Practical Procedures For Evaluating Crop Insurance Policies That Trigger On County Yield Ben Chaffin, Graduate Research Assistant J. Roy Black, Professor.

Practical Procedures For Practical Procedures For Evaluating Crop Insurance Policies Evaluating Crop Insurance Policies

That Trigger On County YieldThat Trigger On County Yield

Ben Chaffin, Graduate Research AssistantBen Chaffin, Graduate Research AssistantJ. Roy Black, ProfessorJ. Roy Black, Professor

Department of Agricultural EconomicsDepartment of Agricultural EconomicsXiaobin Cao, Graduate Research Assistant, Xiaobin Cao, Graduate Research Assistant,

Agricultural Economics/StatisticsAgricultural Economics/StatisticsMichigan State UniversityMichigan State University

IntroductionIntroduction

Personal historyPersonal history– Experience with crop insuranceExperience with crop insurance

county and unit trigger policiescounty and unit trigger policies– Raise sugar beets, corn, soybeans, wheat, & Raise sugar beets, corn, soybeans, wheat, &

cucumbers.cucumbers.

Yield is a 1Yield is a 1stst order proxy for revenue order proxy for revenueIf a county yield trigger policy does not transfer If a county yield trigger policy does not transfer risk, then a county revenue policy will not eitherrisk, then a county revenue policy will not eitherCorn yields and revenues used through the Corn yields and revenues used through the presentationpresentationFeel free to ask questions about clarification Feel free to ask questions about clarification anytimeanytime

Research ObjectivesResearch Objectives

To increase knowledge of county trigger To increase knowledge of county trigger crop insurance policiescrop insurance policies

To provide better tools to support farmers To provide better tools to support farmers in insurance purchase decisionin insurance purchase decision

Target AudienceTarget Audience

FarmersFarmers

Lenders Lenders

Insurance agentsInsurance agents

Extension staffExtension staff

Trade association representativesTrade association representatives

Crop Insurance Policy OverviewCrop Insurance Policy Overview

No InsuranceNo InsuranceGRP (Group Risk Policy)GRP (Group Risk Policy)– County yield insuranceCounty yield insurance

Triggers on county yield indexTriggers on county yield indexCoverage levels 70% - 90%, in 5% incrementsCoverage levels 70% - 90%, in 5% increments

APH (Actual Production History)APH (Actual Production History)– Insurance triggering on farm yieldInsurance triggering on farm yield

Triggers on actual production history Triggers on actual production history – Units and optional unitsUnits and optional units

Coverage levels 50% - 85%, in 5% incrementsCoverage levels 50% - 85%, in 5% increments

Case Study ResultsCase Study Results

Some farms preferred GRP to APHSome farms preferred GRP to APH

Some farms preferred GRP to no Some farms preferred GRP to no insuranceinsurance

Some farms preferred no insurance to Some farms preferred no insurance to APHAPH

Research Approaches UsedResearch Approaches Used

MeasuresMeasures– Net worthNet worth– Net returns or net cash flowNet returns or net cash flow

Evaluation criteriaEvaluation criteria– Mean - Variance Mean - Variance – Expected utility / willingness to payExpected utility / willingness to pay

Measures are adequate, but are not Measures are adequate, but are not farmer friendlyfarmer friendly

Outreach ApproachesOutreach Approaches

MeasuresMeasures– Net worthNet worth– Net returns or net cash flow Net returns or net cash flow

Evaluation criteriaEvaluation criteria– Scenario analysisScenario analysis

Assume perfect correlation between farm and Assume perfect correlation between farm and county yieldcounty yield

Farmer friendly, but criteria does not go far Farmer friendly, but criteria does not go far enoughenough

Another Step: Another Step: Build on previous approachesBuild on previous approaches

Scenario analysisScenario analysis Show tracking between farm yield and county Show tracking between farm yield and county yield yield

Outreach publications and presentationsOutreach publications and presentationsInsurance agency and extension services softwareInsurance agency and extension services software

Use of cumulative probability distributionsUse of cumulative probability distributions– ARMS software to evaluate crop insurance, pre-ARMS software to evaluate crop insurance, pre-

harvest pricing, and enterprise portfolio (King, Black: harvest pricing, and enterprise portfolio (King, Black: 1987) and applications (ND 1993)1987) and applications (ND 1993)

– Price probability distribution forecast WWW (Hilker, Price probability distribution forecast WWW (Hilker, 1996)1996)

Producers have seen this technique beforeProducers have seen this technique before

Case Studies are used to Display Tracking Case Studies are used to Display Tracking Case Farm #1Case Farm #1

40.0

60.0

80.0

100.0

120.0

140.0

160.0

180.0

1980 1985 1990 1995 2000 2005

Year

Yield

County YieldFarm #1 Yield

Cumulative Probability DistributionCumulative Probability Distribution

0 20 40 60 80 100 120 140 160 180 2000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Yield(bu/acre)

Cum

ulativ

e Pr

obab

ility

Yield 1 Unit

Where to?Where to?

Should a county index trigger product be Should a county index trigger product be considered?considered?– Will it transfer significant downside risk?Will it transfer significant downside risk?

How do we get our arms around this How do we get our arms around this decision process?decision process?– How good is good enough?How good is good enough?– Factors influencing performance Factors influencing performance

County index products vs. no crop insuranceCounty index products vs. no crop insuranceCounty index products vs. unit trigger productsCounty index products vs. unit trigger products

Factors Influencing TrackingFactors Influencing Tracking

Homogeneity of farm’s: soils, drainage, Homogeneity of farm’s: soils, drainage, irrigation, and micro climates when irrigation, and micro climates when compared to the countycompared to the county

Farm location within the countyFarm location within the county– Spread across countySpread across county– Center of countyCenter of county– Edge of countyEdge of county– Corner of countyCorner of county

Standardized Mean Standardized Mean

Farm #1 and County YieldFarm #1 and County Yield 1984 to 2003 1984 to 2003 Correlation 0.Correlation 0.85, 1994 to 2003 Correlation 0.96 85, 1994 to 2003 Correlation 0.96

40.0

60.0

80.0

100.0

120.0

140.0

160.0

180.0

1980 1985 1990 1995 2000 2005

Year

Yield

County Yield

Farm #1 Standardized

Yield Insurance Comparison Yield Insurance Comparison Farm Yield + Insurance IndemnityFarm Yield + Insurance Indemnity

Farm #1Farm #1

40.0

60.0

80.0

100.0

120.0

140.0

160.0

180.0

1990 1995 2000 2005

Year

Yield

No InsAPH InsGRP Ins

Standardized Mean Standardized Mean

Farm #3 and County YieldFarm #3 and County Yield 1994 to 2003 Correlation 0.781994 to 2003 Correlation 0.78

40.0

60.0

80.0

100.0

120.0

140.0

160.0

180.0

1990 1995 2000 2005

Year

Yield

County

Farm #3 Standardized

Farm #3, Farm #3, Farm Yield vs. Farm Yield + GRP IndemnitiesFarm Yield vs. Farm Yield + GRP Indemnities

40.0

60.0

80.0

100.0

120.0

140.0

160.0

180.0

1990 1995 2000 2005

Year

Yield

No Ins

GRP Ins

Farm With High Correlation Farm With High Correlation Farm to County Yield Farm to County Yield ≈ ≈ 0.950.95

0 50 100 150 200 2500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Yield(bu/acre)

Cum

ula

tive P

robabili

ty

Farm Yield + APH

Farm Yield No InsFarm Yield + GRP

Farm With Low CorrelationFarm With Low Correlation Farm to County Yield Farm to County Yield ≈ ≈ 0.800.80

0 20 40 60 80 100 120 140 160 180 2000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Yield(bu/acre)

Cum

ula

tive P

robabili

ty

Farm Yield + APH

Farm Yield No InsFarm Yield + GRP

Simulation SetupSimulation Setup

Hypothetical countyHypothetical county– 16 locations16 locations

Pairwise yield correlations across Pairwise yield correlations across locationslocationsYield CDFYield CDF– Shape based upon county and farm dataShape based upon county and farm data– Calibrate to farm insurance unit mean & standard Calibrate to farm insurance unit mean & standard

deviationdeviation

Hypothetical farms Hypothetical farms – The 16 locations that make up the county are treated The 16 locations that make up the county are treated

as individual insurance unitsas individual insurance units..

Hypothetical CountyHypothetical County

NASS sampling error added to county yield NASS sampling error added to county yield

11 22 33 44

55 66 77 88

99 1010 1111 1212

1313 1414 1515 1616

Location CorrelationsLocation Correlations

Observed unit correlations from case studyObserved unit correlations from case study– Actual farm unit correlationsActual farm unit correlations

High 0.96High 0.96

Low 0.46Low 0.46

Simulation used distance to determine Simulation used distance to determine correlationcorrelation– High 0.85 (Nearest neighbor)High 0.85 (Nearest neighbor)– Low 0.50 (Corner to corner: 1 to 16)Low 0.50 (Corner to corner: 1 to 16)

Unit Yield Probability DistributionUnit Yield Probability Distribution

The yield CDF is a composite of 3 countiesThe yield CDF is a composite of 3 counties– Yield data is from 1970 – 2003Yield data is from 1970 – 2003

Mean yield used was: Mean yield used was: – 140140

Standard deviation used was:Standard deviation used was:– 4040

Mean and standard deviation are representative Mean and standard deviation are representative of the county where the case farms are. of the county where the case farms are.

Generation of Yield Guaranties and Generation of Yield Guaranties and Indemnities Indemnities

Model draws 31 outcomes for each of the 16 Model draws 31 outcomes for each of the 16 unitsunitsThe first 30 outcomes are used to calculate ECYThe first 30 outcomes are used to calculate ECYDraws 21 – 30 calculate the expected yield of Draws 21 – 30 calculate the expected yield of each APH uniteach APH unitDraw 31 determines if there is an insurance Draw 31 determines if there is an insurance payment for county and unit insurance policies.payment for county and unit insurance policies.Working model typically takes 10,000 sample Working model typically takes 10,000 sample draws.draws.

Farm Location with Farm Location with ≈0.95 ≈0.95 correlationcorrelation

Farm included locations Farm included locations – 6,7,10, and 11 with equal weight6,7,10, and 11 with equal weight

11 22 33 44

55 66 77 88

99 1010 1111 1212

1313 1414 1515 1616

County Trigger Yield Insurance County Trigger Yield Insurance “GRP”“GRP”

Used 90% coverageUsed 90% coverage

Used maximum protection 100%Used maximum protection 100%– Scale 1.5Scale 1.5

Farms are made up of 1 to 16 of the unitsFarms are made up of 1 to 16 of the units– If county insurance pays an indemnity add it If county insurance pays an indemnity add it

the average farm yield. the average farm yield.

Unit Yield Insurance “APH”Unit Yield Insurance “APH”

Used 75% coverageUsed 75% coverage

Each unit in the county is treated as an Each unit in the county is treated as an insurance unitinsurance unit– Optional unit approachOptional unit approach

Farms are made up of 1 to 16 of the unitsFarms are made up of 1 to 16 of the units

Calculating Insurance PremiumsCalculating Insurance Premiums

County premium chargedCounty premium charged– Averaged the indemnities paid (pure Averaged the indemnities paid (pure

premium)premium)– Multiplied pure premium by (1 – subsidy)Multiplied pure premium by (1 – subsidy)

Unit premium chargedUnit premium charged– Average the indemnities paid (pure premium)Average the indemnities paid (pure premium)– Multiplied by a wedge (1.3, 1.6)Multiplied by a wedge (1.3, 1.6)

Moral hazard + adverse selectionMoral hazard + adverse selection

– Multiplied pure premium by (1 – subsidy)Multiplied pure premium by (1 – subsidy)

Model Tests and Double ChecksModel Tests and Double Checks

Experimented with a correlation matrix Experimented with a correlation matrix based on soil types in the countybased on soil types in the county– Varied means and standard deviations based on soil Varied means and standard deviations based on soil

typetype– Results were relatively the same. Results were relatively the same.

Rates generated are relatively the same Rates generated are relatively the same as insurance rates charged for county and as insurance rates charged for county and unit policies. unit policies.

Farm With High Correlation Farm With High Correlation Farm to County Yield Farm to County Yield ≈ ≈ 0.950.95

0 50 100 150 200 2500

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Yield(bu/acre)

Cum

ula

tive P

robabili

ty

Farm Yield + APH

Farm Yield No InsFarm Yield + GRP

Model ResultsModel Results

Corr.Corr. No InsNo Ins APH Ins APH Ins GRP InsGRP Ins

Downside Downside VarianceVariance

Downside Downside VarianceVariance

Downside Downside VarianceVariance

≈ ≈ 0.950.95 904904 205205 142142

≈ ≈ 0.870.87 10151015 240240 360360

≈ ≈ 0.800.80 10171017 237237 441441

Farm With Low CorrelationFarm With Low Correlation Farm to County Yield Farm to County Yield ≈ ≈ 0.800.80

0 20 40 60 80 100 120 140 160 180 2000

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

Yield(bu/acre)

Cum

ula

tive P

robabili

ty

Farm Yield + APH

Farm Yield No InsFarm Yield + GRP

Model ResultsModel Results

Corr.Corr. No InsNo Ins APH Ins APH Ins GRP InsGRP Ins

Downside Downside VarianceVariance

Downside Downside VarianceVariance

Downside Downside VarianceVariance

≈ ≈ 0.950.95 904904 205205 142142

≈ ≈ 0.870.87 10151015 240240 360360

≈ ≈ 0.800.80 10171017 237237 441441

Increased KnowledgeIncreased Knowledge

Spatial diversity helps GRPSpatial diversity helps GRP

Center of county helps GRPCenter of county helps GRP

As wedge on APH insurance As wedge on APH insurance ↑, the relative ↑, the relative performance of GRP to performance of GRP to APH APH ↑↑

Research and ExtensionResearch and Extension

GRP Evaluation spread sheetGRP Evaluation spread sheet– http://www.aec.msu.edu/agecon/blackj/grp.htmhttp://www.aec.msu.edu/agecon/blackj/grp.htm

GRP Staff PaperGRP Staff Paper– http://http://www.aec.msu.edu/agecon/blackj/grp.htmwww.aec.msu.edu/agecon/blackj/grp.htm

GRP and GRIP MATLAB program GRP and GRIP MATLAB program codecode– http://www.aec.msu.edu/agecon/blackj/grp.htmhttp://www.aec.msu.edu/agecon/blackj/grp.htm

An Improved Model WouldAn Improved Model Would

Use more units in the example countyUse more units in the example county

The model used did not have enough The model used did not have enough detail in the example countydetail in the example county– Used 16 UnitsUsed 16 Units

Actual county has 500+ square milesActual county has 500+ square miles

Perhaps 36 unitsPerhaps 36 units

Continued ResearchContinued Research

M.S. Plan B PaperM.S. Plan B Paper– Insurance Policies that Trigger on Insurance Policies that Trigger on

County IndexesCounty IndexesGRP, GRIP, GRIP HROGRP, GRIP, GRIP HRO

Unit policies will be compared to county Unit policies will be compared to county policies policies

APH, RA and RA HRO APH, RA and RA HRO – http://http://www.aec.msu.edu/agecon/blackj/grp.htmwww.aec.msu.edu/agecon/blackj/grp.htm

AcknowledgementsAcknowledgements

Insurance agents for input and reviewInsurance agents for input and review– Special thanks to Lisa TuggleSpecial thanks to Lisa Tuggle

Farmers for case study information and Farmers for case study information and reviewreview

Practical Procedures For Practical Procedures For Evaluating Crop Insurance Policies Evaluating Crop Insurance Policies

That Trigger On County YieldThat Trigger On County Yield

Ben Chaffin, Graduate Research AssistantBen Chaffin, Graduate Research AssistantJ. Roy Black, ProfessorJ. Roy Black, Professor

Department of Agricultural EconomicsDepartment of Agricultural EconomicsXiaobin Cao, Graduate Research Assistant, Xiaobin Cao, Graduate Research Assistant,

Agricultural Economics/StatisticsAgricultural Economics/StatisticsMichigan State UniversityMichigan State University