The Favorable Impact of Index-Based
Livestock Insurance (IBLI): Results
among Ethiopian and Kenyan
Pastoralists
Christopher B. Barrett, Cornell University
Seminar at USAID Headquarters, Washington, DC
February 10, 2015
Motivation: Poverty Traps And Catastrophic Risk
There is strong evidence of
poverty traps in the arid and
semi-arid lands (ASAL) of
northern Kenya and
southern Ethiopia. These put
a premium on risk mgmt.
Catastrophic herd loss risk
due to major droughts
identified as the major
cause of these dynamics.
Source: Lybbert et al. (2004 EJ) on Boran pastoralists in s. Ethiopia.
See also Barrett et al. (2006 JDS) among n. Kenyan pastoralists,
Santos & Barrett (2011 JDE) on s. Ethiopian Boran.
Motivation: Increased Risk From Climate Change
Pastoralist systems adapted to climate regime. But resilient to a
shift in climate? Many models predict increased rainfall variability
(i.e., increased risk of drought).
Herd dynamics differ b/n good
and poor rainfall states, and so
change with drought (<250
mm/ year) risk.
In southern Ethiopia, doubling
drought risk would lead to
system collapse in expectation
in the absence of any change to
prevailing herd dynamics.
Source: Barrett and Santos (EcolEcon 2014)
Motivation: Standard Responses to Drought
Standard responses to major drought shocks:
1) Post-drought restocking
2) Food aid
Key Problems:
-
Slow
Expensive (in part, because it’s slow)
Targeting challenges
Food aid can reinforce sedentarization/foster dependency
Core issue: If transfers go only to the poor who are already in
the poverty trap, the numbers of poor will grow as shocks
knock others below the poverty trap threshold. In the longrun, the ex ante poor worse off as others join their ranks and
compete for scarce social assistance resources. (see Barrett,
Carter & Ikegami 2012 for more general theory/illustrations)
Alternative Responses: Insurance?
Commercially sustainable insurance can:
•
•
•
•
Prevent downward slide of vulnerable populations
Crowd-in investment and accumulation by the poor
Induce financial deepening by crowding-in credit
Let us focus humanitarian resources on the needy
But can insurance be sustainably offered in the ASAL?
Conventional (individual) insurance unlikely to work,
especially in small scale pastoral/agro-pastoral sector:
• Very high transactions costs, esp. w/little financial
intermediation among pastoralists
• Moral hazard/adverse selection
The Potential of Index Insurance
Index insurance is a variation on traditional insurance:
- Do not insure individual losses.
- Instead insure some “index” measure that is strongly
correlated with individual losses.
(Examples: rainfall, remotely sensed vegetation index, area
average yield, area average herd mortality loss).
- Index needs to be:
- objectively verifiable
- available at low cost in real time
- not manipulable by either party to the contract
The Potential of Index Insurance
Index insurance can obviate the problems that make individual
insurance unprofitable for small, remote clients:
- No transactions costs of measuring individual losses
- Preserves effort incentives (no moral hazard) as no single
individual can influence index
- Adverse selection does not matter as payouts do not depend
on the riskiness of those who buy the insurance
Index insurance can, in principle, be used to create a timely,
financially sustainable, self-targeting safety net to protect
pastoralists against catastrophic drought shocks.
Could also accelerate herd recovery, altering herd dynamics and
averting system collapse if drought frequency increases. Perhaps
crowd in value-adding investments, too!
The Major Challenges of Index Insurance
1. High quality data (reliable, timely, non-manipulable, longterm) to design/price product and to determine payouts
2. Minimize uncovered basis risk through product design. Is it
insurance or a lottery ticket? All turns on basis risk!!!
3. Innovation incentives for insurers/reinsurers to design and
market a new product and global market to support it
4. Establish informed effective demand, especially among a
clientele with little experience with any insurance, much less a
complex index-based insurance product
5. Low cost delivery mechanism for making insurance available
for numerous small and medium scale producers
Index-Based Livestock Insurance: Design
The signal: Normalized Difference Vegetation Index (NDVI) collected by satellite
Response function: regress historic livestock mortality onto transforms of historic
cumulative standardized NDVI (Czndvi) data. In Borana, just NDVI. Designed to
minimize household-level basis risk.
Indemnity payments: In Kenya,(Jensen,
predicted
livestock
Barrett
&2014)mortality >15% according to:
𝑚𝑎𝑥 𝑖𝑛𝑑𝑒𝑥𝑑,𝑡 (𝐿𝑑,𝑡 , 𝜇𝑑,𝑡 ) − 0.15, 0 ∗ 𝑣𝑎𝑙𝑢𝑒 𝑜𝑓 𝑙𝑖𝑣𝑒𝑠𝑡𝑜𝑐𝑘 𝑖𝑛𝑠𝑢𝑟𝑒𝑑
1 year contract coverage
Temporal Structure of IBLI contract:
12 month contract sold during 2month sales windows just prior to
usual start of seasonal rains. Payouts
March 1 and/or October 1.
Chantarat et al. JRI 2013
LRLD season coverage
Jan
Feb
Sale period
For LRLD
Mar
Apr
May
Jun
Jul
Aug
SRSD season coverage
Sep
Oct
Nov
Dec
Jan
Feb
Period of NDVI observations for
constructing LRLD mortality index
Sale period
For SRSD
Period of NDVI observations
For constructing SRSD
mortality index
Predicted LRLD mortality is announced.
Indemnity payment is made if IBLI is triggered
Predicted SRSD mortality is announced.
Indemnity payment is made if IBLI is triggered
Index-Based Livestock Insurance: Implementation
Commercial underwriters: In Kenya: UAP, APA, Takaful. In Ethiopia: OIC
International reinsurers: Swiss Re, Africa Re
Lots of implementation challenges
IBLI team developed extension/(Jensen, Barrett &2014)
financial education programs to
(randomly) inform prospective buyers.
IBLI team (randomly) distributed
discount coupons to induce uptake and
to establish price elasticity of demand.
Payouts in Kenya in Oct 2011, Mar 2012
Payouts in Ethiopia in Nov 2014
IBLI Pilots in Ethiopia and Kenya
IBLI products (surveys) launched in Marsabit, Kenya in Jan 2010
(Oct 2009) and in Borana, Ethiopia, in Aug 2012 (Mar 2012).
(Jensen, Barrett &2014)
Kenya sampling overlaid with HSNP coverage as research design.
IBLI: Significant Basis Risk Remains
Covariate risk is important but
household losses vary a lot …
and the index does not
perfectly track covariate losses.
(Jensen, Barrett & Mude 2014)
Notes: The left figure illustrates the covariate (average) loss rate in each
season. The right figure illustrates the distribution of losses within each
seasons. The boxes depict the interquartile range, the upper and lower adjacent
values are either 3/2 the interquartile range or the value furthest from the
median. The remaining observations fall outside the adjacent values.
Notes: Covariate loss-index observations are seasonal division
average mortality paired with the index value for that divisionseason. Fitted lines and confidence intervals are generated by
regressing livestock mortality rates on the index.
- IBLI hhs still hold most risk: 62-77% of total risk exposure remains
- Most basis risk is idiosyncratic and random, not targetable or correctable.
- Significant spatial variation in covariate share – geographically target IBLI?
Jensen, Barrett & Mude 2014
IBLI: A Highly Imperfect Product
Because of basis risk, esp. false negatives, IBLI cannot
stochastically dominate no insurance.
(Jensen, Barrett &
Histograms of livestock survival rate and net livestock survival rate with full insurance. Tally to the left of zero, between zero and
one, and to the right of one are in green.
Survival rate w/o insurance (L) and net of prem/indemnity payments w/IBLI (R).
Note: - small probability of negative survival rates!
- increased dispersion of outcomes due to false payments>losses
Jensen, Barrett & Mude 2014
IBLI Uptake Significant … But So Is Disadoption
In HH surveys , in Borana (Ethiopia)/Marsabit (Kenya):
- 54/44% ever purchased IBLI within first 4 sales periods
- But repurchase rates low: 18-68%/16-27%
- High rates of disadoption : 26/41% within 2 years
IBLI Uptake Is Also Predictable …
Capacity to predict uptake patterns is reasonably strong:
Unconditional observed / predicted
(Cond. FE) likelihood of buying IBLI
Observed / predicted (Cond. FE)
level of purchases (|buying IBLI)
Key determinants of IBLI uptake
General uptake findings — robust across specifications and surveys
Price: Responsive to premium rate (price inelastic). Price elasticity grows
w/design risk.
Design Risk: Design error reduces uptake; greater effect at higher premium rates.
Idiosyncratic Risk: Hh understanding of IBLI increases effect of idiosyncratic risk
Understanding: Extension/marketing improves accuracy of IBLI knowledge but no
independent effect of improved understanding on uptake.
Herd size: Likelihood of uptake increasing in HH herd size
Liquidity: IBLI purchase increasing w/HSNP participation
Intertemporal Adverse Selection: HHs buy less when expecting good conditions.
Spatial Adverse Selection: Divisions with relatively more covariate risk see higher
uptake and level of coverage increases with variation in division average losses.
Gender: no gender diff in uptake. Women more sensitive to risk of new product.
Bageant 2014; Jensen, Mude & Barrett 2014; Takahashi et al. 2014
IBLI’s Impacts: Herd mortality risk
Proportion of households for whom IBLI improves their
position with respect to each statistic
Statistic
Proportion
Loaded
Subsidized
Proportion of households that are better off
with IBLI&
than without (Simulated
utility analysis)
Mean
Variance
Skewness
Semi-Variance
Unsubsidized
0.232
0.359
0.817
0.374
1.000
0.359
0.817
0.609
Jensen, Barrett & Mude 2014
IBLI’s Impacts: Livestock productivity/income
Dependent
Variable
Production
strategies:
Herd Size
Veterinary
Expenditures (KSH)
Household is
Partially or Fully
Mobile
Production
outcomes:
Milk income (KSH)
Milk income per
TLU (KSH)
HSNP
Cumulative Past Current
Participation
Client
IBLI
Cumulative Current Coverage
Past Coverage
(TLU)
0.0583
(0.470)
-2.531
(2.963)
[4.693]
-5.634***
(1.970)
-0.270
(0.693)
[3.543]
48.22
(57.48)
261.6
(363.9)
[13.78]
584.8*
(324.7)
-46.21
(127.2)
[15.17]
0.0376**
(0.0172)
0.251***
(0.0686)
[17.76]
-0.0669
(0.111)
0.0386
(0.0481)
[14.86]
193.6
(275.6)
307.0
(1,039)
[11.94]
1,688*
(970.0)
840.6*
(473.6)
[11.46]
63.65***
(20.14)
-97.35
423.5***
63.81
(79.24)
(118.1)
(47.23)
[14.72]
[13.05]
A complete list of covariates, coefficient estimates, and model statistics can be found
in Jensen, Mude & Barrett (2014). Clustered and robust standard errors in
parentheses. Model F-stat in brackets. *** p<0.01, ** p<0.05, * p<0.1.
HSNP Participation:
•Improves the likelihood that
target type (poor, old, many
dependents) households
remain mobile, an important
characteristic for pastoralists
IBLI coverage:
•Increases investments in
maintaining livestock
through vet expenditures
•Increases total and per TLU
income from milk.
Note: TLU veterinary expenditures are
pos/sign related to milk productivity
Jensen, Barrett & Mude 2014
IBLI’s Impacts vs. HSNP: Normalized by cost
Cost structure
Total Program
Cost/Participant
Marginal Cost of an
Additional Participant
Income from Milk
Impact Impact/
KSH
992
0.021
2,631
0.067
992
2,631
0.031
1.667
Income per AE
Impact Impact/
KSH1
394
0.083
263
0.070
394
263
0.124
1.666
MUAC
Impact Impact/
KSH2
1.097
0.022
0.337
0.026
1.097
0.337
0.033
0.623
All in real 2009 Kenya Shillings. Impacts are estimated using the average client value and costs provided below, and
parameter estimates in the previous two slides. 1Results are multiplied by 10. 2Results are multiplied by 1,000.
Average values (final survey round, clients)
VOI
Mean VOI in Final Period
HSNP
0.86
HSNPC
3.76
IBLI
0.60
IBLIC
1.26
Values are calculated for the subset of clients
in each program.
Average cumulative cost/client by the final round (KSH)
HSNP
Total Program
47,600
Cost/Participant (2.7BN/57,811 HH)
Marginal Cost of
an Additional
Participant
31,700
(13.9 transfers)
IBLI
43,200
(99MM/3,297
contracts*1.44
contract/HH)
1,580
(3.25 TLUs)
Jensen, Barrett & Mude 2014
IBLI’s Impacts vs. HSNP: Normalized by cost
Total Program
Cost/Participant
Income from Milk
Impact/
Impact
KSH
992
0.021
2,631
0.067
Marginal Cost of an
Additional Participant
992
2,631
Cost structure
0.031
1.667
Income per AE
Impact Impact/
KSH1
394
0.083
263
0.070
394
263
0.124
1.666
MUAC
Impact Impact/
KSH2
1.097
0.022
0.337
0.026
1.097
0.337
0.033
0.623
All in real 2009 Kenya Shillings. Impacts are estimated using the average client value and costs provided below, and
parameter estimates in the previous two slides. 1Results are multiplied by 10. 2Results are multiplied by 1,000.
Average values (final survey round, clients)
VOI
Mean VOI in Final Period
HSNP
0.86
HSNPC
3.76
IBLI
0.60
IBLIC
1.26
Values are calculated for the subset of clients
in each program.
Average cumulative cost/client by the final round (KSH)
HSNP
Total Program
47,600
Cost/Participant (2.7BN/57,811 HH)
Marginal Cost of
an Additional
Participant
31,700
(13.9 transfers)
IBLI
43,200
(99MM/3,297
contracts*1.44
contract/HH)
1,580
(3.25 TLUs)
Jensen, Barrett & Mude 2014
IBLI’s Impacts: Less adverse post-drought coping
Marsabit HHs received IBLI indemnity payments in October 2011,
near end of major drought. Survey HHs with IBLI coverage report
much better expected behaviors/outcomes than the uninsured:
- 36% reduction in likelihood of distress livestock sales, especially
(64%) among modestly better-off HHs (>8.4 TLU)
- 25% reduction in likelihood of reducing meals as a coping
strategy, especially (43%) among those with small or no herds
IBLI appears to provide a flexible safety net, reducing reliance on
the most adverse behaviors undertaken by different groups.
Janzen & Carter 2013 NBER
IBLI’s Impacts: Household subjective well-being
Borana survey HHs report overall life satisfaction. In principle,
insurance helps risk averse people even when it doesn’t pay out.
But an imperfect product with commercial loadings might not.
There had been no payout in Ethiopia (pre-11/14). So use subjective
well-being measures to assess welfare gains even w/o indemnities.
To deal with potential heterogeneity problems associated with SWB
(attitudinal measures), we correct our SWB measures using
hypothetical vignettes, using current best practice, and verify with
alternative measures to ensure robustness of findings.
Hirfrot , Barrett, Lentz and Taddesse. 2014
IBLI’s Impacts: Household subjective well-being
Use randomized treatments to instrument for IBLI and then estimate:
𝑆𝑊𝐵𝑖𝑣𝑡 = 𝛼 + 𝛽 𝐼𝐵𝐿𝐼𝑖𝑣𝑡 + 𝜃 𝑇𝐿𝑈𝑖𝑣𝑡 + 𝜋𝐼𝐵𝐿𝐼𝑖𝑣𝑡−1 + 𝛿`𝑋𝑖𝑣𝑡 + 𝛾𝑖 + 𝑅𝑣 + 𝜀𝑖𝑣𝑡
There are at least two ways IBLI can influence SWB:
1) Non-monetary (psychological) benefits or costs
• Insurance may give peace of mind about adverse outcomes (𝛽 > 0)
• Insurance could increase stress if basis risk is high (𝛽 < 0)
• Buyer’s remorse: 𝜋 < 0
2) Monetary benefits or costs – effect on net income/wealth
• Since premium payment reduces net income/wealth, indemnity
payment increases it, net indemnity payments will influence SWB.
• This effect is captured by 𝜃.
Hirfrot , Barrett, Lentz and Taddesse. 2014
IBLI’s Impacts: Household subjective well-being
Ordered logit regression estimates (extensive margin)
Dependent variable: Vignette
corrected SWB
Predicted IBLI uptake
Model
(1)
0.825***
(0.288)
-0.453***
(0.157)
0.015**
(0.007)
Model
(2)
Model
(3)
0.746***
0.894***
(0.235)
(0.269)
Predicted lapsed IBLI
-0.463***
-0.469***
(0.166)
(0.170)
Number of TLU owned
0.013*
0.013*
(0.007)
(0.007)
Asset index
0.287***
0.243***
(0.066)
(0.067)
Annual income (‘000 Birr)
0.003
0.003
(0.005)
(0.005)
Controls
No
No
Yes
Reera fixed effect
No
No
Yes
Observations
1,530
1,530
1,530
Number of groups (households)
550
550
550
Clustered standard errors in Parenthesis; *** p<0.01, ** p<0.05, * p<0.1
Controls : Household
head gender, age, age
squared, schooling,
household size, and
household composition.
•IBLI has +++ effect
on SWB
•Sig. buyer’s
remorse effect
•But + peace of mind
effect >> - buyer’s
remorse effect
Hirfrot , Barrett, Lentz and Taddesse. 2014
IBLI’s Impacts: Household subjective well-being
Ordered logit regression estimates (intensive margin)
Dependent variable: Vignette
corrected SWB
Predicted TLUs insured
Model
(1)
0.126***
(0.035)
-0.073***
(0.022)
0.015**
(0.007)
Model
(2)
Model
(3)
0.131***
0.136***
(0.033)
(0.033)
Predicted lapsed TLUs
-0.071***
-0.074***
(0.023)
(0.023)
Number of TLU owned
0.012*
0.013*
(0.007)
(0.007)
Asset index
0.331***
0.295***
(0.066)
(0.066)
Annual income (‘000 Birr)
0.003
0.003
(0.005)
(0.005)
Controls
No
No
Yes
Reera fixed effect
No
No
Yes
Observations
1,530
1,530
1,530
Number of groups (households)
550
550
550
Clustered standard errors in Parenthesis; *** p<0.01, ** p<0.05, * p<0.1
Controls : Household
head gender, age, age
squared, schooling,
household size, and
household composition.
•IBLI has +++ effect
on SWB
•Sig. buyer’s
remorse effect
•But + peace of mind
effect >> - buyer’s
remorse effect
Hirfrot , Barrett, Lentz and Taddesse. 2014
IBLI’s Impacts: Household subjective well-being
• IBLI has a positive, statistically significant effect on HH wellbeing, even after premium payment and w/o any indemnity
payments
• IBLI coverage for 3 TLU moves a HH 1 step up the SWB scale
• Insuring 15 TLU (roughly baseline sample mean) shifts HH
from lowest to highest SWB category
• Ex post of contract, purchasers exhibit some buyer’s remorse in
the absence of indemnity payments.
• But the positive effect of IBLI coverage is significantly higher
than the negative effect of buyer’s remorse.
Even with prospective buyer’s remorse, IBLI purchase improves
subjective well-being.
Hirfrot , Barrett, Lentz and Taddesse. 2014
Although IBLI offers incomplete coverage
against herd loss and will not help all people,
uptake is solid and IBLI has clear favorable
impacts on purchasers.
IBLI offers a promising option for addressing
poverty traps that arise from catastrophic
drought risk … and impacts/$ > cash transfers
Thank you for your time, interest and comments!
For more information visit www.ilri.org/ibli/