Waves, Wind and Water: Weather perils and insurance

Dmitry Gorelik

Weather and Insurance• Recent quotes from the press regarding weather-

related losses“…our result for the 2008 financial year has been

affected by…. succession of severe weather events”

“…we have been affected by the increased frequency of natural perils…”

“…raise premiums on personal products to offset more frequent claims after bad weather…”

“…sustained impact of volatile weather …”

Data and Loss adjustmentData• ICA Natural Disasters Claims List• EMA disaster database (www.ema.gov.au)• Benfield market knowledge

Loss Adjustment• CPI alone is not appropriate• RiskFrontiers (Crompton and McAneney) use a

combination of changes in population, wealth and inflation

• CPI and GDP used as a proxy for changes in wealth and inflation

Historical loss adjustment

Crompton and McAneney; 2008Yearly Natural Hazard Losses (in 2008 dollars)

0

500

1,000

1,500

2,000

2,500

3,000

1967

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

tota

l los

ses

($m

)

Original

Yearly Natural Hazard Losses (in 2008 dollars)

0

1,000

2,000

3,000

4,000

5,000

6,000

7,000

1967

1970

1972

1974

1976

1978

1980

1982

1984

1986

1988

1990

1992

1994

1996

1998

2000

2002

2004

2006

2008

tota

l los

ses

($m

)

Normalised

Historical loss adjustment – a few examples

• Comparison of top 10 natural peril insurance losses ($m) as at 2006

670

710

740

820

1430

1480

1610

1790

3300

4120

Normalised Loss(Crompton/McAneney ,2006)

446

465

560

683

886

688

1100

1341

2990

3944

Benfield (CPI and GDP)

51580Sydney1984Flood10

533104Sydney1986Hailstorm9

70840Sydney1976Hailstorm8

128030Multiple1973Tropical Cyclone Madge7

1512180Brisbane1985Hailstorm6

773319Sydney1990Hailstorm5

983176Multiple1983Ash Wednesday Bushfires4

243368Brisbane1974Tropical Cyclone Wanda3

31141700Sydney1999Hailstorm2

3635200Darwin1974Tropical Cyclone Tracy1

Benfield (CPI/GSP

and population)

OriginalLossLocationYearEvent Rank

670

710

740

820

1430

1480

1610

1790

3300

4120

Normalised Loss(Crompton/McAneney ,2006)

446

465

560

683

886

688

1100

1341

2990

3944

Benfield (CPI and GDP)

51580Sydney1984Flood10

533104Sydney1986Hailstorm9

70840Sydney1976Hailstorm8

128030Multiple1973Tropical Cyclone Madge7

1512180Brisbane1985Hailstorm6

773319Sydney1990Hailstorm5

983176Multiple1983Ash Wednesday Bushfires4

243368Brisbane1974Tropical Cyclone Wanda3

31141700Sydney1999Hailstorm2

3635200Darwin1974Tropical Cyclone Tracy1

Benfield (CPI/GSP

and population)

OriginalLossLocationYearEvent Rank

Why is weather hard to model?

998

Natural Peril losses through time (1967 – 2008)Proportion of Natural Hazard Losses (in 2008 dollars)

ex Tracy, ex Newcastle Earthquake

22%

1%

10%

18% 11%

39%

Proportion of Natural Hazard Losses (in 2008 dollars)

20%

9%

9%

16% 10%

36%

bushfirecycloneearthquakefloodhailstorm

• Hail and Storm losses account for more than half of all natural peril insurance losses

• Difficult to differentiate between “Hail” and “Storm” events• The “outliers” cannot be clearly defined where natural perils are

concerned

Natural Peril losses through time

• Almost 80% of natural peril losses in the last 20 years were storm related

• However, this split is influenced by inclusion of large events such as “Ash Wednesday”, Sydney Hail Storm, Cyclone Larry etc

Proportion of Natural Hazard Losses (in 2008 dollars) ex earthquakes ex TC Tracy

12% 6% 7% 11%

10%10% 9%

22%8%

6% 5%

10%

46%50% 47%

39%

24% 28% 32%18%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

last 30 years last 20 years last 10 years total over 40 years

bushfire cyclone flood hail storm

70% 78% 79%

57%

ENSO – El Niño Southern Oscillation• El Niño

– part of a natural oscillation of the ocean-atmosphere system in the tropical Pacific

– commonly referred to as El Niño-Southern Oscillation (ENSO)

• Extreme weather associated with El Niño:– severe droughts and bushfires– devastating floods and landslides– depends on geographic location

• The opposite phase to El Niño is known as La Niña– also linked with extreme weather– storms, hail and cyclones

ENSO – 1967-2007

• A full ENSO cycle, which incorporates one El Niño and one La Niña event - generally takes about four years.

• Approximately a 25% chance of an El Niño (La Niña) event occurring in any one year

SOI index 1967 - 2007 (5 months MA)

-30.0

-25.0

-20.0

-15.0

-10.0

-5.0

0.0

5.0

10.0

15.0

20.0

Jun-67

Feb-69

Oct-70

Jun-72

Feb-74

Oct-75

Jun-77

Feb-79

Oct-80

Jun-82

Feb-84

Oct-85

Jun-87

Feb-89

Oct-90

Jun-92

Feb-94

Oct-95

Jun-97

Feb-99

Oct-00

Jun-02

Feb-04

Oct-05

Jun-07

ENSO and Eastern Seaboard insurance losses

• No clear relationship between losses, type of ENSO phase or the length

Length of ENSO phase vs Industry Loss per month

0

5

10

15

20

25

30

35

40

1/1967 6/1969 4/1972 7/1976 7/1979 3/1985 10/1990 3/1995 3/1997 8/2000 9/2004 10/2005 7/2007

num

ber o

f mon

ths

in a

pha

se

0

50

100

150

200

250

300

350

400

450

loss

per

mon

th ($

m)

La Nina

El Nino

Neutral

Average monthly losses (ex TC Tracy)

• Average monthly loss is higher in the last 10 year• Losses are significantly higher in La Nina phase• However…. These are mainly driven by April

1999 Sydney hail storm

0

20

40

60

80

100

120

140

All Periods Last 10 years Last 20 years

loss

($m

)

La Niña

La Niña ex Sydney Hail

Neutral

El Niño

Average monthly losses – capped at $500m

• A lot less variability when the upper limit is imposed• Arguably a more relevant representation for the

insurance industry – catastrophe reinsurance is likely to cap the industry loss

0

10

20

30

40

50

All Periods Last 10 years Last 20 years

loss

($m

)

La Niña

Neutral

El Niño

Losses by ENSO State

• No bushfire losses in La Nina phase• Large proportion of Hail in La Nina phase• Storm, Hail and Flood should potentially be combined due to the

difficulty in separating these perils

Proportion of Loss Amount

29%12%

21%

15%52%

3%

13%

6%

36%

32%

38%

11%

28%

4%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

El Niño Neutral La Niña

Storm

Hail

Flood

Cyclone

Bushfire

Losses by ENSO State – excluding ‘outliers’

• “Outliers” change the loss distribution in ENSO phases• Observation: approximately 45% of losses in El Nino are Hail –

counter-intuitive

Proportion of Loss Amount (ex TC Tracy, ex Sydney Hail, ex "Ash Wednesday")

14% 12%

25%15% 45%

4%13%

14%

44%

32%

32%

13%28%

9%

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

El Niño Neutral La Niña

Storm

Hail

Flood

Cyclone

Bushfire

ENSO by Peril

• Vast majority of bushfires occur in El Nino phase• Cyclones occur in La Nina phase• Rest is not very clear – similar split to time spent in each phase – slight bias

towards La Nina phase

Proportion of Loss Amount

63%

15%23% 18% 20%

31%

37%

16%

57% 30%

70%

43%

46%

69%

34%47%

12%

37%23%

9%0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Bushfire Cyclone Flood Hail Storm Storm/Hail/Flood Time Split

El Niño Neutral La Niña

ENSO by Peril – excluding “outliers”

• Similar observations when the “outliers” are removed• Even more similarity between storm/hail/flood losses and

the time spent in each phase

Proportion of Loss Amount (ex TC Tracy, ex Sydney Hail, ex "Ash Wednesday")

40%26%

33%18%

25% 31%

60%

28%57%

43% 70% 54% 46%

45%34%

24%12%

22% 23%

9%0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Bushfire Cyclone Flood Hail Storm Storm/Hail/Flood Time Split

El Niño Neutral La Niña

ENSO by Peril – Event Counts

• Size of the event can skew the representation of losses over time• Frequency of losses shows that:

– Bushfires tend to occur during the El Nino Phase– Cyclones during Neutral and La Nina– Storm/Hail Flood during Neutral and El Nino phases

Proportion of Losses

63%

20%32% 38% 32%

38%

40%

69%46%

42% 47%

40%

15%22% 20% 20%

15%0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

Bushfire Cyclone Flood Hail Storm Storm/Hail/Flood

La Niña

Neutral

El Niño

Summary of resultsEl Niño Neutral La Niña

Probability of a loss in a monthBushfire 7% 3% 0%Cyclone 3% 3% 7%Storm/Hail/Flood 21% 31% 13%

Average Monthly Loss ($m)Last 10 years 57 62 45Last 20 year 38 50 36

Average Number of Months in a phase 7 6 7

Analysis of extreme events - Scientific viewPeril La Niña Neutral El Niño 5 year Forecast

N. of Tropic of CapricornCyclones NA Stability in occurrence

Floods NA Below Average

S. of Tropic of CapricornThunderstorms – Sydney NA Below Average

Thunderstorms – Brisbane NA Above Average

Cyclones NA Below Average

East Coast Lows NA Stability

Floods NA Below Average

Bushfire NA Above Average

South Pacific BasinCyclones West 170o NA NA NA Stability

Cyclones East NA Above Average

Floods NA Above Average

Climate Change

Conclusions• Weather has been the primary cause of

Australia’s historic catastrophe losses• Framework for considering losses from

weather perils– Not a definitive study– Some relationships are clear– Others may need more investigation

• Climate Change will make prediction even more difficult