Transcript Regional Climate Forecast Insurance

Disaster Risk Financing
in the Agricultural Sector
Dr. Jerry Skees
H.B. Price Professor, University of Kentucky
President, GlobalAgRisk, Inc., USA
WMO Expert Advisory Group on Financial Risk Transfer (EAG-FRT I)
Geneva, Switzerland
13-14 December 2011
GlobalAgRisk, Inc.

Mission
Improve access to market services for the poor through innovative approaches for
transferring natural disaster risk
 Activities
 Research and development
 Technical capacity building
 Educational outreach
 Supported by
 Multinational donors
 Governments
 Nongovernment organizations
Bill and Melinda Gates Foundation, Ford
Foundation, GIZ, UNDP World Bank, etc.
 Select Country Work
 Peru – El Niño/Flood
 Mongolia – Livestock
 Vietnam – Flood/Drought
 Mali – Drought
 Morocco – Drought
 Mexico – Drought
 Romania – Drought
 Ethiopia – Drought
 Indonesia -- Earthquakes
Weather Index Insurance
 Conceptually similar to weather derivatives.
 Contingent claims based on an underlying weather index.
 Important consumer protection issues: needs to be insurance
 Sold as insurance products.
 Insurance regulation requires that buyers must have an
“insurable interest.”
 Works only in areas that are highly exposed to clearly
identifiable and measurable spatially-covariate weather
perils.
 Two general categories:
 Household products.
 “Risk aggregator” products. Risk aggregators are lenders, input
suppliers, output processors, transporters, etc. whose
financial exposure to catastrophic weather events spans a
larger region.
3
Weather Index Insurance: Challenges
 Basis risk - a mismatch between payment and loss
 Reduces value as a guarantee (loan; a form of collateral)
 High initial and ongoing weather data demands
 High startup costs:
 Pilot projects funded by donors in lower income
countries.
 Tailored to specific perils in a specific place
 Many pilots: Are they Scalable? Sustainable?
4
Two Sources of Basis Risk
 For any given magnitude of the underlying index (e.g.,
rainfall measured at a particular weather station)
there is a conditional probability distribution for the
policyholder’s exposure to the same variable (e.g.,
rainfall measured at the policyholder’s location).
Index
Cause of Loss
 For any given magnitude of the underlying weather
variable measured at the policyholder’s location,
there is a conditional probability distribution of loss.
Cause of Loss
Loss
5
Challenges:
Measuring Basis Risk
 In empirical applications, basis risk is typically
measured as the covariance (or correlation) between
the index and realized losses using all available data.
 Is this really the correct measure?
 We hypothesize that:
 The correlation between the weather event and losses is
greater the more extreme the weather event.
 The spatial correlation of the weather variable increases.
6
Challenges:
Indexes and Individual Losses
 Very unlikely that data will be available to
quantitatively determine risk exposure and
relationships between potential indexes and realized
losses.
 Must rely on available scientific understanding of these
factors; and
 Qualitative data collected from local sources using
carefully structured interviews or focus groups.
 While time consuming, this process is critical for
product design.
7
Challenges:
Pricing and Payment Triggers
 Sufficient quantitative data of an appropriate spatial
specificity are required to price the insurance.
 Is the target market a household or a risk aggregator?
 How much is sufficient? It depends on the temporal
presentation of the risk.
 Is the probability distribution stationary?
 Is the probability distribution homoskedastic?
 Do either of the above exhibit multi-year cycles?
 Understanding the tail of the distribution is critical.
8
Sahelian Rainfall
9
Sahelian Rainfall
10
Challenges:
Data Sources
 Hydro-Meteorological stations
 Alternative Data Sources
 Spatial interpolation of available weather station data.
 Satellite-based Normalized Difference Vegetation Index
(NDVI).
 Satellite-based measures of rainfall.
 Satellite-based synthetic aperture radar (SAR) maps
contours of geospatial environments (e.g., flooding).
 Reanalysis data.
 A class of data products that combines and calibrates
observations from many sources — weather stations, satellites,
weather balloons, etc.
 Plant growth simulation data
Challenges:
Weather Stations
 Density of weather stations is very sparse in many
lower income countries – especially in Africa.
 Outside of South Africa, the only publicly available daily
weather station data for many African countries is from
major airports.
 More stations exist that report less frequently but
density is still sparse – especially in rural areas.
12
Challenges:
Weather Stations
 Minimum cost of an automated rain gauge with data
logger and remote access capability – approximately
US$ 2,000.
 Does not include cost of shipping, installation, power
source, remote access mechanism (e.g., mobile phone
or satellite connection), or any security measures
(e.g., fencing).
 Also not included are routine maintenance costs
which can be quite prohibitive. Mali had 85 different
weather stations in operation between 1951 and
2007. Ten are currently in operation.
13
Challenges:
Alternative Data Sources
 Satellite-based measures of weather variables still




have significant errors when compared to groundbased measures.
Many alternative sources tend to understate outliers.
Insufficient spatial and/or temporal specificity
(possible exception is NDVI).
Limited time series of data and challenges with
calibrating observations across evolving technologies.
Will potential buyers purchase an insurance product
based on satellite data?
14
Future of Alternative Data Sources
 Satellite-based data sources are rapidly improving.
 They are currently used by reinsurers (often in the
form of reanalysis data) to supplement short time
series of weather station data or to cross-check
questionable weather station data.
 Commercial firms are already investing resources in
developing and improving these alternative data
sources — much as the catastrophe bond market
stimulated the development of private-sector firms
that provide earthquake and hurricane modeling
services.
15
Two Implications
 Widespread scale-up will likely require use of
alternative data systems.
 Initial focus should be on products that require less
spatially specific data – risk aggregator products
rather than household products.
16
Current State of Weather Index Insurance
Most weather index insurance pilots have been designed
to protect against reduced yields for a particular crop.
Increasingly, weather index insurance pilots have been
designed to cover more moderate (rather than
catastrophic) losses.
A.
B.


Insurance bundled with loans.
Concern about long-term demand if buyers don’t receive an
indemnity.
Focus is on household products rather than risk
aggregator products.
D. Limited contribution to climate resiliency and
adaptation.
C.

Regional climate forecast insurance
17
(A) Index Insurance is for Consequential Losses:
Livelihoods
 But extreme weather events have impacts that extend
far beyond yield losses for a single crop.
 Increased irrigation cost or quality losses.
 Increased disease and pest pressure.
 Impacts on non-agricultural enterprises.
 Loss of off-farm income opportunities.
 Higher prices for food and other necessities.
 Assets destroyed or liquidated.
 Need to change the focus from yield losses for a single
crop to the multiple consequential losses caused by
extreme events.
18
(A) Index Insurance is for Consequential Losses:
Data Constraints
 Focusing on consequential losses reduces quantitative
data requirements.
 Quantitative measures of the in-sample correlation
between the index and yield of a specific crop are less
important.
 Qualitative data on consequential losses caused by
extreme events become more important.
19
(A) Index Insurance is for Consequential Losses:
Basis Risk
 Basis risk for consequential losses is likely less than
basis risk for yield losses on a single crop.
 For example, Berg and Schmitz (2008) demonstrate that
weather index insurance for a specific crop is a less
effective risk management tool for households with a
diversified portfolio.
 Many smallholder households in lower income countries
have diversified portfolios of both agricultural and
nonagricultural enterprises.
20
(B) Index Insurance is for Catastrophic Losses:
Cost
 Moderate loss insurance is prohibitively costly.
 The most efficient use of insurance is to protect against
extreme catastrophic events which can threaten longterm wealth positions. When, instead, weather index
insurance is designed to protect against more moderate
losses, it raises the price of insurance compared to a
catastrophic policy. As a result buyers purchase less sum
insured and are less well protected when a catastrophe
occurs.
 Savings and borrowing are more economically
efficient mechanism for transferring moderate losses.
21
(B) Index Insurance is for Catastrophic Losses:
Basis Risk
 Spatial correlation of the weather variable may
increase with the severity of the event.
 Evidence: spatial correlation of June Iowa county
rainfalls is higher in the years of drought than in
years of normal or above normal rainfall
(Miranda and Liu, 2010)
 These results support our conclusions that
writing index insurance for catastrophic events
would likely reduce the basis risk problem
(GlobalAgRisk, 2010b).
22
(B) Index Insurance is for Catastrophic Losses:
Demand
 What about buyers losing interest if they don’t receive
frequent indemnities?
 This concern is certainly supported by some psychology
of risk literature.
 People do purchase some forms of catastrophic
insurance – life insurance and accident insurance are the
fastest growing forms of microinsurance.
 Framing matters. Index insurance should be framed as
protecting long-term wealth positions from the multiple
consequential losses of extreme weather events.
23
(C) Risk Aggregator Products have Fewer Data
Constraints
 For the time being, risk aggregator products may be
the only feasible means of extending weather index
insurance products into many regions of the world.
 In many areas, weather station density is not sufficient
to support household products.
 Risk aggregator products require less spatially specific
data so alternative data sources are more likely to be
feasible for risk aggregator products than for household
products.
 Risk aggregators are more likely to understand hedging
and basis risk (lower capacity building needs).
24
(D) Linking Insurance and Risk Adaptation
 Encourage risk management and appropriate adaptation.
 Smooth cash flow between disaster and non-disaster
years.
 Targeted, early payments.
 Insurance payouts can be used to finance adaptation
investments (e.g., infrastructure, livelihoods transitions).
Insurance is only one tool to address climate change
 Insurance can protect against weather extremes, but adaptation is
necessary to adjust to changing climate trends.
Challenge: Which Insurance products can provide the greatest opportunity for
adjusting to changing climate?
(D) Weather vs. Climate
 “The difference between weather and climate is a
measure of time. Weather is what conditions of the
atmosphere are over a short period of time, and climate
is how the atmosphere "behaves" over relatively long
periods of time.
 In addition to long-term climate change, there are shorter
term climate variations. This so-called climate variability
can be represented by periodic or intermittent changes
related to El Niño, La Niña, volcanic eruptions, …”
Source: NOAA Definitions
(D) Regional Climate Change
 At least three drivers to consider:
 Teleconnections (e.g., El Nino Southern Oscillation, N.
Atlantic Oscillations, India Ocean Oscillations, Arctic
Oscillations).
 Regional land use practices (1930s dust bowl in the U.S.;
prolonged drought across the Sahel in the 1960s to 1980s;
Northwest China; other locations?).
 Build up of green house gases (longer term).
Sahel data 1900 – 2007
Courtesy of IRI at
Columbia University
800
700
600
500
Sahel
Semi-arid region below the Sahara400
300
200
100
0
1900
1903
1906
1909
1912
1915
1918
1921
1924
1927
1930
1933
1936
1939
1942
1945
1948
1951
1954
1957
1960
1963
1966
1969
1972
1975
1978
1981
1984
1987
1990
1993
1996
1999
2002
2005
• Dynamic climate largely due
to oceanic oscillations
• Unlike the Sahel, climate
change may lead to more
permanent changes
28
(D) Climate Change Insurance Products?
 Generally weather insurance products involve localized
conditions and are for only one year.
 Modeling climate drivers (e.g. teleconnections) to create
‘forecast insurance’ may be the first regional climate
insurance.
 At the current time, regional climate forecast insurance
for 2 to 3 years may be as close as we can get to creating
‘climate change’ insurance.
(D) Regional Climate Forecast Insurance
 Regional climate insurance can be used to help build
resiliency.
 Most emerging economies have no regional climate
insurance products.
 Due to an improved understanding of how
teleconnections drive regional climate events, the science
of forecasting extreme regional climate events is
improving.
 GlobalAgRisk has developed the financial architecture in
Peru for regional Climate Forecast Insurance – Extreme El
Nino Insurance Product.
Teleconnection:
El Niño and
Sothern
Oscilation
Source:
http://www.grida.no/pub
31
lications/vg/africa/page/
3105.aspx
Piura and Other Areas in Northern Peru
Severely Affected by 1998 El Niño
 Extreme rains (Jan – Apr 1998)
 40x normal rainfall
 Severe floods
 41x normal river volume
 Widespread losses
 Many disrupted markets
 Agricultural production, ↓ 1/3
 Public infrastructure losses
 Cash-flow, debt-repayment problems
 Health problems
 Total losses in Piura estimated at
USD 200 Million
32
Contract Is Written Using NOAA Data
 El Niño estimates derived from —
33
Satellite data, observations of buoys, and readings of the temperature
on the surface and at deeper levels
 Data are publicly available monthly from NOAA
(The U.S. National Oceanic and Atmospheric Administration)
http://www.cdc.noaa.gov/Correlation/nina1.data
Strong El Niño in 1982–83 and 1997–98
27
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
1800
Precipitaciones Aeropuerta Piura
2000
2004
2003
2002
2001
2000
1999
1998
1997
1996
1995
1994
1993
1992
1991
1990
1989
1988
1987
1986
1985
1984
1983
1982
1981
1980
1979
34
21
1600
Two extreme
events in the last
32 years
ENSO 1.2 Nov and Dec
26
25
24
23
22
1400
1200
1000
800
600
400
200
0
DESTRUCCION DE PUENTES
PUENTE SAN MIGUEL Y BOLOGNESI (PIURA)
The Nino Index is Negatively Correlated with Intensity of
Atlantic storms and Hurricanes
Copyright by GlobalAgRisk, Inc. www.globalagrisk.com
28
27
Correlation from 1950 to 2010 = -0.36
Correlation from 1979 to 2010 = -0.47
1997 event
Nino 1.2 (Nov-Dec)
26
1982 event
25
1972 event
24
23
22
21
20
0
50
100
150
200
Accumulated Cyclone Energy (ACE)
NOAA develops the ACE Index to reflect the "total seasonal activity" whiche measures the collective intensity
and duration of Atlantic named storms and hurricanes occurring during a given season. The ACE index is a
wind energy index, defined as the sum of the squares of the maximum sustained surface wind speed (knots)
250
Example of a Payout from the 1997 Event
Niño Region 1.2 (Nov–Dec) Temperature = 26.28°C
Minimum Payment = 5%
The insured selects the Sum Insured
Sum Insured = 10,000,000 Soles
1998 Payment = 76% x 10,000,000 = 760,000 Soles
37
Primary Goal
Improve Access to and Terms of Loans
Capacity building with
 Financial institutions
 Peruvian banking regulator
 Peruvian credit rating agencies
 Sources of social capital flows into Peruvian institutions
Case to be made
1) Strengthen resiliency of the financial institution
2) Financial institution can be ready to lend when the community
needs capital the most — Ex post — After the disaster
38
(D) Teleconnections and Regional Climate Forecasts
 ENSO effects in
atmospheric conditions
is significant– its impact
is not anecdotal but
physically established
 January-March weather
anomalies and
atmospheric circulation
during El Nino and La
Nina phases
Teleconnections and Regional Climate Forecasts
 Precipitation and
Temperature
Anomalies
associated with El
Nino
(D) Teleconnections and Regional Climate Forecasts
 Precipitation and
Temperature
Anomalies
associated with La
Nina
(D) Teleconnections and Regional Climate Forecasts
 What about IOD, NAO,
PDO…etc?
 What is their
signature?
 What about the
combination of these
indices and the
feedback
mechanisms?
 ENSO and IOD have
positive feedback
over Eastern Africa.
• IOD – Indian Ocean Dipole
• NAO – North Atlantic Oscillation
• PDO – Pacific Decadal Oscillation
(D) How can Regional Climate Forecast
Insurance be Climate Change Insurance?
 Getting cash before the impending disaster can motivate
decision makers to take action to reduce the losses
 Most decision makers don’t use forecast information in
an efficient manner (prospect theory: regret)
 Taking action based on impending disasters will result in
improved adaptation that builds resiliency
 As climate change occurs, if the frequency and the
severity of the forecasted events increases, the rising
price of the insurance will increase the dynamics of these
adaptive management practices over time and should
result in stronger systems to cope with climate change
Research Questions
 Does the spatial covariance of specific weather events
change depending on the severity of the event?
 Does the covariance between specific weather events
and realized losses change depending on the severity
of the weather event?
 Does the covariance in losses across different
livelihood activities change depending on the severity
of the weather event?
44
Research Questions
 Does index insurance designed to protect against
various consequential losses have lower basis risk
than index insurance designed to protect against yield
losses for a specific crop? (Will likely depend on the
weather peril and local context.)
 Will potential buyers purchase index insurance based
on novel (e.g., satellite-based) data sources?
 Will they purchase index insurance that protects
against only catastrophic losses?
45
Research Questions
 In what regions can estimates of oceanic
anomalies such as ENSO be used as
forecasts for regional climate risk transfer?
 ENSO for regions other than Peru?
 Gulf of Guinea SST?
 North Atlantic Oscillation?
46
Conclusion
 Important research questions remain, we believe that:
 Wide spread scale-up will likely require alternative data sources.
 Index insurance should target consequential losses (not just crop





yield shortfalls).
Index insurance should target catastrophic losses.
For the near future, risk aggregator products are likely the only
feasible means of extending weather index insurance products
into many regions of the world.
There is potential to use climate forecast insurance to improve
resiliency and adaptation.
Developing regional climate forecast insurance will take time.
The science of regional climate forecasting is improving and will
enable more widespread application.
47
Sources
 Collier, B., B.J. Barnett, and J.R. Skees. “State of
Knowledge Report – Data” produced by GlobalAgRisk, Inc.
for the Bill and Melinda Gates Foundation. Available soon
at http://www.globalagrisk.com
 Barnett, B.J., C.B. Barrett, and J.R. Skees. 2008. “Poverty
Traps and Index-Based Risk Transfer Products.” World
Development 36:1766-1785.
 Collier, B., J.R. Skees, and B.J. Barnett. 2009. “Weather
Index Insurance and Climate Change: Opportunities and
Challenges in Lower Income Countries.” Geneva Papers
on Risk and Insurance — Issues and Practice 34:401–424.
48