There are different ways to aggregate loss data. You can discover lots of interesting information about your dataset just by looking at it differently. Every claim should have information on:
- Accident date – The date of loss, when the accident occurs. For chronic illnesses or other losses that occur over time, the discovery date is typically used
- Report date – The date the claim gets reported to the claims administrator.
- Paid date – The date of the transaction. This probably won’t be shown to you as there could be many transactions, or it may show the latest transaction date. Only needed for grouping by calendar year.
- Policy Year dates – The beginning and end dates of the policy year associated with the claim. This may not always be in the loss data but you should know it.
You can now group claims together by calendar years, report years, accident years, or policy years. Each have special purposes and characteristics. An example will be shown below.
Policy Year – This is used to make actuarial projections to assist in projecting liabilities for the upcoming policy year. If you’re going on a “per occurrence” basis, you use the loss date. If you go by a “claims made” basis, you use the report date. Then you group the losses by the policy years’ start and end dates. This has the most opportunity for development, especially for policies with long tails.
Development refers to changes over time to the loss amounts, either because of new transactions, changes to reserves, and/or new claims. The way you group the claims will have different development rates due to the way claims come in. When the claims are fully developed and closed out, those years are at ultimate.
Claim Number | Loss Date | Report Date | Paid Date | Paid Amount |
1 | 7/1/2014 | 10/1/2014 | 11/1/2014 | 100 |
2 | 4/1/2015 | 7/1/2015 | 8/1/2015 | 200 |
3 | 7/1/2015 | 12/1/2015 | 1/1/2016 | 300 |
4 | 10/1/2016 | 1/1/2017 | 2/1/2017 | 400 |
1 | 7/1/2014 | 10/1/2014 | 4/1/2017 | 500 |
Grouped by Calendar Year, valued as of 12/31/15, 12/31/16, and 12/31/17.
Calendar Year | 12/31/2015 | 12/31/2016 | 12/31/2017 |
1/1/2014 | 100 | 100 | 100 |
1/1/2015 | 200 | 200 | 200 |
1/1/2016 | 300 | 300 | |
1/1/2017 | 900 |
Grouped by Report Year, valued as of 12/31/15, 12/31/16, and 12/31/17. Note the development is a result of IBNER.
Report Year | 12/31/2015 | 12/31/2016 | 12/31/2017 |
1/1/2014 | 100 | 100 | 600 |
1/1/2015 | 200 | 500 | 500 |
1/1/2016 | |||
1/1/2017 | 400 |
Grouped by Accident Year, valued as of 12/31/15, 12/31/16, and 12/31/17.
Accident Year | 12/31/2015 | 12/31/2016 | 12/31/2017 |
1/1/2014 | 100 | 100 | 600 |
1/1/2015 | 200 | 500 | 500 |
1/1/2016 | 400 | ||
1/1/2017 |
Grouped by Policy Year, valued as of 12/31/15, 12/31/16, and 12/31/17.
Policy Year | 12/31/2015 | 12/31/2016 | 12/31/2017 |
6/1/2014 | 300 | 300 | 800 |
6/1/2015 | 300 | 300 | |
6/1/2016 | 400 | ||
6/1/2017 |
Were you able to get the same answers? In another article we will look at how to observe trends, and the appropriate methods used to address them.
Share this: