[quote kurtrussellsbeard]@Scottishskifun thank you. Is it not the case though that they have a duty to look at the worst case scenario.
This is what all the drama was about prof Ferguson wasn't it. Part from his affair I mean. Incorrect statistical modelling? It must be relatively common? [/quote]
You need to be aware of the worst case scenario but you don't necesarily make all decisions on the basis of it, no.
To consider a very different example; let's say I'm modelling interest rates, so I'm projecting what interest rates might be in the future. I'm using a stochastic model so it produces thousands of possible outcomes and they're then analysed to give percentiles and stats. So I have the median, the mean, the range, the different percentiles, the volatility and so on. My charts might then show up to the 99.5th percentile which would be a 1 in 200 possibility of happening.
Let's say I'm projecting 10 years ahead. I could easily have a percentile chart which says the median interest rate is 2%, the bulk (say the 10th to 90th percentiles) of the results are between -0.5% and 4%, and the 99.5th percentile is 15%. Should I make decisions on the basis that interests rates will be 15%? No. You take it in to account, you plan how to manage the risk that it could be, but it's not your central scenario for decision making.
And in the process of doing that projection I will have had to make numerous assumptions and judgements - I'll need to choose which statistical model to use, I'll need to determine what each of the variable in that model will map to, I will need to chose which data to use to set the different values, some will be an expert judgement e.g. what is the realistic long term average for interest rates, others will be based much more on current data which involves knowing where and how to get good reliable data.
Ask 10 economists/actuaries/statisticians to model the same thing and you will get 10 different answers.