Floods, Fires, Earthquakes and Pine Beetles:
JARring Actions and Fat Tails
Berkeley Symposium on
Real Estate, Catastrophic Risk, and Public Policy
UC Berkeley, March 22-23, 2006
Thanks to Alan Berger, Carolyn Kousky, Erzo Luttmer
Mistreating the Mississippi:
JARring Actions that Fueled the Floods
Work with Carolyn Kousky
Mistreating the Mississippi
• There has been much examination of what went
wrong once Katrina was on the radar screen
heading for New Orleans.
• Our analysis is on decisions made long before
Katrina but that nonetheless exacerbated her
impacts. For example:
 1850: Swamp and Overflow Land Act (filling wetlands)
 1944-63: 6 dams planned for and constructed on the
Missouri (“Big Muddy”)
 1965: Completion of Mississippi River Gulf Outlet
• These are examples of JARring actions.
Overview of Presentation
• What are JARring actions and what
challenges do they pose for policy?
• Hurricane Katrina as an example of a
disaster that was exacerbated by JARring
actions
• Potential policy responses
JARring Actions
• Acronym JARring: Jeopardize Assets that are Remote
• A particular type of externality: cost imposed on
others who are spatially or temporally distant
• Could also be distant due to the probabilistic nature of
imposed costs
• Can often lead to a loss of “ecosystem services” such
as flood mitigation and storm surge attenuation by
coastal wetlands
• Building in wildfire zone or on fault
• Suppressing fires
The Problem
Normal mechanisms to control
externalities (e.g., liability,
contracting, political regulations for
those in jurisdiction) do not work
well for JARring actions.
Policy Challenges
• Hard to assign responsibility for
consequences  plausible deniability
• Collective action problems
• People tend to look for local causes
• Scientific uncertainty
• Hard to calculate changes in risk levels
Applications
• Here, we focus on JARring actions that
increase the vulnerability of groups to
natural disasters, particularly flooding, but
the concept has much wider applications.
• A key application is climate change.
Greenhouse gas emitting actions may come
to be viewed as the ultimate JARring
actions.
Examples of JARring Actions
• Private entities face no incentive to consider how
their actions affect those distant from them.
• Examples:
–
–
–
–
–
Filling wetlands
Increasing impervious surface area
Construction of agricultural levees
Increasing drainage of agricultural land
Lobbying for structural projects that though privately
beneficial mistreat the Mississippi
Public JARring Actions
• Historically, government (particularly USACE)
has not considered the external costs imposed by
large structural projects.
• Examples:
–
–
–
–
Levees
Canals
Dams
Responding to lobbying efforts of concentrated interest
groups
Coastal wetland loss and Katrina:
losing a storm buffer
(PHOTO BY ELLIS LUCIA / The Times-Picayune)
Each year an area of marsh close to the size of Manhattan is lost.
Tracking the Loss
1973
Source:
http://www.publichealth.hurricane.lsu.edu/Louisian
a%20Coastal%20Land%20Loss.htm
2000
Causes of Loss:
Lack of Sediment Reaching Gulf
Sources: http://www.lewis-clark.org/content/content-article.asp?ArticleID=1412
http://www.industcards.com/hydro-usa-ne-dakotas.htm
Causes of Loss:
Sediment Sent Out to Sea
The sediment is “shot over the shelf like peas through a peashooter, and
lost to the abyssal plain.” - John McPhee
Causes of Loss:
Dredging of Canals
Source: http://marine.usgs.gov/fact-sheets/LAwetlands/lawetlands.html
Other causes of loss: subsidence due to lack of sediment and
exacerbated by withdrawals of oil and natural gas
MRGO
(Mississippi River Gulf Outlet)
http://www.saveourlake.org/wetlands.htm
Issues Left to the Reader
(Due to time limitations)
• Government Action Needed
• Possible Policy Responses
• The Mistake of Planning for the X-Year
Flood, Fire, Quake, or Infestation
• Incentives, Insurance and Planning for the
Future
ec
e
N mb
ov er
em , 2
0
O ber 05
c
Se tob , 20
pt er 05
em , 2
be 00
Au r, 5
gu 200
st
,2 5
Ju 00
ly, 5
Ju 20
ne 05
,
M 200
ay
,2 5
Ap 0
05
r
M il, 2
a
Fe rch 005
br , 2
ua 00
Ja ry, 5
D nua 20
ec
0
em ry, 5
2
N
ov ber 005
em , 2
0
O ber 04
c
Se tob , 20
pt er 04
em , 2
be 00
r, 4
20
04
D
Number of times
Contagion in Disaster Attention
Number of times "avian flu" mentioned in U.S.
news sources as tallied from Lexis-Nexis
500
450
400
350
300
250
200
150
100
50
0
Month
What’s Wrong with This Picture?
Phoenix suburb
This Picture Too.
Littleton, Mass.
Historically Significant Wildland Fires 1988-1999
Continental U.S.
Date
Name
Location
1988
Yellowstone
Montana and
Idaho
1988
Canyon Creek
June 1990
Acres
Significance
1,585,000
Large Amount of Acreage
Burned
Montana
250,000
Large Amount of Acreage
Burned
Painted Cave
California
4,900
641 Structures Destroyed
June 1990
Dude Fire
Arizona
24,174
6 Lives Lost
63 homes destroyed
October
1991
Oakland Hills
California
1,500
25 Lives Lost and 2,900
Structures Destroyed
August
1996
Cox Wells
Idaho
219,000
Largest Fire of the Year
1998
Flagler/St. John
Florida
94,656
Forced the evacuation of
thousands of residents
August
1999
Dunn Glen
Complex
Nevada
288,220
Largest Fire of the Year
86,700
Hundreds of people were
evacuated by this complex of
fires that burned for almost 3
months
September November
1999
Kirk Complex
California
FAT TAILS: Observe massive multiple of largest to next in three categories:
Largest fire of year, people evacuated, and structures destroyed.
Largest and Deadliest Earthquakes by Year 1990-2005
Largest Earthquakes
Deadliest Earthquakes
Year
Date
Magnitude
2005
03/28
8.7
1,313 Northern
10/08
7.6
2004
12/26
9.0
283,106 Off West
12/26
9.0
283,106 Off West Coast
2003
09/25
8.3
0 Hokkaido,
12/26
6.6
31,000 Southeastern
2002
11/03
7.9
0 Central
03/25
6.1
1,000 Hindu Kush
2001
06/23
8.4
138 Near Coast
01/26
7.7
2000
11/16
8.0
2 New Ireland
06/04
7.9
1999
09/20
7.7
2,297 Taiwan
08/17
7.6
1998
03/25
8.1
0 Balleny
05/30
6.6
4,000 Afghanistan-
1997
10/14
7.8
0 South of Fiji
05/10
7.3
1,572 Northern Iran
12/05
7.8
0 Near East
1996
02/17
8.2
166 Irian Jaya
02/03
6.6
1995
07/30
8.0
3 Near Coast
01/16
6.9
5,530 Kobe, Japan
10/09
8.0
49 Near Coast
1994
10/04
8.3
11 Kuril Islands
06/20
6.8
795 Colombia
1993
08/08
7.8
09/29
6.2
9,748 India
1992
12/12
7.8
12/12
7.8
2,519 Flores Region,
1991
04/22
7.6
10/19
6.8
2,000 Northern India
06/20
7.4
1990
12/22
7.6
07/16
7.7
Fatalities Region
Sumatra,
Indonesia
Coast of
Northern
Sumatra
Japan
Region
Alaska
of Peru
Region,
P.N.G.
Islands
Region
Islands
Date Magnitude Fatalities Region
80,361 Pakistan
of Northern
Sumatra
Iran
Region,
Afghanistan
20,023 India
103 Southern
Sumatera,
Indonesia
17,118 Turkey
Tajikistan Border
Region
Coast of
Kamchatka
Region
Indonesia
of Northern
Chile
322 Yunnan, China
of Jalisco
Mexico
0 South of
Mariana
Islands
2,519 Flores
Region,
Indonesia
75 Costa Rica
0
Indonesia
Kuril Islands
1,621 Luzon,
Philippine
Islands
50,000 Iran
FAT TAILS: Observe massive multiple of largest to next in fatalities.
Earthquakes and People
• We routinely build in areas at risk from
earthquakes – consider the Bay Area.
• This is largely because earthquakes are an
“amenity risk.”
• Amenity Risks: risks that have associated
with them a benefit
• Compare to Noxious Risks
Comparing population density to
earthquake hazard
Source: Wikipedia, USGS
USGS time dependent map of earthquake
hazard in the next 24 hours
Source: USGS
Protection responds to investment.
Investment responds to protection.
• Housing pressures in the Bay Area have led to
increased development in the Sacramento-San
Joaquin Valley.
• The area is prone to flooding and levees in the
region are in need of repair.
• State lawmakers are considering upgrading
protection in the region and the state’s
congressional delegation is working to secure
federal funding.
• Local officials are also requiring builders to
increase protection.
Forest Fires
• As we expand into the urban-wild interface,
our investments face greater risk of fire
• This is another example of an “amenity
risk” – along with higher risk of fire, people
experience peace and quiet, scenic views,
etc.
• Risk reduction includes actions such as
clearing fuel around homes
Pine Beetles
• Largest infestation in North American history.
• Infest area in British Columbia three times the size of
Maryland.
•
•
•
•
•
•
12 western states
Attacks old, even-aged trees
Result of clear cutting and wild fire suppression
Dead trees also spur wildfires
Associated with warming winters (global warming?)
Timber companies may benefit, millions of trees to be cut
down
Fire is Our Enemy
Dendroctonus ponderosae
Loss Distributions and Fat Tails
1.
2.
3.
4.
5.
6.
7.
8.
For the distributions of some individual outcomes, e.g., human heights, variability is
small relative to the mean.
Other outcomes are aggregates, determined by large numbers of trials with low
probabilities. These will also have small variability relative to the mean.
These distributions are the norm we carry around in our head. They are
fundamentally misleading when we think about low probability catastrophes.
Catastrophes are low probability outcomes. Their magnitudes, e.g., acres lost to the
largest wildfire, people lost in a terrorist attack, damage due to a flooding incident,
are highly variable relative to their mean.
When individual outcomes have high variability relative to their mean, there is rarely
symmetry in distribution of the magnitude of losses.
Rather, there is likely to be symmetry in the logarithm of the magnitude of loss.
Thus, an outcome of roughly twice the median is as likely as one half the median.
We should think of these outcomes in terms of multiples. You think the median loss
is M. What multiple of the mean K would you require to have a ½ chance of
observing the outcome? That is, for what K do you believe the outcome would be as
likely as not to be between M/K and KM?
Even when we think in terms of multiples, our thinking is likely to be influenced
excessively by the Normal (bell-shaped) distribution. The tails of real world
distributions tend to be thicker than that.
This is bad news, if true. The most people lost in a terrorist accident in the US is less
than 3,000. But there is a good chance that 100,000 could be lost.
People Estimate Distributions Too Tightly
Real-Time Estimation Exercise
Your job is to assess:
First, your median estimate (50th percentile). This means the
outcome is as likely to be above it as below it.
Second, your 25th and 75th percentiles. The 25th percentile means
the outcome has one chance in four of being below it. In other
words, if it is below your median, it is as likely as not to be
below your 25th percentile.
Third, your 1st and 99th percentiles, what are sometimes called
surprise points. There should only be a 1% chance that the true
value lies below your 1st percentile. There should only be a 1%
chance that the true value lies above your 99th percentile.
YOUR ESTIMATION QUESTION
Only 7% of land in California is in the wild-urban interface
where the risk of fire damage is highest. How many homes
are located in that 7% of California land?
(Roger Kennedy, Wildfire and Americans: How to Save Lives, Property,
and Your Tax Dollars, 2006.)
99th percentile __________
75th percentile __________
50th percentile __________
(median)
25th percentile __________
1st percentile __________
In Conclusion
While we can never contract with the
future or accurately predict all of the
consequences of our actions and policies,
policymakers must extend their thinking
about their impacts and the impacts of
private entities beyond the local, the near
term, the likely, and the recently
newsworthy.