Spring 2016 #24

carrotscarrots
in F-15-Evaluation

In the reporter's examination report, they said that many did the error to use latest diagonal reported and divide by the selected ultimate to get the %reported.

I thought that since it doesn't specify in the question that the ultimate was selected from the development method we should indeed use the above method because the development CDFs may not be appropriate.

This what is mentioned in the Wiki article for chpt 15.

image.png

But it also makes sense that if the question mentioned the ultimate uses the development method, we could divide the reported and the ultimate to get the % reported.

in what situations can when can we use (reported / selected ultimate) ?

Thank you!

Comments

  • GrahamGraham
    edited April 2022

    I think the examiner's report is wrong about this. The fact that this was listed as a common mistake and that the examiner's report provided such a detailed explanation indicates to me there must have been a lot of discussion about this. So here's what I think:

    • The question provides "Selected Ultimates" so the natural assumption is that these are best estimates, regardless of which method or methods were used. The examiner's report makes it seem like these selected ultimates were just any old numbers that could have been pulled out of thin air but that's just to justify them not accepting dividing reported losses by ultimates. That isn't a reasonable assumption based on the way the information was presented.
    • If you use the CDF method as the examiner's report suggested, it's possible the reported loss data was distorted and not even used in the final selection of ultimates, and in that case calculating %reported using the CDFs would also be distorted.
    • Now, the examiner's report is indeed correct when they say dividing the reported loss by the ultimates uses only the latest diagonal and ignores the rest of the triangle. This is a valid argument, but there is no optimal solution. Dividing reported losses by ultimates has potential problems but using the CDFs also has potential problems.
    • So it's just a matter judgment which way you do it. I don't think either is clearly better than the other.
    • Actually, a better way to do it would be to divide the whole triangle by the appropriate ultimates. Then you could select the %reported for each age as an average of each column. This has the advantage that all of the data is used and also takes into account the given selected ultimates.

    I will ink to this post from the wiki.

    Final point: If you look at 2016-Fall-Q27(b) you'll see that sample answer 2 somewhat contradicts the examiner's report for 2016-Spring-Q24.

  • carrotscarrots

    ahh ok this makes sense, thank you for the explanation!

Sign In or Register to comment.