2. Collapsed steel pylon.
Cover: Aerial view of ruins of buildings
which collapsed in the massive earthquake
and a quake-induced lake in Tangjiashan,
Beichuan county, Mianyang city, southwest
Chinas Sichuan province, 30 May 2008.
Contents
Low-temperature freezing rain and snow weather
in China in January 2008 Page 1
The 2008 Sichuan earthquake: A predictable
disaster with unpredictable consequences? Page 7
3. The China Meteorological Administration gave the
system the name “low-temperature freezing rain
and snow weather – ”, which
clearly indicates that there were no high wind
speeds involved in this event. This detail is very
important since in a strictly technical sense the
terms “storm” and “blizzard” are not valid without
wind speed being involved.
The total direct economic losses reached
RMB 151.6bn (>US$ 21bn) according to the Civil
Affairs Minister (24 February 2008). The Chinese
insurance industry was quite hard hit, with
estimated losses in the range of RMB 5–12bn
and a record number of over 1,000,000 claims.
Even though not all the losses have yet been
settled, it is already clear that the snow weather
event will become one of the largest insured
losses in China’s insurance industry.
Low-temperature freezing rain and snow weather
in China in January 2008
Introduction
In the past, weather-related natural catastrophes
like floods and typhoons have led to large losses
for the insurance industry worldwide as well as
for China. The winter of January 2008 proved to
China that loss potentials from snow and ice
catastrophes can create large insured losses and
that these loss potentials are not negligible.
From 10 January 2008, unusually low tempera-
tures, icy rain and heavy snow affected over 18
provinces for several weeks. Badly impacted were
air, rail and road traffic, as well as the food supply.
Downed power lines brought central, eastern and
southern China to a virtual standstill. Worst hit
were the provinces Hubei, Hunan, Anhui, Guizhou,
Jiangxi and Guangxi, while the provinces Zheji-
ang, Sichuan, Henan, Jiangsu, Shanghai, Xinjiang,
Gansu, Shaanxi, Qinghai, Ningxia, Yunnan,
Chongqing and Guangdong also suffered.
Fig. 1: Illustration of the affected areas
Unusual low temperature, heavy snow and
freezing rain had affected a large part of
China (beige), some provinces suffering
severely (red). Locations of severe snow
fall and ice rain are indicated.
Mainly affected regions
Affected regions
Snow
Ice rain
Qinghai
Xinjiang
Ningxia
Henan Jiangsu
Zhejiang
Jiangxi
Hunan
Guangdong
Guangxi
Yunnan
Sichuan
Guizhou
AnhuiHubei
Shaanxi
Gansu
Chongqing
1Munich Re Severe winter and earthquake
4. Meteorological situation
The typical January weather pattern in China is
characterised by a high-pressure system above
Siberia with cold air continuously flowing from the
north. The unusually strong flow in 2008 brought
the Siberian cold air further south than in a normal
year. At the same time, warm moist air from the
Gulf of Bengal penetrated into southern China,
sustained by a rather strong high pressure zone in
the northwest Pacific.
Many meteorological records were set during the
cold spell. In January and February 2008, the
ground temperature in most of southern China
was well below average. For example, in parts of
Hubei, Hunan, Guizhou and Gansu the tempera-
ture was over 4°C lower than normal and the mean
temperature reached a historical low.
Freezing rain occurs in the wintertime when rain
produced in a warm air mass falls through a
shallow cold air mass below, freezing on contact
with the ground or any other surface (if it is below
zero degrees) and instantaneously forming an
ice layer. In January 2008, these conditions were
present.
In short, the general weather situation in January
2008 led to a weather pattern in which continu-
ously cold air from the north met with a warm and
moist air mass from the sea.
What turned the event into a catastrophe was the
unusually long duration of the cold spell (over
three weeks), a massive area affected by snow
(over 1.3 million km2) and the phenomenon of ice
accumulations. Central and southern regions –
especially in Guizhou, Hunan and Jiangxi with a
climate that usually does not have prolonged
periods with temperatures around freezing point –
experienced ice accumulations measuring
30–60 mm.
Whether there is a strong connection between the
catastrophic event of January 2008 and the per-
sisting La Niña situation is not easy to say. La Niña
conditions supported the high-pressure system in
the northwest Pacific that blocked the cold air
mass over land in southern China and played a
supporting role, but cannot be held responsible as
the actual cause.
Loss analysis
Altogether 129 people were killed, many of them
in car accidents on frozen roads. Over 1.6 million
people were evacuated and many passengers
were stranded. For example, at Guangzhou train
station, over 800,000 people had to wait several
days for transportation home for the Chinese New
Year. Due to power outages and snowed-in tracks,
over 130 trains were out of commission. In China,
electricity is primarily generated using coal, the
transport of which was delayed, thus causing
further blackouts. In the extreme weather, more
than 480,000 homes were destroyed and nearly
12 million hectares of farmland damaged in the
regions afflicted by snow and ice rain.
Insured losses
The insured loss pattern was characterised by a
huge number of motor and property damage
claims, especially within the industrial and power
sector. The present insured loss estimate lies
within the range of RMB 5–12bn. Claims amount-
ing to roughly RMB 5–6bn have already been
regulated, mainly losses from motor and industrial
lines business as well as some of the T&D line
losses. A further RMB 6–7bn concerns claims in
the power sector mainly due to snapped T&D lines
and collapsed transformer towers. Hence, in terms
of loss amount, property claims dominate by far,
with over 70% of losses relating to the power
sector. The farming and residential sectors account
only for a marginal loss amount. BI and CBI also
played only a minor role.
Fig. 2: Schematic representation of the
atmospheric condition
The January weather situation had a high-
pressure system above Siberia causing
continuous flow of cold air from the north
(blue), while from the south warm moist air
(red) penetrated into southern China.
Cold air
Subtropical
high
Siberian high
Warm SST
rm moist airWarmrm
2 Munich Re Severe winter and earthquake
Low-temperature freezing rain and snow weather in China in January 2008
5. Motor business – Many road accidents
Snow and ice on the roads accounted for a higher
than normal number of car accidents at the
beginning of the cold period. As the snowy
weather continued, many cars could not be
moved due to blocked roads. The average loss
per claim was around RMB 900, and the massive
number of claims (approx. 800,000) was quite
quickly regulated. In severely affected areas such
as Hunan, motor insurance density is still quite
low. By comparison, during the Canada ice storm
in 1998 about 22% of the motor premium was
consumed by losses from the ice event.
Property building and contents – Collapsed roof
and lightweight metal construction
The loss pattern was characterised by numerous
collapsed roofs and consequential water damage
on machinery and/or stock. Especially the indus-
trial sector was hard hit, with many manufacturing
plants and warehouses being affected, even
though the average snow depth was below 50 cm
and in most parts even below 30 cm. The main
reason was that the standard metal construction
could not withstand the weight on top. In Asia,
it is common to use a so-called lightweight metal
construction for the roofing. This type of construc-
tion is clearly not designed to withstand either
the weight of wet snow and/or ice or strong wind
speeds in the case of storms and so has to be
considered as highly vulnerable to these perils.
Property transmission and distribution lines –
Collapse mainly caused by ice
The massive ice accumulations severely affected
the power sector, with many snapped T&D lines
and collapsed transformer towers. Consequently,
a large number of cities were without power,
in many cases over several weeks. From the tech-
nical point of view, it is quite clear that the
transmission lines were not built to withstand
such an amount of ice.
Loss evaluation
For the Chinese insurance market, the winter of
2008 was one of the costliest events ever.
Historically, only the large floods (1998, 1996 and
1991) were more costly based on 2008 values.
Compared to the loss experience in other coun-
tries, for example Canada in 1998 (>US$ 2bn at
today’s values) or the 1993 blizzard affecting the
USA and Canada (>US$ 4bn at today’s values), the
Chinese snow and freezing rain event ranks third
among winter events worldwide.
In China, only 3–5% of industrial and commercial
risks have insurance cover. In the USA, almost
all risks are covered, while the figure in Canada
stands at over 50%. Given that the Chinese insur-
ance market is one of the fastest growing, it is
evident just how enormous the loss potentials are.
Assuming future growth, similar building quality
and the same quality in loss prevention and
insurance regulation, a future winter event of
equal dimensions would affect much higher
insured values. Therefore, an insured loss poten-
tial of over US$ 5bn from a snow storm in
China is easily conceivable.
The three most expensive winter events worldwide (1980–2008)
Period Event Losses (US$m, original values) Deaths
Overall losses Insured losses
11.–14.3.1993 Blizzard US and Canada 5,000 1,975 270
10.1.–13.2.2008 Winter damage China > 21,000 1,600 129
4.–10.1.1998 Ice storm Canada 2,900 1,150 45
The most expensive weather-related events for China´s insurance industry (1980–2008)
Period Event Losses (US$ m, original values) Deaths
Overall losses Insured losses
10.1.–13.2.2008 Winter damage > 21,000 1,600 129
May–Sept. 1998 Floods 30,700 1,000 4,159
27.6.–13.8.1996 Floods 24,000 445 3,048
May–Sept. 1991 Floods 13,600 410 2,628
9.–16.8.2006 Typhoon Saomai – No. 8 1,400 200 460
18.–20.9.2007 Typhoon Wipha (Goring), No 8 1,000 200 7
24.–25.6.2001 Typhoon Chebi 470 120 150
19.–20.7.2005 Typhoon Haitang (No. 5) 1,000 85 5
6.– 8.8.2005 Typhoon Matsa (No. 9), floods 850 85 19
3Munich Re Severe winter and earthquake
Low-temperature freezing rain and snow weather in China in January 2008
6. Underwriting aspects
Special care has to be taken when writing natural
catastrophe business. For example, in order to
minimise uncertainties in estimating the risk, the
need to secure good information on the exposure
(such as complete, up-to-date and high-quality
data) is absolutely essential. Another very basic
recommendation is “never give a cover for free”:
any natural peril covered has to have a price, even
those that have not occurred in the recent past.
Worldwide experience has repeatedly proven that
catastrophes tend to get forgotten rather quickly.
In order to be able to control exposure, working
with limits (loss occurrence limits, annual aggre-
gate limits, deductibles as a percentage of the
value) is highly recommended. Controlling poten-
tial loss accumulations is another essential step,
i.e. limiting exposure in high-risk zones of each
respective peril. It is also necessary to monitor
exposure of highly vulnerable risks, e.g. light-
weight construction types when covering snow
pressure or underwriting T&D lines.
How to deal with snow pressure and
freezing rain losses
Attention has to be paid not only to losses due
to snow pressure, freezing rain and other winter
perils but also to their regulation. In view of the
many roofs that collapsed under the weight
of snow and the many power lines downed due
to the weight of ice, one of the key challenges
concerns how to handle these losses. Wherever
(in primary insurance) snow and ice pressure
are covered perils, the handling of mass claims
resulting from the peril should be regulated.
Among the challenges are how to find a clear
definition of a loss occurrence and how reinsurance
can be offered for snow pressure and freezing
rain losses. Usually an occurrence definition con-
tains causal, geographical or temporal criteria.
A causal criterion requires individual losses to be
ascribable to a common cause.
In the case of snow pressure, it is very difficult to
define the common cause: Is it the last snowfall?
Or the total snowfall of the season? Or the general
weather pattern of the winter? In principle, the
general weather pattern of the winter can be
regarded as the common cause for all snow-pres-
sure losses, and hence all roof collapses can be
subsumed under one loss occurrence, assuming
the insurer and reinsurer agree that the general
weather pattern is the common cause. Clearly, dis-
putes are pre-programmed in such a situation.
In some wordings, all individual losses must arise
from one event. In that case, there is a further
difficulty: is a prolonged cold winter an event?
The term general weather pattern cannot be
considered as one event. Using the last snowfall
as the trigger for the loss occurrence resolves the
challenge of defining a common cause (and
event). So each loss occurrence would involve
roofs that collapsed from one and the same
snowfall.
A practical difficulty is that very often the exact
time of the loss cannot be determined and hence
it is not easy to link the loss directly to the various
snowfalls. Furthermore, meteorological data are
not always easily accessible in every country and
hence individual snowfalls cannot be mapped.
Scientifically speaking, it is the total weight of all
preceding snowfalls that causes the roof to
collapse, so the accumulated weight of many
snowfalls (events) remains on the roof without
causing any damage, and then an additional
normal snowfall is sufficient to cause collapse.
Reinsurance for losses caused by snow pressure
and freezing rain is mostly provided under
programmes containing standard occurrence
definitions (168-hours clause for all events that are
not typhoons, floods, etc.). However, practice has
shown that these wordings are not always suit-
able. In order to reinsure losses caused by snow
pressure, other possibilities could be considered,
i.e. a special winter aggregate XL or an extension
of the loss occurrence clause.
How to deal with transmission and
distribution lines
The loss amount due to T&D line failure was
unexpectedly high and accounted for the lion’s
share of the property loss. One of the challenges
is how to deal with such losses in the post-
event claims regulation. Furthermore, (primary)
insurance terms have to be assessed.
4 Munich Re Severe winter and earthquake
Low-temperature freezing rain and snow weather in China in January 2008
7. The owners of T&D lines like power producers,
power distributors, and network operators play a
very important role in distribution and service
supply. When they fail, these services are quite
difficult to put back into production.
Besides property damage in the form of physical
damage to the T&D line network (mast, line,
transformer), the end consumer can be impacted
very severely following power failure. For ex-
ample, heavy industry relying on cooling facilities
could experience a total loss (e.g. explosion of a
steel plant due to overheating) or, as happened
in Canada during the great ice storm, the coverage
of frozen goods in household policies led to a
significantly higher loss.
T&D lines are characterised by a widespread
physical distribution network. Due to their high
susceptibility to natural hazards, especially wind-
storm and ice, these above-ground lines harbour
a major loss accumulation potential. The age of
the network, the quality of maintenance, and the
design standard (distance of poles, material used)
of the masts and lines used have an impact on
the actual exposure and vulnerability. Besides
natural hazards themselves, other factors with a
substantial influence on T&D line exposure
include, for example, the capability to minimise
service downtime (i.e. having a back-up network
of another provider). Of course, the market
standards of the primary insurance policies play
an important role as well.
Special attention has to be given to risk definition,
risk quality and sum insured as well as claims
handling in the event of a loss. Experience has
shown that underwriting practice and policy
wordings are not always fully customised to the
special requirements of T&D risks.
Sums insured under policies written years ago
have not been adjusted to the new replacement
value, which resulted in considerable under-
insurance. Also no provision was made in the
policies when the T&D lines were upgraded
following reconstruction to meet higher quality
standards. Furthermore, there were no provisos in
the policies with regard to increased reconstruc-
tion costs because of the urgency to get the power
grid back on stream. All these factors should be
given due consideration with a view to obtaining
adequate terms and prices and to avoiding
difficulties with claims settlement for the future.
From the reinsurer’s perspective, these high-risk,
above-ground T&D lines can be covered within a
reinsurance treaty subject to certain restrictions.
In general, a distance of 1,000 m from the insured
premises (power plant) for coverage of T&D lines
should not be exceeded (covering the physical
loss, i.e. wires, cable, poles, pylon, masts, trans-
former).
Coverage for T&D lines on a standalone basis is
also possible, but the underwriting requires
full transparency of the risk and coverage on the
insurance and reinsurance side. From a rein-
surance perspective, non-proportional coverage
appears to be more appropriate because pro-
portional reinsurance could result in undesired
risk accumulations.
Conclusion
The Chinese snow and freezing rain event has
shown the enormous loss potential from such
events. The risk due to winter perils was ap-
parently underestimated. Consequently, there is
need for more appropriate conditions and prices
on the primary and reinsurance side.
Only 15 mm of ice were sufficient to down
distribution lines.
5Munich Re Severe winter and earthquake
Low-temperature freezing rain and snow weather in China in January 2008
8. Many buildings were not constructed to
withstand the strong shaking of the
Sichuan earthquake. Total failure of the
building structures, like in this 40-year-old
building from Leigu Town, was common in
the epicentral area.
9. Introduction
On 12 May, at 2.28 p.m. local time, the earth
started shaking heavily in the Longmenshan
mountains in western Sichuan. About one minute
later, when the rupture came to an end almost
250 km further to the northeast at the southern
tip of Gansu province, one of the biggest disasters
in Chinese recent history had happened. Accord-
ing to official figures, more than 69,000 people
had been confirmed dead by the end of July,
and an additional 18,000 are still missing. 374,000
people were injured, some 50,000 of them left
with physical disabilities. At least 4.8 million have
been made homeless.
Rivers and roads were blocked by innumerable
landslides, many bridges destroyed, and elec-
tricity, water supply, and telecommunications
interrupted. The resulting damage caused was
tremendous. Official figures put the direct eco-
nomic damage at about 845 billion Yuan: values
approximately equivalent to the GDP produced in
the province in Sichuan in 2007 were destroyed in
only one minute of intense shaking in a small
and remote part of Sichuan province, without
seriously affecting any major city in the province.
The final costs of reconstruction will be even
higher than the direct economic damage.
Reconstruction will take years and require tremen-
dous efforts from the whole country. Only then
will the impact of this catastrophe on Chinese
society and the local economy be fully revealed.
The effect on the insurance industry will be felt
earlier. Only a small portion of the direct damage
seems to be covered by insurance.
Due to sub-standard data about the exposures
covered and their exact properties, such as
location, initial loss estimates turned out to be
very difficult and unreliable. Even now, four
months after the event, the exact amount of the
insured loss is still unknown as various issues
about coverage had not been clarified in advance.
So far, only 386 million Yuan have been paid
out to insureds.
The Sichuan earthquake:
An expected event?
The epicentre of the Sichuan earthquake is located
in one of the steepest mountain areas of the world,
at the eastern rim of the Himalayas. Mountains are
often the expression of an active tectonic regime,
of ongoing compression, often going hand in hand
with strong and frequent earthquakes.
Strong earthquakes have in fact been recorded
in the area in the past. The latest occurred in
August 1976, when a strong M 7.2 earthquake and
two strong aftershocks of M 6.7 and 7.2 struck
some 100 km further north, in the vicinity of the
city of Songpan. The 1933 Diexi earthquake in the
same region also led to significant damage.
All this seems to provide an easy answer.
But only in 2007 was it published in a scientific
publication that this particular region probably
contributes significantly to the earthquake hazard
in Chengdu, the capital of Sichuan province,
located about 60–70 km from the fault zone.
The actually measured convergence between the
Himalayas and the Sichuan basin to the east is less
than 3 mm/year along the Longmenshan fault.
In a recent hazard research project by the MIT, the
area was even excluded from further research as
the hazard was assumed to be too low. The fault
zone itself is not expressed as an obvious, single,
and clear fault line. Historically strong earthquakes
did not happen along the Longmenshan fault
itself, but north or south at different fault systems.
The 2008 Sichuan earthquake:
A predictable disaster with unpredictable
consequences?
7Munich Re Severe winter and earthquake
10. This is why the earthquake turned out to be a sur-
prise even to so many experts who were focusing
on the neighbouring faults. The shaking during the
Sichuan earthquake by far exceeded the shaking
assumed in the official building code map for
China, where the area was classified as being as
hazardous as Hong Kong. Contrary to Munich Re’s
internal model, which is based on a refinement
of the Global Seismic Hazard Assessment model
(GSHAP), the magnitude of this event was not
captured by all commercial earthquake risk
models.
A rough indicator for the return period of this
earthquake is provided by the offset. According
to the slip distribution model of the rupture
published by Caltech University (Sladen, 2008),
the offset along the 250 km rupture area was up
to 14 m, with an average of approximately
4–5 m.
This would lead to return periods far above
1,000 years for a similar earthquake along this
specific fault zone, provided that the measured
convergence is representative and released in
one single event. Paleoseismic analysis (Dens-
more et al, 2007) leads to roughly similar results.
This is also in agreement with the Munich Re
model, where a probability of once in 1,000 years
was assigned for such an event. In this model,
it is assumed that the deformation energy along
the Longmenshan fault will be released with
characteristic earthquakes of magnitude 7.9 ± 0.1.
The size of the fault system apparently does not
allow for significantly stronger earthquakes to
happen along this fault system. The hazard map
for a return period of 1,000 years from Munich Re’s
model is shown in Figure 1.
Clearly, this earthquake led to the redistribution
of stress along the many faults in this area.
Some of them will be more probable to rupture
after this event than they were before. Therefore,
it is likely that more earthquakes will occur in
the region. They cannot be predicted, but we can
prepare ourselves for them.
The building code and its
enforcement
The Chinese building code is described by
independent experts as fairly rigorous as well
as modern and up to date. Obviously, however,
this was not enough to avoid this dramatic
damage caused by the Sichuan earthquake.
There are several reasons that can be identified
as contributing to the disaster.
A main feature of a building code is a hazard
zoning map. The map underlying the Chinese
building code classifies the affected area as low
to medium hazard. Buildings built according
to the latest code are required to withstand 0.1 g
of horizontal acceleration. This is approximately
equivalent to an intensity of VII. Measurements
indicate that the real shaking in the epicentral area
was at least 4–5 times higher. Although this is
rather low for such a strong and shallow earth-
quake, especially as only little surface rupture
could be observed, the anticipated shaking as per
the code did not match the observed one.
In this respect, it has to be kept in mind that
building codes are based on a probability of
exceedance of approximately once in 500 years.
As mentioned before, the Wenchuan earthquake
was a very rare event, and produced correspond-
ingly a higher ground motion. In conclusion,
even if buildings were built according to the code,
significant damage could be expected.
Fig. 1: Hazard map
Hazard map of Munich Re’s probabilistic
earthquake model for Mainland China
for a return period of 1,000 years.
Red is high hazard, green is low hazard.
8 Munich Re Severe winter and earthquake
The 2008 Sichuan earthquake
11. In addition, calculations using Munich Re’s
internal model show that the expected shaking
for the code map seems to be underestimated.
But the zoning map was not the only problem with
the code: Many cases of failure to enforce the code
have been described; this became obvious during
the three-day field trip by Munich Re’s experts
six weeks after the earthquake. Especially outside
of towns and bigger villages, buildings were
constructed often with sub-standard materials
and insufficient steel reinforcements.
It was obvious that compliance with the building
code was taken more seriously in (major)
cities than even in their suburbs, not to mention
the rural areas. This led to a clear difference in the
damage pattern inside and directly outside
villages, where the risk of damage and death
was far higher.
Thirdly, it was clear that the implementation of
a stringent building code for new structures has
not been supplemented by reinforcement of older
structures. This was previously identified as a
serious problem in many earthquakes – especially
in economically under-developed and rapidly
growing regions.
Most of the destruction, however, can be at-
tributed to unregulated, non-engineered con-
structions. These were buildings built with the
simplest means and materials, consisting most
often of unreinforced concrete blocks, loosely
connected with some mortar, a wooden frame
structure and a light roof with shingles, offering
very little lateral resistance. Widespread
destruction of this kind of construction was
observed and contributed significantly to the
devastation, as shown in the top picture.
With respect to commercial and industrial risks,
it has to be concluded that none of the observed
damage really came as a surprise. Many damaged
structures consisted of unreinforced masonry
(URM) or non-ductile concrete. Both are known
to perform badly in strong earthquakes. Often,
insufficient bracing of the ceiling could be
observed, resulting in the total collapse of the
roof, as shown in the bottom picture.
Residential area close to Mianzhu.
Damaged industrial plant.
9Munich Re Severe winter and earthquake
The 2008 Sichuan earthquake
12. Is this earthquake over now?
When the main rupture came to a stop in Gansu
province, the main factor leading to loss of lives
also ended. However, the hazard is still above
average. Many aftershocks occurred since 12 May,
the biggest so far with a magnitude of 6.0 and
causing additional damage and casualties.
The main shock and recorded aftershocks are
shown in Figure 2.
Many of the steep slopes in this mountainous
area failed, resulting in tremendous landslides.
The remaining soil is not at all stable. Landslides,
triggered by aftershocks or even a light amount
of rain will be a permanent hazard in the coming
months. These landslides often impound the water
of the rivers in the valleys, forming natural, but
most often instable dams and creating the so-
called “quake lakes”. Hundreds of these could be
observed directly after the earthquake, and they
posed a significant risk. It has been possible
through enormous efforts to prevent such a lake
from bursting, especially the dangerous Lake
Tangjiashan, which threatens more than one
million people living downstream. However, that
meant inundating some of the already heavily
destroyed and evacuated cities, like Beichuan.
As the risk of landslides is likely to persist for a
long time to come in this area, the creation of
quake lakes will also be a risk we have to keep our
eye on for a long time.
How dangerous these temporary lakes can
become was manifested by an earthquake in 1786
also in Sichuan province, where a landslide dam
burst, killing more than 100,000 people down-
stream of Dadu River.
The release of stress along the Longmenshan fault
might also have led to a shift of the stress within
the earth’s crust towards the northeast and/or the
southwest, increasing the likelihood of a strong
earthquake there significantly.
Risk management
In Chinese history, several earthquakes have
caused far more damage and devastation than the
Sichuan earthquake. They are listed in Figure 3.
It is simply a question of time until the next
devastating earthquake hits China, causing even
more fatalities or economic damage.
According to Munich Re’s earthquake model,
economic damage approximate to that caused by
the Sichuan earthquake is likely to be reached
or even exceeded by another earthquake in China
every 20–25 years. Risk management of capital
at risk should focus on events with a far lower
probability of occurrence, like once in 200, 500,
or even 1,000 years, depending on the risk appetite
of individual companies. Market losses calculated
for a probability of once in 200 years are about
10–15 times higher than calculated for the Sichuan
earthquake, according to the results from Munich
Re’s earthquake model.
Fig. 2: Modelled intensity field and
aftershocks
Main (blue) and aftershocks (white) with
magnitude higher than 5 up to 12 Septem-
ber from the 12 May M 8.0 Sichuan earth-
quake underlain by intensity field as per
Munich Re’s earthquake model.
High
Low
10 Munich Re Severe winter and earthquake
The 2008 Sichuan earthquake
13. Fig. 3: List of earthquakes with more than 20,000 fatalities in China since 1,000 A.D.
Source: Munich Re
Date Area Magnitude Number of deaths
9.1.1038 Shaanxi 23,000
1057 Gulf of Chihli M 6.7 25,000
Aug. 1289 Che Li 40,000
27.9.1290 Gulf of Chihli M 6.7 100,000
1303 Shanxi M 8.0 200,000
23.1.1556 Shaanxi M 8.0 830,000
1654 Gansu M 8.0 31,000
25.10.1662 Anxiang 150,000
25.7.1668 Shandong M 8.5 50,000
1679 Hebei M 8.0 45,000
May 1695 Shanxi M 8.0 30,000
19.6.1718 Gansu 43,000
30.11.1731 S of Beijing M 6.5 100,000
3.1.1739 Ningxia M 8.0 50,000
1879 Gansu M 8.0 30,000
16.12.1920 Gansu M 8.5 235,000
22.5.1927 Nanshan M 8.0 40,000
25.12.1932 Gansu M 7.6 77,000
1.5.1974 Yunnan M 6.8 20,000
27.7.1976 Tangshan M 8.2 242,000
12.5.2008 Sichuan M 8.0 70,000
Assessing and writing cat risks
The Sichuan earthquake is part of the event set in
Munich Re’s probabilistic earthquake model for
China, which translates direct and indirect effects
from more than 250,000 stochastic earthquakes
into losses for an individual client’s portfolio.
This model, based on more than 50 different
area source zones and more than 200 faults, is
designed to reflect the complete earthquake
hazard in China.
From what is known so far, it can be said that the
Sichuan earthquake has been taken into account
accurately. Despite the high degree of uncertainty,
the estimated loss from the model turns out to
be in line with observations. The modelled inten-
sity field is shown in Figure 2 as well.
However, the earthquake highlighted a couple
of critical issues for the current market.
Pricing
Due to high competition and in order to attract
business, earthquake extension in insurance
policies was often given away for free. This is in
line with recent experience. For quite a number
of years, no significant earthquake has happened
in China; losses from natural catastrophes have
been restricted to typhoons and inundations.
This earthquake made very clear that this percep-
tion is distorted. Significant losses can and do
arise from earthquakes. This has to be adequately
reflected in the price of insurance cover.
Data quality
Far too little information about the risks has been
captured by the insurance industry – and even
less has been communicated to the reinsurers and
brokers. Reinsurers had to rely on one exposure
figure only for the province of Sichuan. In some
cases, these figures had been broken down into
different lines of business as well as into buildings,
contents and business interruption. Information
about types of construction was not available
at all.
In the aftermath of the earthquake, no company
could immediately say with a minimum of
certainty how much exposure it had in the affected
area and what quality the business was. That
made loss estimates and thus loss reserving very
difficult and dependent on many assumptions.
It should be understood that, among other things,
the earthquake served as a warning bell in this
respect. In consequence, clear data standards for
risk management and reporting between insurers
and reinsurers have to be defined, as is common
practice in other countries. This certainly includes
zones for exposure reporting (CRESTA zones).
The currently used status of province-level
reporting has been proven by this earthquake to
be virtually useless. Four- or even six-digit postal
code zones would enable a much more realistic
risk assessment.
11Munich Re Severe winter and earthquake
The 2008 Sichuan earthquake
14. Lack of coverage for private households
The very low insurance density in Sichuan led to
the low contribution of the insurance industry
in the financial management of the losses. This
unsatisfactory situation is partly due to the fact
that earthquake coverage was not offered for
private house owners before the event occurred.
Providing minimum protection for people seeking
protection for their belongings and houses is
a field-tested approach. As in other regions after
comparable events (Taiwan 1999, Turkey 1999),
the establishment of a pool providing minimum
natural catastrophe coverage could be a suitable
solution.
Final remarks
“This disaster should serve as a warning to
countries in high-seismic-risk areas. It is vital to
identify seismic hazards and to develop retrofit
programmes for hazardous structures. The
immense damage seen in Sichuan could have
been avoided through seismic risk management”
(Miyamoto, 2008).
Perhaps not entirely avoided, but certainly signifi-
cantly reduced! As the next destructive earthquake
in China is only a matter of time, coordinated
efforts for reducing the devastating effects of
natural catastrophes should be initiated as soon
as possible.
Apart from the suggested reinforcement
programme for existing buildings, the insurance
industry should also be encouraged to provide
solutions as mentioned above. The mismatch
between the enormous economic loss and the
small fraction of it taken by the insurance sector
fuels the discussion on how to distribute the
risk between the private insurance sector and the
public in future.
In the aftermath of this earthquake, it became
clear that the main aspirations are gaining more
transparency with regard to coverage conditions
and knowing what payments can be expected
from which risk carrier.
Insurance and reinsurance have a long tradition
in managing and financing risk from natural
catastrophes in many parts of the world. It can
provide the tools with a view to playing an im-
portant role in this process in China as well.
References
Densmore, A.L., Ellis, M.A., Li, Y, et al (2007):
Active tectonics of the Beichuan and Pengguan faults at the
eastern margin of the Tibetan Plateau, Tectonics, Vol. 26.
Miyamoto (2008): 2008 M8 Sichuan, China earthquake field
investigation report. 45 pages.
Sladen, A. (2008), Caltech. Copied on 15 September 2008
from http://www.tectonics.caltech.edu/slip_history/2008_e_
sichuan/e_sichuan.html
12 Munich Re Severe winter and earthquake
The 2008 Sichuan earthquake
