1
3rd CRAM workshop 20110923-26, RCMRD, Nairobi,
Kenya
Weather index-based crop insurance a giant with
clay feet?
René Gommes, JRC/Ispra
2
Overview

• Why is there so much talk of insurance and why
  are there so many projects?
• Example of determination of crop yield indices in
  ETH
• What are the problems and how serious are they?

3
Farm household risk avoidance mechanisms

• Risk is a main factor of agricultural production
  stagnation
• Households choose low-risk, low-return
  activities new technologies are not adopted
  because of risks
• Financial institutions restrict lending to farm
  households (because of widespread defaults
  following a disaster)

Source Skees et al., 2007
4
Agricultural insurance products classification
Source Iturrioz, 2009
5
How does index insurance work?
6
Advantages of index insurance

• Large populations dependent on (non -irrigated)
  agriculture with direct dependence on rainfall
• Addresses correlated weather risks (uncorrelated
  risk car accident)
• Easy to administer low overhead, standard
  contracts (no individual loss-assessment),
  reduced monitoring cost ? maximum payout

Source Skees et al., 2007
7
Factors of growth of agricultural insurance market

• Increase of value of production (higher prices)
• Increase in value of agric. Assests, which leads
  to greater risk of losses and demand for
  insurance
• New markets and public sector support new markets
• 62 in US/Canada (1 in Africa) 74 multiple
  peril crop insurance, 17 hail

Source Iturrioz, 2009
8
Overview

• Why is there so much talk of insurance and why
  are there so many projects?
• Example of determination of crop yield indices in
  ETH
• What are the problems and how serious are they?

9
General approach

• Develop yield functions
• Use yield functions to reconstruct yield time
  series
• Determine probabilities of yields

10
2007-2009 enumeration areas
11
Maize yield by EA (2008-09)
12
Teff yield factors
13
Yield functions
14
(No Transcript)
15
Maize yield (calibrated 2001 census 2008-09
EA)
16
Rainfall-reconstructed maize yield in Lalo-Midir
17
Woreda clusters based on yield time series of 6
crops
18
Wheat yield percentiles, group G
Regression graph prepared with CurveExpert
19
Overview

• Why is there so much talk of insurance and why
  are there so many projects?
• Example of determination of crop yield indices in
  ETH
• What are the problems and how serious are they?

20
The main message...

• There is a mismatch between the level of
  sophistication of the index technology and the
  insurance technology the first is usually
  given or improvised while the second get lots
  of scientific attention
• In other words, the weak link (i.e. base) of many
  index-based insurance schemes may be the indices
  themselves

21
Eight methods compared...
22
IDW with alt. correc-tion (200907-I rainfall)
23
NN spatial pattern (200907-I rainfall)
24
Nearest neighbour pros cons

• Pro
• Simplicity
• Limited data requirements
• Transparency
• Cons
• funny spatial behaviour sharp limits no
  altitude effects
• Sensitive to missing data (cannot work without
  missing data estimation!)
• Not very tamper-resistant

25
Local index calculation
red circles spatial intrepolation
26
Conclusions 1/2

• Indices can be significantly better than some
  rainfall-based proxy borrowed from the nearest
  weather stations indices must and can be made
  location-specific
• Data collection for yield index development must
  be integrated in the overall data collection
  schemes of projects
• We need insurance-oriented index calibration
  methods (e.g. to minimize false negatives)

27
Conclusions 2/2

• Simple and transparent indices are not
  compatible with
• deficient data and sparse data (deficiency
  includes uncertainty about station location)
• agronomic performance of indices (rainfall, and
  rainfall probabilities are not sufficient to
  function as proxies for yield)
• The fact that indices must be tamper-proof

28
Thank you!