical analysis performed on the results
Allows for uncertainty and covariance between parameters
Spurious accuracy
Longer construction and run time
Interpretation and communication difficulty
Complex programming/ Costly
Additional capital intensity
Higher risk of model and parameter error due to complex nature
Uses of a model PROF P:
Pricing - setting premium of charging structure Risk management Options and guarantees valuation Set Financing strategies Individual Provisions valuation
Using models for pricing consider BAD CoMPA DiSCo
Business strategy
Assess capital requirements
Discount rate used
Competitiveness
Model point used
Profit requirements
Assumptions of future experience
Distribution channel
Contract design
Size of market
Factors affecting assumptions: DIE FULCEN
When setting assumptions it is important to consider:
Demographic factors
Investment strategy
Economic factors
Financial significance of assumptions
Use of model/ Use to which the assumptions would be put
Legislation and regulation / Level of prudence
Consistency of assumptions / Correlation between assumptions
Expert guidance should be allowed for
Needs of client and company in terms or risk appetite
Factors that increase the risk of product design:
GLOC U
Guarantees and options Lack of historical data Overheads Complexity Untested market
When considering accuracy and prudence of assumptions BIPACS O
Best estimate vs Including uncertainty(Prudence) (overstatement)
Implicit assumptions:
-consistency of population distribution
-closed or open to new business
Purpose of the valuation
Accuracy of assumptions vs. Accuracy of outcome
Correlation between assumptions
Significance of assumption error
Once off cashflows
Assumptions made form historic data/ Current data
BIDS/ REIS FW
Benefit growth - past inflation
Investment returns
Demographic data
Salary levels and growth
Regulation and legislation Economic factors – central bank policy Inflation - Index linked bonds interest rates Scheme sponsor provides info on: Future salary increase Withdrawals
Demographic assumptions:
Morty Will Never FallS
Mortality/ Morbidity Withdrawal rates New business volumes/mix Future contributions Salary/Promotional increases
Examples of financially significant economic assumptions
5CIET - Also for monitoring!
Claim amounts Claim numbers Investment returns Expenses such as legal fees in court claims Commission Claims inflation Claims expense inflation Tail length of business
Contract Features Captured in a model point BAGTSAD
Basis Age Gender Term of policy Sum Assured Duration in force / Demographics
Setting the risk discount rate when modelling RASS
Required return of the company Average risk faced by business put into a single rate (simplicity) Statistical risk (Risk causing variation in cashflows and other model risks) Stochastic Discount rate
The risk discount rate (project appraisal) SyNC WACS
Systematic risk considered Nature of cash flows (real vs. nominal) Current cost of raising capital WACC (Explain fully) Alternative investments considered Compare riskiness – similar projects, other companies Sensitivity testing
Requirements on the data used RRASH VoC
High volume Stable Consistent - similar form, source and time Homogeneous Reliable Automated Regularly received
WACC Formula
(Market value of equity* Cost of equity)/(Market value of equity Plus market Value of Debt) + (market value of debt* Cost of debt)/(Market value of equity +Market value of debt)(1- Company tax rate)
WACC = (MVECOE)/(MVE + MVD) + (MVDCOD)/(MVE+MVD) *(1-T)
How to determine the Cost of equity (CAPM model)
COE = Risk free rate of return (Yield on G bond with similar duration) + Beta(sensitivity of equity to the market return) *(Market rate of return - Risk free rate of return)
COE = RF + B * ( RE - RF)
Usually around 10% for a listed company
Around 20% for a private company
Causes of differences between past experience form future events/company to company/products/before and after changes to a contract design CUBER DRIC
Contract design differences – the benefits provided (so many ideas!)
Underwriting – discuss the different rating factors to consider
Business mix (the demographic/rating factor differences – ideas!)
Externals – regulation, tax
Reinsurance agreements
Distribution channels
Investment strategy
Claims controls and experience
Rate changes
Data governance:
Definition
guidelines:
RCaSCM
ADACSIA SAS
A data governance policy is a documented set of guidelines for ensuring the proper management of an organisations data.
Guidelines:
The roles and responsibilities of individuals in the organisation with regards to the data
How data would be Captured, processed and analysed
Issues with respect to data security and privacy
Controls put in place to ensured that the required data standards are applied
How the adequacy of the controls will be monitored on an ongoing basis with respect to data usability , accessibility, integrity and security
Access to data controlled
Disclosure on the use of data
Accountability held of managers for data
Confidentiality agreements with service providers
Sharing of data controlled between departments, subs and industry
IT requirements
Anonymity in the data used
Structure of the data should be good
Auditing done of the data
Sensitive information protected
