Index Insurance, food security, and poverty

1

Index Insurance, food security, and poverty
Daniel Osgood deo_at_iri.columbia.edu The
International Research Institute for Climate and
Society Columbia University
2
Index Insurance

• Problems with traditional insurance have kept it
  from being available to most of the world
• Traditional Crop insurance
• Undermined by Private Information problems
• Almost always subsidized (SUBSIDIES CAN CAUSE
  PROBLEMS)
• Recent index innovation
• Insure weather index (eg seasonal rainfall), not
  crop
• Only partial protection (basis risk), should not
  oversell
• Cheap, easy to implement, good incentives
• Implementations only a couple of years old
• Exploding popularity--dangerous if misused
• Structure to target each particular goal
• Design complex
• Only a naive partner would reveal all their cards
• All partners must play active role in a
  cooperative design
• Client must know what is not covered

3
Multiple poverty challenges, multiple index
strategies

• Insurance is not for its own sakeit is to reduce
  poverty, improve food security, and encourage
  development
• Implementation strategy driven by context, type
  of risk
• Damage dropping people into poverty traps
• Risk preventing people from moving forward
• Immediate damage
• http//iri.columbia.edu/publications/search.php?id
  556

4
Projects I have experience with

• Participated in design of contracts purchased by
  farmers in Africa and Latin America (micro)
• Worked on micro contracts for
• Malawi, Kenya, Tanzania, Honduras, Nicaragua,
  supported MVP (macro) insurance design, working
  on contracts for Ethiopia
• IRI specific
• Roundtable on Index insurance, poverty, and
  climate risk
• Climate and Society Publication, Vol II
• Work with partners to bring existing and new
  research to solve practical implementation
  problems
• Statistical/stakeholder based design
• Explicitly build climate into product
• Stakeholder communication and contract
  verification
• Forecasts
• Data poor environments
• Limitations of and potential for
• Remote sensing
• Of crops
• Of precipitation

5
Macro Example
MVP Index

• Early warning vs early action?
• IRI projects
• Index product for Earth Institute MVP (led by
  Neil Ward, with Asher Seibert, Eric Holthaus in
  IRI, many other partners)
• Index to ensure development goals of MVP for each
  village cluster
• If MVP lifts people out of poverty traps
• Prevent climate from them falling back in
• Also exploring Locust, fire, malaria, livestock
  disease and international trade, forage, water
  management

6
Drought Index Insurance A contribution to
managing the climate risk
Millennium Villages and Index Insurance
Including C. Palm, D. Osgood, A. Siebert, E.
Holthaus, J. Anttila-Hughes, J. Puri, W.
Baethgen, partnership with Swiss Re, and
partnerships with Meteorological Services in
Africa (processing station rainfall data) and
satellite data sources (NASA and NOAA)
7
Variables that can be a proxy for impact on rural
population
Satellite Regional NDVI
Ground-based Local Rainfall
Rainfall
8
Seasonal rainfall total is not the best indicator
for crop yield Alternative is to use a simple
crop model, e.g. Water Requirement Satisfaction
Index (WRSI)
Water requirement varies through crop growth cycle
Eg for 180-day maize (as used for Sauri)
9
1984
2000
10
Micro Example

• Malawi Groundnut
• Farmer gets loan (4500 Malawi Kwacha or 35)
• Groundnut seed cost (25, ICRSAT bred, delivered
  by farm association)
• Interest (7), Insurance premium (2), Tax
  (0.50)
• Prices vary by site
• Farmer holds insurance contract, max pay is
  loansize
• Insurance payouts on rainfall index formula
• Joint liability to farm Clubs of 10 farmers
• Farmers in 20km radius around met station
• At end of season
• Farmer provides yields to farm association
• Proceeds (and insurance) pay off loan
• Remainder retained by farmer
• Farmers pay full financial cost of program (with
  tax)
• Only subsidy is data and contract design
  assistance

http//iri.columbia.edu/deo/IRI-CRMG-Africa-Insur
ance-Report-6-2007/
11
Insurance design methods

• Previous methods were based on crop biology and
  did not have a mechanism for systematic inclusion
  of climate information
• Methodology overview
• Stakeholders determine
• Premium constraint
• Payout frequency target
• Set initial guess for optimizer
• Pursue strategies that target alternate risks (eg
  sowing vs flowering)
• Computer optimization (tuning)
• Using performance measures, WRSI based loss
• Optimize upper triggers to
• Minimize variance of (losses - insurance
  payments)
• Subject to specified maximum insurance price
• Compare contracts performance against conflicting
  information sources looking for contract
  strengths and vulnerabilities
• Adjust parameters to round numbers so that client
  does not get misimpression that design
  information is higher accuracy than it is
• Communicate results with stakeholders
• Correct years for correct reasons
• Is coverage what clients demand?
• Adapt contracts and models

12
(No Transcript)
13
(No Transcript)
14
Ranking of losses and payouts
RANK YEAR LOSS PAYOUT? 1, 1995 7641.140
1 2, 1973 6542.680 1 3, 1966 6324.398
0 4, 1996 6315.617 1 5, 1990
5903.817 0 6, 1984 5660.633 1 7,
2005 5598.026 1 8, 1970 4929.469 1 9,
1992 4904.982 0 10, 1997 4459.438 1 11,
1968 4400.516 0 12, 1969 4296.916 1
13, 1980 4235.219 0 14, 1994 4136.128
0 15, 2004 3921.972 0 16, 1979 3513.749
0 17, 2000 3399.898 0 18, 1983
3399.299 0 19, 2001 3367.294 0 20, 2006
3347.076 0 21, 2002 3218.283 1 22, 1967
3070.731 0 23, 1962 0.000 0 24, 1963
0.000 0 25, 1964 0.000 0

26, 1965 0.000 0 27, 1971 0.000
0 28, 1972 0.000 0 29, 1974 0.000
0 30, 1975 0.000 0 31, 1976 0.000
0 32, 1977 0.000 0 33, 1978 0.000
0 34, 1981 0.000 0 35, 1982 0.000
0 36, 1985 0.000 0 37, 1986 0.000
0 38, 1987 0.000 0 39, 1988 0.000
0 40, 1989 0.000 0 41, 1991 0.000
0 42, 1993 0.000 0 43, 1998 0.000
0 44, 1999 0.000 0 45, 2003 0.000
0
15
Stakeholder input drives contracts

• Look for
• Do stakeholders understand contracts?
• Do stakeholders show evidence of negotiating in
  their own interests?
• Do stakeholders understand basis risk and what is
  not covered?
• Look for insightful complaints
• Malawi stakeholders have been very active, driven
  design
• Original CRMG project proposal was for stand
  alone Maize Insurance
• Malawi stakeholders proposed groundnut bundle

16
Insurance index developed with farmers
Nicole Peterson, NSF-DMUU
17
Insurance and forecasts

• Insurance is exact compliment for forecast
• Community has focused on insurance problems due
  to forecast
• Overlooked fundamental relationship between
  Insurance, forecasts, and investments
• Benefits to building forecast into insurance to
  improve use of information in making investments
• Need to research how to do this
• Exploring for Malawi case
• Insurance can be tied to early warning (forecast)
  systems to finance and formalize response
• Examples WFP and MVP
• Potential
• Fire management, flood response, health response,
  etc.
• With Jim Hansen, Pablo Suarez, Miguel Carriquiri,
  Ashok Mishra, and others

18
Climatology important

• Northern and Southern Malawi
• opposite Enso phase response
• Location of north-south dividing line challenging
  to forecast
• But climate info still very valuable for
  insurance
• Natural hedge
• By strategic pooling of contracts harnessing
  negative correlations of climate
• Total risk can be reduced, reducing costs of
  insurance
• Global insurance pool reduces cost of managing of
  risk everywhere
• Difficult to design based on payout data alone
• Few payouts in historic record
• Dont know if statistical have physical forces
  driving them
• Climate science understanding may guide and
  verify insurance strategies
• Investigating for Central America
• With Megan McLauren, Marta Vicarelli, Alessandra
  Giannini

19
Global implications

• With increasing climate risk need to leverage
  whole world
• Extreme events
• Much of Climate Change?
• Negatively correlated across globe?
• Whole world distributes risk
• Need to develop global risk markets
• Companies want to serve a global insurance market
  that does not yet exist to reduce costs
• Insurance premiums lower
• With global markets incentives for optimal global
  production diversification
• International climate sensitive suppliers

(Based on Ropelewski Halpert, 1987)
20
Learning, insurance and experimental economics

• Can people cooperatively transfer risk through
  insurance negotiations?
• Experimental economics pilot
• Examined alternate educational strategies for
  cooperation
• Found that people may think of risk more
  rationally when negotiating with others
• Do people understand the the index?
• Upcoming basis risk pilot experimental economic
  pilot in Brazil
• With CRED

21
Monte Carlo, insurance, and pricing uncertainty

• Monte Carlo simulations useful in insurance
  design methods
• Apply design optimization to simulated rainfall
• Price using probabilities from monte carlo payout
  distributions
• Model uncertainty can be built into insurance
• Could map uncertainty into insurance product
  using modified simulations
• Estimate uncertainty in parameters
• Draw from distribution of parameters
• Draw from distribution specified by parameters
  drawn
• Build into rainfall simulator for educational
  tool
• Could condition on forecast information
• With Kenny Shirley, Andy Robertson

22
Insurance and paleoclimate data

• Hypothetical Tree ring insurance
• Explore consequences of using recent data when
  longer term data is available
• Impacts of using 50 years of record
• vs full 370-year record
• Preliminary findings
• Size of 100 year disaster overstated.
• Uncertainty in 50-year estimate brackets true
  risk.
• Properly including uncertainty in insurance price
  could lead to sustainable product
• Including information from paleodata may make
  insurance more affordable
• With Art Greene, Lisa Goddard, Kevin Anchukaitis

N 370 50 Lev -0.56 (-0.49)
Lower -0.58 (-0.63) Upper -0.52
(-0.34)
N 370 50 Lev -0.40 (-0.31)
Lower -0.46 (-0.49) Upper -0.30 (-0.13)
23
Index insurance data

• Rainfall data is short, with gaps
• Limited spatial coverage
• How far is too far from station?
• Common to many applications
• Need technique for new stations
• Most places do not have long met station history
• Must address for scale-up

24
Micro level projects at scale

• Develop techniques to utilize remote sensing in
  product design, pricing, maintaining
  uncertainties
• Treatment of spatial basis risk
• Increased focus on farmer education, involvement
  in monitoring and design
• With Paul Block, Tufa Dinku
• Must have design methodologies with local
  expertise so that contracts and packages can
  adapt every year
• Eg of these issues Ethiopia micro project with
  Oxfam
• Compliments WFP and Ethiopian Government national
  index insurance project
• But need to stop using pilot scale temporary
  solutions solutions for scale up
• If we want industry we need industrial strength
  solutions and industrial strength processes for
  technique development

25
Training tools (under development with WB-CRMG)
Megan McLauren and Lulin Song
26
Design training tools (under development)
27
Design training tools (under development)
28
Design training tools (under development)
29
Thank you