Any probabilistic model is going to project distributions. It's presented here as "cones of uncertainty", but we get the same with percentile results from any Monte Carlo modeling in financial models, too. I usually plot it as line graphs of specific percentiles, such as the median, the 25th and 75, 10th and 90th, etc. percentile results from simulation...
I can think the uncertainty in parameterization of their chosen model may be driving a lot of that width, fwiw.
And yes, for some models the 10th - 90th percentiles (or whatever) can give you very broad results.
I remember talking with the appointed actuary on one set of business re: IBNR and I said "well, there is a range of legit IBNR we could give... but I'm not telling the CFO the range".
I haven't seen cones of uncertainty before. These seem like very wide variation possibilities. Is that usual?
For projections into the future?
Any probabilistic model is going to project distributions. It's presented here as "cones of uncertainty", but we get the same with percentile results from any Monte Carlo modeling in financial models, too. I usually plot it as line graphs of specific percentiles, such as the median, the 25th and 75, 10th and 90th, etc. percentile results from simulation...
I can think the uncertainty in parameterization of their chosen model may be driving a lot of that width, fwiw.
And yes, for some models the 10th - 90th percentiles (or whatever) can give you very broad results.
I remember talking with the appointed actuary on one set of business re: IBNR and I said "well, there is a range of legit IBNR we could give... but I'm not telling the CFO the range".