View Full Version : Selecting development factors
01-23-2003, 09:36 PM
There seems to be different schools of thought to choose loss development factors. How do you select a set of these factors?
01-24-2003, 11:11 AM
For me, I look at the resulting completion factor pattern (completion factor = 1 / (factor to ultimate). I use my judgment to select the factors.
A few thoughts:
1. Don't rely on one consistent average for your selections, such as using the 3 year average for all selecteds. Use informed judgment.
2. Be careful with averages excluding highs and lows. They may have a downward bias. I believe their was a paper written that tries to demonstrate this mathematically.
3. When doing your analysis, spend a majority of your time on pre-selection and post-selection diagnostics. Before making selections, review closed-to-reported ratios and average case reserves for potential changes in settlement rates are reserving practices, respectively. After making my preliminary selections, I review a few statistics for reasonableness. Are the projected freq/sev levels reasonable given expected trends? Are the projected loss ratios reasonable given rate change activity? Review the "indicated unpaid per pending claim" triangle implied by your ultimate selections.
4. Further on down the road, use retro tests to evaluate your selected LDF's.
01-24-2003, 12:37 PM
First step is keeping communication lines open with claims. Have they made any changes to the way they settle claims in the last year. Then look at the data and start asking questions about any strange looking data - is it real or bad data? Then start your analysis and start asking more questions...
Also, it depends what you are using the factors for - developing the data in the triangle to ultimate, or projecting other data. If the former, you may want to make specific adjustments to known oddities in the data. If the later, you are more likely to want to remove the efffects of the oddities.
vBulletin® v3.7.6, Copyright ©2000-2013, Jelsoft Enterprises Ltd.