Principles of Financial Modelling

Sensitivity Testing

Aims to determine which parameter has the biggest impact on our output. Useful as the more sensitive the model is to a certain parameter, the more energy should be spent on getting accuracy on that parameters estimation.

Sensitivity Testing Helps us to:

• Develop an understanding of the risks faced

• Provide insight into the dependence of the output on subjective assumptions

• Satisfy a supervisors requirement.

Stress Testing

Like an. extreme version of sensitivity testing. A parameter’s value is drastically increased or decreased to determine if the output stays within a predetermined range.

Stress Testing Can:

• Help us to create plans of actions by assessing the possible stress impact both pre and post implementation of response strategy.

• Enable regulators to compare the impact of the same stresses on different organisations.

• Help senior management to compare the same impact on different business units within the same organisation

• Examine extreme events that might not have been considered if a stochastic approach was adopted

Issues with stress testing:

• It doesn’t align probabilities to the events considered

• Might look at extreme situations but doesn’t consider the full range of more likely outcomes that a stochastic approach can do

• The choice of assumptions to stress and the degree to which the relating parameters are stressed is subjective.

Scenario Analysis

When we think about a set of possible future realities and then change multiple parameters and inputs simultaneously.

Scenario Analysis Can:

• Helps us evaluate plausible future events

• Its not restricted to historical data and so can consider the vulnerability of high impact, low probability events

• Also helps us create plans of action by assessing the possible stress impact both pre and post implementation of response strategy.

Issues with Scenario Analysis:

• It is a complicated process

• As need to generate hypothetical scenarios that are extreme and plausible

• There is uncertainty on whether the scenarios are Representative or exhaustive

• Again no probabilities are assigned to any of the scenarios

Stochastic or Deterministic Model:

• Budget - stochastic modelling will likely be more expensive

• Time Available - stochastic modelling will likely take longer to build

• Problem - some issues are too complicated to be modelled deterministically such as options of guarantees, dealing with skewed loss distributions and allowing for interactions between variables

• Audience - How will the results be communicated?

Risk Awareness

Models can be used to determine the risk appetite of an organisation by incorporating utility functions, indifference curves and answers to a questionnaire about risk preferences.

Risk Identification

The risk register provides a list of risks as well as their dimensions, such as frequency and severity that will then house the output of the various risk models

Risk Assessment

Here risk models are created for each and every risk and used to estimate the frequency, severity and other dimensions of risk.

Risk Management

Models need to measure the modifications that risk responses and policies have made to risks’ frequency, severity and other dimensions

Risk Profile

All the risks need to be aggregated together using a model. This needs to be compared to risk appetite and if it is not aligned, action needs to be taken. This aggregated model can also be used to determine the amount of capital needed.

Risk Monitoring

A model needs to compare expected incidents to actual experience and be used to help analyse the difference and determine if it is significant enough to warrant action or feedback back into the risk awareness stage.

Enterprise Risk Management

Uses models at almost every stage to get details on individual risks and it uses them to get a holistic view of the organisation’s risk profile.

Downside Risks of using Models

Models require specialised skills and are expensive to develop.

Models with flaws can mislead the team and be costly

Even the right model can contain parameter risk which could have serious consequences if not picked up early.

Risk that the model is used/applied incorrectly or misinterpreted by Humans.

Mitigating Model Risks

Purchasing ready made commercial products or modifying/reusing existing models

Having the model peer-reviewed by a specialists.

Prevention of parameter risk can be done through data and estimation methods being audited and requiring checking from a supervisor.

Ensure staff are trained adequately to use the model effectively.

Steps of the Modelling Process

1. Objectives - Why are we doing this?

2. Plan - How are we doing this

3. Data - What is going to be modelled?

4. Parameters - Define what is needed and estimate appropriate parameter values taking into account any correlation between them

5. Version 1 - Define the model at a high level that aims to capture the essence of the real-world system.

6. Experts - Who is going to be involved? Provide feedback on the validity of the conceptual model.

7. Software - Where will we do this? Choosing a simulation package or programming language that can implement the model.

8. Program - Write the code for the model

9. Debug - Make sure it performs the intended operations defined earlier.

10. Test - Check the reasonableness of the output from the model.

11. Review - Make changes in the input and parameters and see how the model behaves. If deterministic: Sensitivity, Stress, Scenario Tests

12. Run and Analyse - If stochastic, simulate the model a number of times and analyse the output of the model.

13. Professional Guidance - Ensure that the model complies with the Technical Actuarial Standards on Data, Modelling and reporting

14. Documentation - Make sure the process, calculations and parameters have been documented and explained

15. Communication - Communicate the results of the mdoel in a manner that is appropriate for the audience so that they can draw conclusions

Use of Models

Models enable the risk professionals to give an organisation appropriate advice so that it can manage its risk in a sound financial way. They assist in the day-to-day running of a company by assisting in decions making and being part of the checks and controls system.

Model Use Cases

• Pricing financial products

• Valuing organisations and possible projects

• Estimating future volatility

• Determining Capital Requirements

• Projecting solvency position

• Assessing various risk management techniques

• Assessing different strategies, form investment portfolios to business decisions

What do Modellers need to consider when making a good model

Since models contain a degree of subjectiveness modellers need to balance various competing objectives.

General Requirements actuarial model

• Valid

• Adequately documented

• Rigorous, refinable, developable

• Appropriate inputs and parameters that allow for all significant features

• As few parameters as possible

• Arbitrage free

• Reasonable behaviour

• Not too long to run

• Understandable

• Communicable workings and output, need clear results

• Reflects risk profile of contracts being modelled

• Independent verification of outputs

• Correlation - sensible joint behaviour of the varaibles