Actuarial Outpost
 
Go Back   Actuarial Outpost > Actuarial Discussion Forum > Property - Casualty / General Insurance
FlashChat Actuarial Discussion Preliminary Exams CAS/SOA Exams Cyberchat Around the World Suggestions


Fill in a brief DW Simpson Registration Form
to be contacted when new jobs meet your criteria.


Reply
 
Thread Tools Search this Thread Display Modes
  #1  
Old 07-08-2013, 10:56 AM
me13 me13 is offline
CAS
 
Join Date: Oct 2011
College: University of Waterloo
Posts: 13
Default Loss Development in GLMs

Suppose you are creating a GLM for modelling loss cost for a long tail line using loss experience data. Would you generally applying LDFs to this loss experience data before modelling, account for development after creating the GLM, or use some combination of the two?
Reply With Quote
  #2  
Old 07-08-2013, 11:03 AM
MountainHawk's Avatar
MountainHawk MountainHawk is offline
Member
CAS AAA
 
Join Date: Dec 2001
Location: Salem, MA
Studying for Nothing!!!!
College: Lehigh University Alum
Favorite beer: Yuengling
Posts: 64,850
Default

Developing individual claims to go into a GLM is ... tricky. How do you go about doing it? At a minimum, you would need separate development factors for open claims and closed claims I would think. Maybe closed claims development is 1, but it wouldn't be for a line like WC. And how do handle IBNR claims? Do you just create claims out of thin air?

I haven't done this often in my current role, but my theoretical preference would be to use slightly outdated info valued at a point where claim count IBNR is minimal. If 98% of claims are reported in the first 24 months, for example, then I might use X years of data ending 6/30/11 all valued as of 24 months, and then try to adjust for the 24 -> Ultimate development later.
__________________

Play in the AO Prediction Game now!



1
Reply With Quote
  #3  
Old 07-08-2013, 11:26 AM
Buck's Avatar
Buck Buck is offline
Member
CAS AAA
 
Join Date: Mar 2005
Posts: 3,773
Default

There is an article co-authored by Mark Shapeland on GLM and loss development. You may do a search on the site CASACT.org. I recall Mark proved that the chain ladder process is actually intended to manually solve a GLM problem.
Reply With Quote
  #4  
Old 07-10-2013, 10:05 PM
3.14 3.14 is offline
Member
CAS
 
Join Date: Sep 2005
Posts: 983
Default

http://www.actuarialoutpost.com/actu...serving&page=2
Reply With Quote
  #5  
Old 07-10-2013, 10:06 PM
3.14 3.14 is offline
Member
CAS
 
Join Date: Sep 2005
Posts: 983
Default

Guszcza and Lommele also describe a model that contemplates pure IBNR.
Reply With Quote
  #6  
Old 08-22-2019, 09:14 AM
sticks1839 sticks1839 is offline
Member
CAS AAA
 
Join Date: Jul 2005
Posts: 1,653
Default

Quote:
Originally Posted by MountainHawk View Post
Developing individual claims to go into a GLM is ... tricky. How do you go about doing it? At a minimum, you would need separate development factors for open claims and closed claims I would think. Maybe closed claims development is 1, but it wouldn't be for a line like WC. And how do handle IBNR claims? Do you just create claims out of thin air?

I haven't done this often in my current role, but my theoretical preference would be to use slightly outdated info valued at a point where claim count IBNR is minimal. If 98% of claims are reported in the first 24 months, for example, then I might use X years of data ending 6/30/11 all valued as of 24 months, and then try to adjust for the 24 -> Ultimate development later.
Ancient bump, but I'm in the midst of considering this same question. Should we develop losses prior to modelling via a GLM and how do we minimize bias or distortion introduced from that development?

My thought is similar to MH. Use slightly dated loss experience to limit the missing IBNR claims, then use a report period development analysis to fully develop the known claims. I hadn't thought of the idea that closed claims should be considered separately, which complicates the development process. What about using an individual claim level model to adjust case reserve values to an expected "Ultimate" basis?

Also, has anyone done any sensitivity testing on the impact of various development methods/processes on the GLM performance/coefficients?
Reply With Quote
  #7  
Old 08-26-2019, 01:14 PM
Actuarially Me Actuarially Me is offline
Member
CAS
 
Join Date: Jun 2013
Posts: 191
Default

Quote:
Originally Posted by sticks1839 View Post
Ancient bump, but I'm in the midst of considering this same question. Should we develop losses prior to modelling via a GLM and how do we minimize bias or distortion introduced from that development?

My thought is similar to MH. Use slightly dated loss experience to limit the missing IBNR claims, then use a report period development analysis to fully develop the known claims. I hadn't thought of the idea that closed claims should be considered separately, which complicates the development process. What about using an individual claim level model to adjust case reserve values to an expected "Ultimate" basis?

Also, has anyone done any sensitivity testing on the impact of various development methods/processes on the GLM performance/coefficients?
I don't have any specific experience with it, but this book may point you in the right direction. Starting at page 208.

https://ferassamain92.github.io/MyCV...%20Volume2.pdf
Reply With Quote
  #8  
Old 08-26-2019, 01:17 PM
Actuarially Me Actuarially Me is offline
Member
CAS
 
Join Date: Jun 2013
Posts: 191
Default

Looks like he references Monograph 1, so this may point you in the right direction too until someone with experience jumps in:

https://www.casact.org/pubs/monograp.../01-Meyers.PDF
Reply With Quote
  #9  
Old 08-26-2019, 04:17 PM
Colonel Smoothie's Avatar
Colonel Smoothie Colonel Smoothie is online now
Member
CAS
 
Join Date: Sep 2010
College: Jamba Juice University
Favorite beer: AO Amber Ale
Posts: 49,282
Default

Quote:
Originally Posted by 3.14 View Post
Guszcza and Lommele also describe a model that contemplates pure IBNR.
This one?

https://www.casact.org/pubs/forum/06fforum/115.pdf
__________________
Recommended Readings for the EL Actuary || Recommended Readings for the EB Actuary

Quote:
Originally Posted by Wigmeister General View Post
Don't you even think about sending me your resume. I'll turn it into an origami boulder and return it to you.
Reply With Quote
  #10  
Old 08-30-2019, 04:12 PM
DamSon's Avatar
DamSon DamSon is offline
Member
CAS
 
Join Date: Sep 2018
Location: NYC
Studying for 8
Favorite beer: Alcoholic
Posts: 221
Default

This is probably a dumb question, but if you're building a PP model (or even freq/sev) does developing claims even matter other than if used as a credibility standard? Effectively it cancels out right? I feel like I'm missing something if those way smarter than me have posted papers on the topic.
Reply With Quote
Reply

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off


All times are GMT -4. The time now is 08:57 PM.


Powered by vBulletin®
Copyright ©2000 - 2019, Jelsoft Enterprises Ltd.
*PLEASE NOTE: Posts are not checked for accuracy, and do not
represent the views of the Actuarial Outpost or its sponsors.
Page generated in 0.24324 seconds with 9 queries