It's a simple idea, somewhat obvious when you stop to think about it, but risk attitudes aren't stationary. I'm using the term 'stationary' in the statistical sense: if stationary, it doesn't matter when you make an observation; the observed phenomenom is time invariant.
So, if not stationary, then attitudes must change with time; and indeed, they do. This is one interpretation of the cone of uncertainty:
- In the far future, when there's plenty of time to make an adjustment, and all options are on the table, we're more optimistic about a risk (it might be a big deal if it happens, but we'll probably be able to fix things before it does)
- In the near future, when most of the options are off the table, we're more pessimistic that we'll be able to fix whatever goes wrong.
- In between these two, we're more or less centered: might happen, might not; either way, we can deal with it.
You can see that it's the usual triangle model for risk, skewed towards pessimism, so that the expected value is a bit more pessimistic than the single trial most likely outcome. Of course, the real world is not a world of triangles; but the triangle is a good model nonetheless because it's mathematically simple and the central limit theorem tells us that the distribution model is irrelevant to Monte Carlo results. So, we might as well use something simple!
Now, put on the temporal dimension and you get something like this:
1. Estimates about the far future are usually overly optimistic, leading to underestimates and then overruns. The sales staff on the project are notorious for this practice; who's not heard: Win it first, then fix it!
2. Keep an eye on those expressing a risk attitude: sponsors, stakeholders, even cost account leaders. Their notions, and thus the risk register itself, are not stationary. Everyone says risk management is constantly iterative, and so it is, and this is one of the reasons why so.