loss model - Bosna RE
loss model - Bosna RE
loss model - Bosna RE
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
Određivanje cijene reosiguranja<br />
viška štete za zelenu kartu /<br />
Pricing of Green Card XL<br />
Reinsurance<br />
Dr. Torsten FISCHER<br />
Domenico SEMERARO<br />
September 1st, 2009<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
1<br />
Overview of the Pricing Process for Green Card XL<br />
2<br />
Simple Stochastic Model for Large Losses<br />
3<br />
Calculation of the Price for XL Reinsurance<br />
4<br />
Outlook on Refined Loss Modelling<br />
5 Discussion<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
Overview of the Pricing Process for Green Card XL<br />
1. Submission of pricing request by cedent / broker:<br />
description of portfolio, slip, (rates).<br />
2. Creation of a <strong>loss</strong> <strong>model</strong> for single large <strong>loss</strong>es.<br />
3. Calculation of a price for reinsurance based on <strong>model</strong><br />
and contract features.<br />
Possible: quotation for alternative layerings, using the<br />
same <strong>model</strong>.<br />
4. Negotiations, finalization of the contract.<br />
3
1<br />
Overview of the Pricing Process for Green Card XL<br />
2<br />
Simple Stochastic Model for Large Losses<br />
3<br />
Calculation of the Price for XL Reinsurance<br />
4<br />
Outlook on Refined Loss Modelling<br />
5 Discussion<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
Simple Stochastic Model for Large Losses<br />
Fictitious Example for Green Card XL Reinsurance<br />
• Reinsurance for 3 layers:<br />
1. Layer: 200.000 EUR xs 200.000 EUR<br />
2. Layer: 600.000 EUR xs 400.000 EUR<br />
3. Layer: unlimited xs 1.000.000 EUR.<br />
• Data:<br />
new contract period: 2009<br />
same portfolio for 10 years<br />
<strong>loss</strong> history for the years 1999 – 2008:<br />
complete list of all <strong>loss</strong>es above 100.000 EUR.<br />
What is the price for reinsurance?<br />
Problem: We do not know the future <strong>loss</strong>es.<br />
5
Simple Stochastic Model for Large Losses<br />
• Approach: Create and apply a <strong>model</strong> for the (future)<br />
<strong>loss</strong>es, based on the <strong>loss</strong> history.<br />
• What should the <strong>model</strong> predict?<br />
1. number of <strong>loss</strong>es above the priority (200.000 EUR),<br />
2. the severity (<strong>loss</strong> amount) distribution.<br />
• Assumptions (for simplicity):<br />
no inflation of premiums or <strong>loss</strong>es,<br />
no development of <strong>loss</strong>es (final incurreds are known).<br />
For more realistic assumptions, see below.<br />
6
Loss Amount in EUR<br />
Simple Stochastic Model for Large Losses<br />
Fictive Loss History (above 100.000 EUR in 1999-2008)<br />
1.200.000<br />
1.000.000<br />
800.000<br />
600.000<br />
400.000<br />
200.000<br />
0<br />
1999 2000 2001 2002 2003 2004 2005 2006 2007 2008<br />
Accident Year<br />
7
Number<br />
Simple Stochastic Model for Large Losses<br />
All Loss Amounts above 200.000 EUR, ordered<br />
51<br />
41<br />
31<br />
21<br />
11<br />
1<br />
0 200.000 400.000 600.000 800.000 1.000.000 1.200.000<br />
Loss Amount in EUR<br />
8
Cumulative Probability<br />
Simple Stochastic Model for Large Losses<br />
Empiric Severity Curve<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
0 400.000 800.000 1.200.000 1.600.000<br />
Loss Amount in EUR<br />
9
Cumulative Probability<br />
Simple Stochastic Model for Large Losses<br />
Empiric Severity Curve and Pareto Fit<br />
100%<br />
80%<br />
60%<br />
40%<br />
The largest historic<br />
<strong>loss</strong> is below 1 mio.<br />
In the Pareto <strong>model</strong> also<br />
<strong>loss</strong>es above 1 mio are<br />
possible.<br />
20%<br />
0%<br />
0 400.000 800.000 1.200.000 1.600.000<br />
Loss Amount in EUR<br />
10
Number of Large Losses<br />
Simple Stochastic Model for Large Losses<br />
Number of Losses above 200.000 EUR per Year<br />
10<br />
8<br />
6<br />
4<br />
History<br />
Average 5,4<br />
2<br />
0<br />
1998 2000 2002 2004 2006 2008 2010<br />
Accident Year<br />
11
Simple Stochastic Model for Large Losses<br />
We have estimated the parameters of a<br />
frequency-severity-<strong>model</strong>:<br />
1. Model threshold: 200.000 EUR<br />
2. Frequency:<br />
average 5,4 <strong>loss</strong>es (above 200.000 EUR) per year<br />
Poisson-distribution (not shown here)<br />
3. Severity (<strong>loss</strong> amount)<br />
Pareto-distribution<br />
Pareto-alpha: 2,3.<br />
In this <strong>model</strong> even very large <strong>loss</strong>es are possible, e.g. <strong>loss</strong><br />
amounts above 5 mio EUR.<br />
12
1<br />
Overview of the Pricing Process for Green Card XL<br />
2<br />
Simple Stochastic Model for Large Losses<br />
3<br />
Calculation of the Price for XL Reinsurance<br />
4<br />
Outlook on Refined Loss Modelling<br />
5 Discussion<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
Cumulative Probability<br />
Calculation of the Price for XL Reinsurance<br />
Expected Loss to each Layer = Area * Frequency 5,4<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
0%<br />
0 400.000 800.000 1.200.000 1.600.000<br />
Loss Amount in EUR<br />
14
Cumulative Probability<br />
Calculation of the Price for XL Reinsurance<br />
Annual Loss per Layer<br />
100%<br />
80%<br />
60%<br />
40%<br />
20%<br />
With probability 88% the 3. layer<br />
has no annual <strong>loss</strong>.<br />
With probability 56% the 1. layer has an<br />
annual <strong>loss</strong> of 500.000 EUR or less.<br />
1. Layer 0,2 xs 0,2<br />
2. Layer 0,6 xs 0,4<br />
3. Layer unlimited xs 1<br />
0%<br />
0 500.000 1.000.000 1.500.000 2.000.000<br />
Loss Amount in EUR<br />
15
Costs in EUR<br />
Calculation of the Price for XL Reinsurance<br />
1. layer: “working layer”,<br />
mainly driven by expected <strong>loss</strong>.<br />
Expected Costs per Layer<br />
600.000<br />
3. layer: rarely hit, small expected <strong>loss</strong><br />
mainly driven by cost of capital.<br />
400.000<br />
200.000<br />
Cost of Capital<br />
Internal Expenses<br />
Brokerage<br />
Expected Loss<br />
0<br />
0,2 xs 0,2 0,6 xs 0,4 unlimited xs 1<br />
Layer<br />
16
1<br />
Overview of the Pricing Process for Green Card XL<br />
2<br />
Simple Stochastic Model for Large Losses<br />
3<br />
Calculation of the Price for XL Reinsurance<br />
4<br />
Outlook on Refined Loss Modelling<br />
5 Discussion<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
Outlook on Refined Loss Modelling<br />
• Loss Inflation: How much would a historic <strong>loss</strong> cost<br />
today?<br />
Possible approach: <strong>loss</strong> amounts are inflated like the<br />
wages or medical costs in the country of occurrence.<br />
• Loss development: Consider that the ultimate <strong>loss</strong><br />
amount for an accident is unknown for a long time. Split of<br />
<strong>loss</strong> into paid and reserved. Discount of later payments,<br />
capital allocation for period of uncertainty.<br />
• Application of market <strong>model</strong>s: in particular if the cedent’s<br />
data is sparse, not suitable for <strong>model</strong>ling; application to<br />
non-working layers.<br />
18
Outlook on Refined Loss Modelling<br />
• Separate <strong>model</strong>s for different type of risks: segmentation<br />
by type of vehicle, destination<br />
- however: one needs a lot of good data for that.<br />
• Comparison of historic portfolios:<br />
- Inflation of premiums, e.g. by Premium Index (tariff<br />
changes).<br />
- Preferred exposure measure: number of vehicles or<br />
vehicle years.<br />
19
Outlook on Refined Loss Modelling<br />
• Consideration of structural changes:<br />
- changes in the cedent’s portfolio, e.g. from passenger<br />
cars to trucks, merger with another company, change<br />
in underwriting policy.<br />
- changes in law: higher limits, change in jurisdiction<br />
→ estimation of quantitative effect is difficult.<br />
Facit: Refined Loss Modelling leads to a better estimation of<br />
the fair price for Reinsurance.<br />
The better the data (level of information, number of<br />
reported years,…) the better the <strong>model</strong>.<br />
20
1<br />
Overview of the Pricing Process for Green Card XL<br />
2<br />
Simple Stochastic Model for Large Losses<br />
3<br />
Calculation of the Price for XL Reinsurance<br />
4<br />
Outlook on Refined Loss Modelling<br />
5 Discussion<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009
Hvala na pažnji!<br />
-<br />
Thank you for your<br />
attention!<br />
Dr. Torsten FISCHER<br />
Phone: +49-221-2928-1452<br />
TFischer@scor.com<br />
Domenico SEMERARO<br />
Phone: +49-221-2928-1304<br />
DSemeraro@scor.com<br />
Pricing of Green Card XL Reinsurance - Sarajevo<br />
September 1st, 2009