Guidance notes - Risk modelling

This section concentrates on the risk modelling exercise after the risks have been quantified.

The first consideration in designing the model is how the risks should shape the structure of the model. For example, a cash flow model might be normally modelled in yearly units; however, the risks may well be quite different in summer than in winter. For the risk analysis it makes much more sense to separate the years into halves or quarters. This is a matter of judgement, but in large risk projects simple prototypes are constructed with different levels of detail to see what the impact the model structure has on the outputs. Another factor to consider in structuring the model is the timing of the risks and when they are likely to occur.

Simple valuation technique

The Public Sector Comparator Guidance provides an example of the simple valuation technique, and this guidance material provides an example in the financial model (simple risk valuation) in Appendix C. The risks are modelled from the risk estimates detailed in the Risk assumptions worksheet and are categorised into retained and Transferred Risk. In this example, timing flags are used to model the risks. However, other modelling techniques are available which can achieve the same result.

Advanced valuation technique

Monte Carlo simulation works by selecting a random value within the described probability distribution such that, over a large number of iterations, the distribution of the selected values reflects the input probability distribution. For example, if there a discrete distribution with a 20 per cent chance of a '0', a 50 per cent chance of being '1' and a 30 per cent chance of being '2', for each iteration the simulator will select either '0', '1' or '2', and after a large number of iterations, approximately 20 per cent of the values will have been '0', approximately 50 per cent '1' etc.

As mentioned in passing above, the accuracy of the estimates of the output parameter (i.e. the particular risks or risk category) depends on the number of iterations, and not the number of inputs, as the greater the number of iterations (as described above), the more likely an output distribution is formed with the risk probability estimates as described by the risk expert. Also Monte Carlo simulation, unlike most simpler methods, does not require that the relationships between the inputs and outputs are linear, i.e. do not involve division, multiplication or IF statements. It is for these two reasons that it is such a powerful and widely used method.

Example of risk quantification - advanced valuation technique

This is an example of how a risk is quantified for the advanced valuation technique.

(a) Phase one and two: A risk identified during the initial risk workshop is 'risk of adverse geological ground conditions'. The workshop participants assessed this risk as having a 'low' probability of occurrence and 'high' capital cost impact if the risk were to eventuate. The risk was assessed as a Transferred Risk under a PPP methodology.

(b) Phase three and four: The risk expert undertook a further review of the risk and placed it within the 'low' probability range at 15 per cent probability of occurring.  Its three-point estimate of 'best case' 'most likely case' and 'worst case' were $300 000, $375 000 and $700 000.  Note that this estimate straddles the 'high' and 'medium' boundaries set in the risk workshop.  The risk expert estimates that the risk would occur once (or it may not at all) during the construction phase, i.e. if the risk were to occur at the beginning of the construction phase, it would not occur again.

(c) Phase five: These details are entered into the risk register as shown in the Risk assumptions of the worked example financial model (advanced risk valuation method, refer to Appendix D).

The worked example shows that when the risk simulation is run, this risk will only occur once (if at all) during the construction period. During simulation the Monte Carlo function will select a value of zero 85 per cent of the time. The remaining 15 per cent of the simulation runs will be a value from the distribution as described by the risk expert.

The outputs of the simulation in this worked example are the NPC of the Retained Risks and Transferred Risks in total, as the PSC is expressed as retained and Transferred Risk shown in total. However, it is possible to select every risk as an output from a Monte Carlo simulation if required.