free cover

[Becoming An Actuary]

the question where part b) said the mgmt team thought no premium should be charged for the 10M X 10M layer....

was a) 24ish%?

I couldn't remember if we had to apply the loss trend to premiums as well? I think we did

[Becoming An Actuary]

also for part b) did you suggest price per million approach?

[The Sad Man]

Yeah I debated doing that for a long time because I remember All10 does that with construction cost trend or whatnot. In the end, I decided not to apply the loss trend to premium, although it was a coin flip.

[The Sad Man]

also for part b) did you suggest price per million approach?

That's a good answer. I think I said something about using XS ILFs to price the higher layers since that's what I do on my job.

[Sox34]

That's a good answer. I think I said something about using XS ILFs to price the higher layers

that's what i said too :)

[Warren Schmidt]

I think I said something about experience rating the lower layer and then multiplying that rate by the relativity of the exposure rated higher layer compared to the exposure rated lower layer.

[KidCA]

I said to use exposure rating and relativities based on the experience rating of the layer with loss info. Since the question didn't have much detail, I think the price per million answer would get some or all credit.

[Levin]

1. I'd agree that answer to part (b) did not require much detail, just a general description of the sort that has been posted here.

2. Was part (a) of this question where you calculated the experience rate for a 500 xs 500 layer? I think I answered it correctly, but had to explicitly state some assumptions. I think the question was ill-posed.

[MUAct]

the question where part b) said the mgmt team thought no premium should be charged for the 10M X 10M layer....

was a) 24ish%?

I couldn't remember if we had to apply the loss trend to premiums as well? I think we did

I24ish% sounds familiar

[Becoming An Actuary]

2. Was part (a) of this question where you calculated the experience rate for a 500 xs 500 layer? I think I answered it correctly, but had to explicitly state some assumptions. I think the question was ill-posed.

yes that was the question...here's what i did

1. sum of (incurred loss * 500x500*ULDF * loss trend)

2. sum of (subject premium*on level factor*loss trend)

1/2 to give rate = .24 or .25

anyone?

[radrach]

why (1) and (2) by loss trend?
my concern was, denominator SEP or SEP * onlevel. Since SEP is the Exposure, not the reins premium.

[frummie]

I think I said something about experience rating the lower layer and then multiplying that rate by the relativity of the exposure rated higher layer compared to the exposure rated lower layer.

You are correct, sir.

[chic squirrel]

Yeah I debated doing that for a long time because I remember All10 does that with construction cost trend or whatnot. In the end, I decided not to apply the loss trend to premium, although it was a coin flip.


I did the same-did not apply the loss trend to the premium figures.

[Becoming An Actuary]

Yeah I debated doing that for a long time because I remember All10 does that with construction cost trend or whatnot. In the end, I decided not to apply the loss trend to premium, although it was a coin flip.

page 101 of CASS 9, exhibit 9-11

[ahow]

You are correct, sir.I did the same as Frummie, but did not give a specific answer. I just gave the procedure...

[Levin]

yes that was the question...here's what i did

1. sum of (incurred loss * 500x500*ULDF * loss trend)

2. sum of (subject premium*on level factor*loss trend)

1/2 to give rate = .24 or .25

anyone?
1a. Why I think the question was ill-posed, is that trends are typically based on ground-up losses. I did not directly apply the trend to the either the losses or premiums, but made an assumption about how many losses hit the layer.

1b. I did not trend premium. Perhaps real-life got in the way of my thinking a little bit. Not all lines of business exhibit premium trend equal to the loss trend, and some don't exhibit an explicit premium trend at all. Nowhere in the problem was the line of business identified. In the absence of an identified premium trend, I would not assume it was equal to the loss trend.

2. But I still have this unanswered question: Didn't the ground-up loss ratio have to be at least 500%? There was about $100,000 of subject premium for each year, and there were losses excess of $500,000 for each year. Even if the only loss was the loss that hit the layer, you would still be at a 500% gross loss ratio. This is an unreasonable number, even though it is "just an exam problem." I'd like to say that I'm being dense here. What am I missing?

[Levin]

why (1) and (2) by loss trend?
my concern was, denominator SEP or SEP * onlevel. Since SEP is the Exposure, not the reins premium.
Denominator = on level Subject Premium. You are projecting a rate for the prospective policy year, so you want your premiums on the rate level of that prospective year.

[Mel-o-rama]

I think part b) has more than one answer. I answered that you can use exposure rating to get relativities to get a price for the upper layers.

[The Sad Man]

1b. I did not trend premium. Perhaps real-life got in the way of my thinking a little bit. Not all lines of business exhibit premium trend equal to the loss trend, and some don't exhibit an explicit premium trend at all. Nowhere in the problem was the line of business identified. In the absence of an identified premium trend, I would not assume it was equal to the loss trend.


Levin,

I didn't trend premium either and I agree that because no LOB was given, it made the problem ambiguous. I saw that you're writing a letter to the CAS. Are you going to complain about this question?

[jorgensquelch]

1a. Why I think the question was ill-posed, is that trends are typically based on ground-up losses. I did not directly apply the trend to the either the losses or premiums, but made an assumption about how many losses hit the layer.
This question was not asked well. You don't trend losses unless they are ground up. Trending losses that have already been portioned into the layer is nonsense. I actually did this though. I wonder what will come out of it.

[Levin]

Levin,

I didn't trend premium either and I agree that because no LOB was given, it made the problem ambiguous. I saw that you're writing a letter to the CAS. Are you going to complain about this question?
Yes. Hopefully not enough to get it thrown out, because I did a bang-up job of explaining my rationale, and would expect full credit. I am confident now that I did not misunderstand the problem; a number of my FCAS co-workers have responded to me that "layering, then trending" is definitely wrong. (And some of them would have responded by doing it for the problem anyway, while some would have made assumptions like I made.) These are all reinsurance actuaries, by the way.

The ambiguity of whether you should trend premium the same as losses was not seen as a defect. I will include it in my letter, since it was in the Reinsurance Practices book. (Though not in Clark; his Property example had a different trend for TIV than for losses, and he specifically mentioned that only inflation-sensitive exposure bases require premium trend.)

I am also confident now that if any of my coworkers have read this thread, I have outed myself ... if I hadn't previously.:wave:

[Levin]

This question was not asked well. You don't trend losses unless they are ground up. Trending losses that have already been portioned into the layer is nonsense. I actually did this though. I wonder what will come out of it.
I'm hoping that what comes out of it is both methods getting credit. I have not yet talked to anyone, or seen anyone here, who did it my way.

And, as stated above, there are people who experience rate for a living who would have answered it the "wrong way," knowing, as you did, that they had to do so to answer the question the way it was intended.

[Levin]

Okay, one more comment then I'm done for the night. I checked over on the CAS e-mail forum. No one has really been discussing stuff on the Part 6 list since the exam.

But I am impressed that they have turned a formerly ugly, cumbersome system into a smart-looking, eminently-navigable forum. The postings are still email-based, but they are displayed online in a discussion forum style. Nice job, CAS.

[Examinator]

Here's a thought on part A. I did it the way you all did, if I remember correctly, but someone else at work had this observation. Trend is to apply to ground up losses. Then cap/layer them. Then develop them. That's hit hard in the article, and in past questions. My coworker said we should have added 500,000 to all the losses (which were all in the 10k-50k range, or something similar), then trended them, subtracted 500,000 (capped/layered), then developed. The more I look at the problem, the more I think that's the approach they were after. Any thoughts?

[jorgensquelch]

...My coworker said we should have added 500,000 to all the losses (which were all in the 10k-50k range, or something similar), then trended them, subtracted 500,000 (capped/layered), then developed. The more I look at the problem, the more I think that's the approach they were after. Any thoughts?
I see what you are saying...but you need to apply that methodology to individual losses. The problem looks like it aggregates the layer losses by policy year, and if that is the case it wouldn't be possible to do what you are suggesting. For example, is the $25,000 in policy year from one claim, or 5 claims?
In the end, this problem needed to have more description.

[Levin]

Here's a thought on part A. I did it the way you all did, if I remember correctly, but someone else at work had this observation. Trend is to apply to ground up losses. Then cap/layer them. Then develop them. That's hit hard in the article, and in past questions. My coworker said we should have added 500,000 to all the losses (which were all in the 10k-50k range, or something similar), then trended them, subtracted 500,000 (capped/layered), then developed. The more I look at the problem, the more I think that's the approach they were after. Any thoughts?
That is is exactly how I did it, and it is the right way to experience rate. But to use it on this problem, you have to make some assumptions. I assumed, as did your co-worker, that there was a single excess loss per policy year and that no losses below the retention will pierce it after trend.

This is tacitly assuming it is a per-occurrence excess cover. One of my coworkers suggested that you could do it with fewer assumptions about the claim counts by assuming that it is a catastrophe excess cover. You would get the same answer in each of these cases.

You could theoretically produce a trend factor for the layer. We often talk informally about the "leveraging" of trend in excess layers. But we rarely, if ever, quantify it for use in projections. The distribution of losses below the retention is one complicating factor. The low credibility of excess losses is another.