This topic has come up several times in the last few months, most recently chatting with a fellow risk analyst Steve Jewell  who has 30+ years of experience in project risk analysis. Finding out he finds reasonable the approach I often advocate encourages me to share it with you.
Let’s say that you have an investment project that involves some initial construction with attendant cost and delivery time uncertainty, followed by uncertainty in the compensating cashflows. Let’s also say that your company does quite a few projects.
During the construction period, we’ll also say that a risk has been identified with a low probability (1%) of occurrence but an impact that would blow apart your project (let’s say it doubles the cost).
Now, if we perform a Monte Carlo simulation risk analysis model of the outturn cost, the histogram might look something like this:
(results generated using ModelRisk)
The blue bars show the cost distribution without the risk, the red line shows the total cost including the risk. The only difference is the small hump on the right. Viewed as a cumulative distribution, the difference is still very slight:
If we set a target budget for the project manager at say the P50, and a contingency equal to (P70 – P50) for example, the numbers that come out will be the same using either curve.
So, it wouldn’t make much sense to run a risk analysis with low probability, high impact financial loss risks together and then consider that one has accounted for such risk in the project’s budget targets.
My thought is this:
- Take out these types of risk from the cost risk analysis;
- Move them up to the corporate budget;
- Devise and implement a risk management strategy can be applied to reduce probability and/or impact (of course) – a bowtie approach can be very helpful;
- For the residual risk, take out specific insurance if possible and charge back to the project that premium as a fixed cost, or include it into the corporate insurance coverage if that’s feasible, and charge back a suitable portion of the corporate premium;
- If insurance is not possible, treat the risk as being self-insured, and charge an appropriate commercial premium to the project 
- Put the cost of the premium into the financial performance model for the investment (not the PM’s budget)
- Add the residual risk to the corporate risk register
Aside from the advantage of making the risk more visible at the corporate level, and of perhaps getting a more cost-effective premium, for a company that performs many such projects there is also a ‘portfolio effect’ – meaning that when you have enough such risks the aggregate loss distribution becomes more predictable (proportionally less spread relative to the average expected loss.
 Steve is a project risk analysis consultant. If you are searching for such assistance, I recommend you include him in your list.
 ModelRisk, the risk analysis add-in for Excel, has functions for calculating such premiums.