index insurance: structure, models, and data daniel osgood (iri) [email protected] material...

20
Index insurance: structure, models, and data Daniel Osgood (IRI) [email protected] Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole Peterson, Megan McLaurin The International Research Institute for Climate and Society

Upload: earl-blair

Post on 12-Jan-2016

214 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Index insurance: structure, models, and dataIndex insurance: structure, models, and data

Daniel Osgood (IRI)[email protected]

Material contributed by:

Miguel Carriquiry, Ashok Mishra, Nicole Peterson, Megan McLaurin

The International Research Institutefor Climate and Society

Page 2: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Examples from groundnut in Malawi

Page 3: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Contract Structure

• Rainfall summed over 10 day periods (dekads)

• Dekadal maximum ‘cap’

• Sowing rainfall condition – Starts contract clock

– Or triggers ‘failed sowing’ payout

• Season split into phases

• Payouts each phase – From capped dekadal rainfall total over phase

Page 4: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Phase sum payout function

Payout = (1 – (Rainfall Sum – Exit) / (Upper trigger – Exit)) Max Payout

Phase Payout function 2006

0

25

50

75

100

125

150

175

200

225

250

275

300

0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000

Kwacha

Mv

ula

Ra

infa

ll (

mm

)

Page 5: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Nicole Peterson, CRED

Insurance Contract developed with Farmers

Page 6: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Contract parameters

• Sowing– Sowing window beginning, end

– Sowing trigger

– Failed sowing payout

• Phases– Number of phases

– Beginning, end of each phase

– Upper trigger, exit

– Maximum payout per phase

• Maximum total payout

Page 7: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

New obligations with index products

• Traditional insurance--Triggered on loss– Pricing and financing on losses

– If payments not closely linked to losses• Provider and client both face consequences

– Adjuster is responsible for agreement

– Insurance providers experienced assessing losses

• Index insurance--Triggered on index– Insurer pricing and financing built on index

– If there is an error linking payments to losses• Only client faces consequences

– Contract must emulate adjustor

– Much more client interaction

Page 8: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Crops and Climate

• Crop models – Summarize the biological drought vulnerability

of crops during a season

• Well selected crop– Adapted for little vulnerability during the dry

spells in local climate

• Drought stress:– Combination of biology and local climate

Insurance contracts must address this balance

Page 9: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Financial features of insurance

• Deductible, payout frequency: – Insurance only protects against the largest losses– Insurance pays out rarely

• Insurance must target losses that are important in client’s risk management

– Client may prefer protection against 100 year loss, or 5 year loss

– Client may prefer protection against late season losses because sowing problems might be better addressed through practice changes

• Price constraints– Insurance must be affordable– Risk coverage must be most cost effective option

These features must be addressed in design

Page 10: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Water Stress Information

• Multiple information sources– WRSI

– Process based crop models (eg DSSAT)

– Historical regional yield

– Farmer and expert feedback

– Field trials

Each has strengths, limitations for design

Page 11: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

WRSI

• Powerful tool for ‘water stress accounting’ – Well known– Assumptions intuitive – Results are accounting of

• Rainfall• With storage, loss assumptions

• Not best for direct yield simulation– Its developers at FAO use related statistical

techniques instead of model outputs for yields

• In contract design useful – Weigh relative water stresses due to crop genetics

and climate– Platform for communication of crop features in design– Starting point for contract parameters– Statistically link local climate to crop vulnerabilities

Page 12: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

WRSI Issues

• Key parameter assumptions– Timing of growth stages is assumed– Relative vulnerability over season is assumed

• Limited capabilities—`Simple but honest’– Often inaccurate for small losses– Not accurate quantification of

• Risk faced by individual farmer

• Yield losses

– Excess water impacts not modeled– Crop failure is not modeled

Targets limited coverage to most important risk

Must verify using additional sources of info

Page 13: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Stress models

0.0

200.0

400.0

600.0

800.0

1000.0

1200.0

0

1000

2000

3000

4000

5000

6000

7000

Daily Yield Ave Ky

Daily Yield Ky(t)

DSSAT Yield

What is ‘truth’?

Page 14: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

DSSAT and WRSI Simulated Yields and Historical Yields for Chitedze Groundnut Crop

100

300

500

700

900

1100

19

62

19

64

19

66

19

68

19

70

19

72

19

74

19

76

19

78

19

80

19

82

19

84

19

86

19

88

19

90

19

92

19

94

19

96

19

98

20

00

20

02

20

04

20

06

Year

Cro

p Y

ield

(k

g/h

a)

0.60

0.80

1.00

WR

SI c

rop

yie

ld (

at

1 s

ca

le)

DSSAT Crop Yield Historical yield WRSI

  Hist. Yields

  DSSAT 0.30

WRSI 0.35 0.52

Correlations

WRSI, DSSAT, Historical Yields

Page 15: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Insurance targets covariate risk

EPA Historical Groundnut Yields

0

200

400

600

800

1000

1200

1400

1600

Year

Yie

ld (k

g/ha

)

CHILAZA

DEMELA

KAMBANIZITHE

MING'ONGO

MLOMBA

M'NGWANGWA

MPINGU

NTHONDO

SINYALA

UKWE

CHILAZA DEMELA KAMBANIZITHE MING'ONGO MLOMBA M'NGWANGWA MPINGU NTHONDO SINYALA UKWE

0.78 0.92 0.78 0.69 -0.52 0.69 0.74 0.89 0.81 0.89

Correlations with average yield:

Note: ~2-3 worst years most important for insurance

Page 16: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Questions for farmers and experts

• What are the best years and the worst drought years that you can remember?

– In which years did you have yield problems because of drought, and for each year, what was the reason for the problem

(eg dry sowing/weak start of rains or drought during the filling phase)?

• When do you typically plant?– When is the earliest that you have planted?– When is the latest that you have planted?

• What do you do if rains are insufficient for planting?

• For what growth phases is rainfall most important? – In what months?

• Do the historical payouts from this contract – Match the years you had reduced yields from drought?– Connect to the growth stage that your crops were in when

they were impacted?

Page 17: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole
Page 18: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole
Page 19: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Use of Water Stress Information Sources

• WRSI– Somewhat insensitive, direct product of assumptions– Good benchmark– Use as an accounting system for relative water stress, not a direct

simulation of yields

• Process based crop models (eg DSSAT)– Must be carefully calibrated– Data intensive– Representative of very specific situation– Good for identifying and understanding for losses missed by WRSI

• Historical regional yield– Not only water stress – Often low quality– Short time series– Different varieties, practices– Use to see if important historical losses are covered

• Field trials– Artificial production situation, very limited availability– Detailed and reliable specifics of crop/climate interaction

• Farmer and expert feedback– Qualitative, strategic– Use to tune and verify WRSI and model timing, gauge how well

coverage addresses important years for correct reasons – But remember it may be strategic, unreliable

Page 20: Index insurance: structure, models, and data Daniel Osgood (IRI) deo@iri.columbia.edu Material contributed by: Miguel Carriquiry, Ashok Mishra, Nicole

Contract design?

• Different data sources--different information

• Because of moral hazard in traditional insurance:– Only naïve players show all of their cards

– We can only approximate client• risk preferences, productivity, self-insurance,

production details, microclimate, practices, consumption needs, hedging strategies, other sources of income, etc…

– Design is negotiation process

• Iterative statistical system for design• Strategic use of information