Q2008 #6

[Digweed]

Question says there is an uneven distribution of business along other class dimensions not being analyzed.

In this case, after we derive the LR, shouldn't we rebase to (t1,a2) before multiplying by the current relativities?
Otherwise, how do we adjust for dimensions not considered?

[Present Value]

I think the current raletivities given are the current average relativities across other class dimensions. So as soon as LRs have been multiplied by them they are now adjusted for class distribution differences.

However, I agree that I would rebase to the base class given, but that I don't think has got anything to do with uneven class distribution.

[Digweed]

2002 Q45 rebases so I thought that is the norm, I really start to hate this paper now.

[Ironside]

Originally Posted by Digweed

2002 Q45 rebases so I thought that is the norm, I really start to hate this paper now.

I think re-basing ends up being equivalent to multiplication by a scalar, across the entire minimum bias equation. As such, you can do this at the beginning or the end of the calculation, without altering the outcome.

[Present Value]

Only if convergence occurs at one iteration. Otherwise, the first iteration (or non-converged relativities) are not the same before and after rebasing

I would always rebase before iteration if I know what I need to rebase to...and then keep rebasing at each step since ietration doesn't guaranty base will be 1 at each step

[Digweed]

CAS would accept answers w/ or w/o rebasing as long as you state your assumptions correct?

[Present Value]

I would think so

[rebeccap]

http://www.actuarialoutpost.com/actu...d.php?t=174286
Is this what you get when rebasing? What would the assumption be for not rebasing?

[Present Value]

If you just do Loss Ratio times Curent relativities then what you end up with is loss ratios. These are not relativities. If the question says what are the realitivities, you would need to "pull out" the base class loss ratio to make your answer a relativity.

If you do Loss Ratio*CR/Base class LR then you have relativities right upfront adjusted for distributional biases. If you proceed with this you will end up with loss ratio relativities.

Does that help?

[rebeccap]

Originally Posted by Present Value

If you just do Loss Ratio times Curent relativities then what you end up with is loss ratios. These are not relativities. If the question says what are the realitivities, you would need to "pull out" the base class loss ratio to make your answer a relativity.

If you do Loss Ratio*CR/Base class LR then you have relativities right upfront adjusted for distributional biases. If you proceed with this you will end up with loss ratio relativities.

Does that help?

Thanks for replying, but I am not sure I get your point...my answers from the other thread use L/R relativities adjusted to base class, as you described in your 2nd paragraph above. I am really inclined to say the official answer is wrong for not making the base class adjustment, but I'd like to see if anyone out there disagrees with me and has a valid assumption/reason for not rebasing.