DAT-16 Data For Modeling

1
DAT-16 Data For Modeling

• Nolan Asch
• Discussion of how data , or the LACK of it,
  influences how reinsurers use Reinsurance
  Pricing Models

2
DATA FOR REINSURANCE PRICING

• WHAT IS WANTED
• Limits Profile
• Losses excess of half the attachment
• Loss versus LAE

• WHAT YOU GET
• Sometimes
• EVERYTHING
• Sometimes
• Much Less

3
What To Do If You Get The Proper Data

• Exposure Rating REI-18
• Experience Rating REI-17

4
What To Do If You Do NOT Get The Proper Data

• Twist the data to try to fit the model inputs

• Twist the model to adapt to your data

5
What to do if you do NOT get proper data

• DEMAND PROPER DATA!

6
How Models Are Typically Twisted

• Subject Premiums
• Adjusted Subject Premiums
• Limits Profiles
• Locations NOT premiums

7
Typical Limits Profile ( claims) Twisting

• Excludes business no longer written
• Too General ( All GL / All auto)
• Substitutes locations for premiums

8
TYPICAL MODEL TWISTING

• As If Analyses
• ( Heroic) Assumptions
• Comparison to Benchmarks

9
HOW SENSITIVE ARE MODELS TO THE DATA?

• GIGO

10
How Sensitive are models to proper data?

• Run sensitivity analysis

11
HOW SENSITIVE ARE MODELS TO DATA?

• Use various dimensions to find the key variable
  or variables.

12
COMMON DIMENSIONS TO TEST

• Trend Factors

• Loss Development Factors ( or any other input
  variable that is key to the rating models)

13
WHAT DATA ISSUES RELATE TO A VENDOR?

• Does the vendor have a lot of experience in the
  data field and do they have a great deal of
  experience in the insurance industry??