Post on 21-Jun-2020
transcript
Weather Cycles: What’s in it for Insurers?
Prepared by Tim Andrews, Sean West and Kamal Pun
This presentation has been prepared for the Actuaries Institute 2012 General Insurance Seminar.
The Institute Council wishes it to be understood that opinions put forward herein are not necessarily those of the Institute and the Council is not responsible for those opinions.
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
-25
-20
-15
-10
-5
0
5
10
15
20
25
1901
1903
1905
1907
1909
1911
1913
1915
1917
1919
1921
1923
1925
1927
1929
1931
1933
1935
1937
1939
1941
1943
1945
1947
1949
1951
1953
1955
1957
1959
1961
1963
1965
1967
1969
1971
1973
1975
1977
1979
1981
1983
1985
1987
1989
1991
1993
1995
1997
1999
2001
2003
2005
2007
2009
2011
0
100
200
300
400
500
600
700
SOI I
ndex
Financial Year
Tota
l R
ainf
all (
mm
)
Total Rainfall (October to March) SOI Index Series1
The Link Between ENSO and Rainfall
The Link Between ENSO and Rainfall
0
100
200
300
400
500
600
El Nino Neutral La Nina Strong La Nina
Ave
rage
Rai
nfal
l Oct
ober
to M
arch
(mm
)
Average Rainfall (October to March) Average all years
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
The Link Between Rainfall and Insurance Claims
Home - Brisbane
GIOSydney & Brisbane Regions (2010-11)
Home & Motor Cost Per Policy Relative to No Rain
Home - Sydney Motor - Sydney
0%
200%
400%
600%
800%
1000%
1200%
Nil 0-1mm 1-10mm 10-20mm 20-30mm 30mm+ No Rain Rain
CPP
Rela
tivi
ty to
No
Rain
Daily Rainfall Intensity
0%
100%
200%
300%
400%
500%
600%
700%
800%
Nil 0-1mm 1-10mm 10-20mm 20-30mm 30mm+ No Rain Rain
CPP
Rela
tivi
ty to
No
Rain
Daily Rainfall Intensity
90%
95%
100%
105%
110%
115%
120%
125%
Nil 0-1mm 1-10mm 10-20mm 20mm+ No Rain Rain
CPP
Rela
tivi
ty to
No
Rain
Daily Rainfall Intensity
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
Influence of ENSO on Storms - Brisbane
0
1
2
3
4
5
6
-30 -20 -10 0 10 20 30
Stor
m S
core
SOI
StrongLa Nina
Weak La Nina
1
2
3
4
5
El Nino Neutral La Nina Strong LaNina
Brisbane Flood History
0
1
2
3
4
5
6
7
8
9
10
1877 1886 1895 1904 1913 1922 1931 1940 1949 1958 1967 1976 1985 1994 2003 2011
Floo
d Le
vel (
met
res)
Year ending 30 June
Peak SOI > 8 SOI > 14
River Peak vs SOI
Weather Cycle vs Annual Industry Costs (ICA Data)
0
5
10
15
20
25
30
0%
50%
100%
150%
200%
250%
El Nino Neutral La Nina Strong La Nina
Num
ber
of Y
ears
in W
eath
er P
hase
Ave
rage
Ann
ual C
ost R
elat
ive
to A
vera
ge
Relativity Average Number of Years in Weather Phase
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
• Two types of models available to predict sea-surface temperatures in the equatorial Pacific Ocean:
1. Dynamic Models • Physical equations representing ocean and atmospheric
behaviour used to determine future conditions • Computationally intensive
2. Statistical Models: • Rely on large volumes (30-50 years) of past observations
to predict the future • Simpler and cheaper to implement
ENSO Predictions: Types of Models
Red = Actual
Blue = Forecast (6 months prior)
A Particular Dynamic Model: Lamont-Doherty (LDE05)
Source: “Predictability of El Nino over the past 148 years” (Cane 2004)
A Particular Dynamic Model: Lamont-Doherty (LDE05)
Source: “Predictability of El Nino over the past 148 years” (Cane 2004)
Dynamic Model Predictions • “Ensemble” models tend to be the best performers • Could be improved further by excluding certain weaker
models from the ensemble • 86% correlation at 6 months lead time
Source: Springer-Verlag 2008
-3
-2
-1
0
1
2
3
Dec-07 Apr-08 Aug-08 Dec-08 Apr-09 Aug-09 Dec-09 Apr-10 Aug-10 Dec-10 Apr-11 Aug-11 Dec-11 Apr-12 Aug-12 Dec-12 Apr-13
IRI CS - 6 months ALL IRI CS - 6 months DYN POAMA - 6 months ESSIC Intermed Coupled Model
Japan Met Agency - 6 months Japan Frontier Coupled Actual NINO3.4
…How Accurately were the 2011/12 La Nina Events Predicted?
Presentation Outline
The Link Between ENSO
and Rainfall
The Link Between Rainfall
and Insurance Costs
The Link Between ENSO and Insurance
Costs
Is ENSO Predictable?
If Yes, What Can We Do With the
Predictions?
What can we do with the Predictions?
• Assume we can perfectly predict the stage of the weather cycle months in advance…
• What is the benefit to insurers?
A Simple Model
• Considered range of pricing strategies
• Only new business premiums adjusted in light of forecast; retention book untouched
• Measured the potential saving in weather loss ratio under each option
A Simple Model: Key Assumptions Number of Years Simulated: 20Number of Simulations: 1,000Initial Number of Policies: 1,000
Base Premium: 400Base Weather CPP (October-March): 200Base Weather CPP (April-September) 120
Type of Weather CPP PremiumPeriod Oct-Mar Apr-Sep Relativity OutlookStrong La Nina 10% 0% 200% Very BadLa Nina 15% 0% 130% BadNeutral 50% 100% 100% AverageEl Nino 25% 0% 90% Good
Probability
Premium Retention New BusinessOutlook Rate (p.a.) Growth (p.a.) Renewal New BusinessVery Bad 85% 10% 400 500Bad 85% 15% 400 450Average 85% 25% 400 400Good 85% 30% 400 350
Premium Rate ($)
Cycle Prediction: Strategies Considered
1. Do nothing – base strategy
2. Reactive – change rates based on previous period
3. 6 month forecast – change rates based on the stage of the
cycle 6 months in advance
4. 6 month forecast with 2 year outlook – if there is a Strong
La Nina within the next 2 years, change the rates now, otherwise
maintain 6 month strategy
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
Num
ber o
f Pol
icie
s
Do Nothing
Reactive
6 months
6 months, 2 year outlook
A Sample Run
2013 2014 2015 2016 2017 2018 2019 2020 2021 2022 2023 2024 2025 2026 2027 2028 2029 2030 2031 2032
Strong La Nina La Nina Neutral El Nino Series5
Strategy Avg pol growth (p.a.) Average LRDo Nothing 10.25% 42.67%Reactive 9.27% 42.88%6 months 9.27% 42.67%6 months, 2 yr outlook 7.05% 42.48%
Results Across all Simulations:
“Best” strategy Loss Ratio
Improvement = 0.19%
Premium Retention New BusinessOutlook Rate (p.a.) Growth (p.a.) Renewal New BusinessVery Bad 82% 10% 500 500Bad 84% 15% 450 450Average 85% 25% 400 400Good 87% 30% 350 350
Premium Rate ($)
A Simple Model: Alternate Assumptions
Number of Years Simulated: 20Number of Simulations: 1,000Initial Number of Policies: 1,000
Base Premium: 400Base Weather CPP (October-March): 200Base Weather CPP (April-September) 120
Type of Weather CPP PremiumPeriod Oct-Mar Apr-Sep Relativity OutlookStrong La Nina 10% 0% 200% Very BadLa Nina 15% 0% 130% BadNeutral 50% 100% 100% AverageEl Nino 25% 0% 90% Good
Probability
Strategy Avg pol growth (p.a.) Average LRDo Nothing 10.25% 42.63%Reactive 9.31% 42.66%6 months 9.31% 42.41%6 months, 2 yr outlook 6.63% 40.89%
Results Across all Simulations:
“Best” strategy Loss Ratio
Improvement = 1.74%
Conclusions
• ENSO does affect weather costs
• It is predictable to some extent
• Insurers need to be clever to achieve a significant impact on profit
Any Other Potential Uses for ENSO Predictions?
• Budgeting
• Monitoring
• Normalising historical results