PDA

View Full Version : Premium and Loss Trends


leemathewson
01-17-2012, 10:13 PM
Reading through the Werner and Modlin text, I get the impression that the trend period for premiums is the same length as the trend period for losses. Is this accurate?

I understand that premiums should be trended from the average earned date of the historical period to the average earned date of policies written during the effective period (not the average date of premium earned during the effective period - a mistake I've caught myself making a couple times.) Further, losses should be trended from the average date of loss in the historical period to the average date of loss of policies written during the effective period. Wouldn't these be the same?

Here's an example: Suppose you are using CY-AY 2010 earned premiums and losses to price policies with an effective date of 1.1.2012. Suppose further there is a year between rate revisions and you write annual policies. Then, for premiums, I would want to trend from 7.1.2010 to 1.1.2013 (average date of premiums earned during CY 2010 to the average date of premium earned from policies written in the effective period). Same dates for the trending period for losses. Am I wrong here? Wouldn't the length of time for premium trend be the same as for losses?

Sorry about the long post. Thanks!

david849
01-18-2012, 12:06 PM
average written date for CY2010 EP would be 1.1.2010.
average written date for PY2012 would be 6.30.2012.
For trending i think you always use the average written date instead of average earned date.

leemathewson
01-18-2012, 01:25 PM
I don't think average written or average earned date matters. Both ways you will end up with the same trend period. See Werner and Modlin pg. 85.

Just curious if it is too simplifying to say that the amount of time to trend losses will be the same as the amount of time to trend premium.

david849
01-18-2012, 03:00 PM
right, i revise my statement, both written and earned date will produce the same trend period.

I believe you will get the same trend period for premium and loss if you are using CY for prem and AY for loss for experience period, and PY for projected period. But if you are using PY for loss in experience period (P114) you will get a shorter trending period for loss.

leemathewson
01-18-2012, 06:11 PM
This is assuming losses are PY and premiums are CY, right? Wouldn't you want losses and premiums to be aggregated the same?

right, i revise my statement, both written and earned date will produce the same trend period.

I believe you will get the same trend period for premium and loss if you are using CY for prem and AY for loss for experience period, and PY for projected period. But if you are using PY for loss in experience period (P114) you will get a shorter trending period for loss.

Almost_Done
01-18-2012, 07:51 PM
I suppose you would want them to be aggregated the same way in practice. But in theory I don't think they have to be, besides for consistency. After all these are two separate processes and eventually you arrive at the trended perm and loss for the projected period separately

Vorian Atreides
01-19-2012, 10:04 AM
This is assuming losses are PY and premiums are CY, right? Wouldn't you want losses and premiums to be aggregated the same?
If losses are PY aggregated, you'll want premiums PY aggregated as well. Note that in the latter case, earned premium (or exposures) = written premium (or exposures).

The other options one can consider are:

CY aggregated losses compared to CY aggregated earned premium (exposures).
AY aggregated losses compared to CY aggregated earned premium (exposures).
Note that PY aggregation is the :qunq:most accurate:qunq: matching of losses to corresponding exposures; but AY/CY matching will often do well enough to approximate PY comparisons; especially for short-tail lines with a stable book of business.

Side note: regardless of how you're aggregating your losses/premiums in the experience period, your forecast period is always on a PY basis. You should understand why this is the case. (IIRC, there have been Exam problems that ask for the trending dates and not just the length of time.)