smallholder access to weather securities: demand and impact on consumption and production decisions...
TRANSCRIPT
Smallholder access to weather securities: demand and impact on consumption and
production decisions
Tirtha Chatterjee, Isaac Manuel, Ashutosh Shekhar
Centre for Insurance and Risk Management, CIRM-IFMR
Ruth V. Hill, Peter Ouzounov, Miguel Robles
International Food Policy Research Institute - IFPRI
Netherlands – April 2012
Research problem:• Smallholders in developing countries are exposed to weather shocks.
• Weather shocks have large impact on output
• In the absence of efficient mechanisms to transfer/share risks then impact on welfare
– Negative impact on investment decisions– Volatile income and consumption
• Smallholders have none or very limited acces to weather insurance markets
• Weather index-based products is an effort to provide access to smallholder to weather insurance markets
• Uptake on existing weather index-based products is low
Research problem:
• We propose a new approach in providing weather index–based insurance products
– Multiple weather securities that pay a fixed amount as opposed to a unique policy
– Weather securities are simple and flexible
• We run a pilot project to provide weather securities and understand demand (uptake) and impact on consumption and production decisions
• In this presentation: What is the impact of three interventions on weather securities uptake? (preliminary results)
– Price discounts
– Insurance literacy training
– Distance to weather station (basis risk)
Research questions:
Product: Basic concept
• Basic product: weather security (rainfall excess)…
Payout (Rs.)
Triggervalue
ExitIndex
Price (premium)
Product: basic concept…
• Basic product: weather security (rainfall deficit)…
Payout (Rs.)
Triggervalue
ExitIndex
Price (premium)
Product: multiple securities• We identify 3 cover periods:
• For each cover period we have multiple (at least two) products:– Different trigger values– Different prices– Same payouts
• Farmers are free to choose among different products!
Cover period
Jun 25 – Jul 20 Jul 21 – Sep 15 Sep 16 – Oct 15
Crop stage Sowing and germination Vegetative, reproductive and maturity
Harvest
Peril • Excessive rainfall • Excessive rainfall• Deficit rainfall
• Excessive rainfall
Index Maximum rainfall on any single day
Total cumulative rainfall Maximum rainfall on any single day
Final products: Dewas district
Security Cover period index strike Exit Payout 1 condition
Payout 2 condition Premium
incl of ST(Rs)
(Rs 1000) (Rs 4000) Security 1 Jun 25 – Jul 20 maximum rainfall on any
single day (mm)95 200 Index >
strikeIndex > exit
352Security 2 Jun 25 – Jul 20 maximum rainfall on any
single day (mm)120 200 Index >
strikeIndex > exit
265Security 3 Jul 21 – Sep 15 Total cumulative rainfall
(mm)280 130 Index <
strikeIndex < exit
265Security 4 Jul 21 – Sep 15 Total cumulative rainfall
(mm)340 130 Index <
strikeIndex < exit
352Security 5 Jul 21 – Sep 15 Total cumulative rainfall
(mm)635 960 Index >
strikeIndex > exit
352Security 6 Jul 21 – Sep 15 Total cumulative rainfall
(mm)700 960 Index >
strikeIndex > exit
265Security 7 Sept 16 – Oct 15 maximum rainfall on any
single day (mm)70 160 Index >
strikeIndex > exit
352Security 8 Sept 16 – Oct 15 maximum rainfall on any
single day (mm)85 160 Index >
strikeIndex > exit
265
Not implemented
triggered
Location and sample
• Product was marketed in 3 districts of Madhya Pradesh, India: Dewas, Bhopal, Ujjain
• Research focus: 30 landowning households per village
District Villages Households (total)
Households (sample)
%
Dewas 29 5356 881 16.4%
Bhopal 30 7264 904 12.4%
Ujjain (only last cover period)
13 2777 398 14.3%
Total 72 15397 2183 14.2%
Data: oversampled hhs with larger land holding and higher education
Treatment (Insurance)
Control (No insurance)
Total
Villages 72 38 110
Sample households (30 per village)
2183 1156 3339
Land holding (acres) In sample
8.4 8.9 8.6
Land holding (acres)Out of sample
3.5 3.7 3.6
Schooling head hh (yrs) In sample
5.4 5.7 5.5
Schooling head hh (yrs)Out of sample
4.3 4.2 4.3
Data
• Average 8.6 acres of land, 90% sown with soy• Over last 10 years, 15% experienced flood and 40%
experienced drought• 35% trust private insurance schemes• Low knowledge of insurance (1/2 correct)• 26% believe closest weather station is a good measure
of rain for their field
Exogenous (randomized) treatments
1. Insurance literacy training
– Basic training (2 hours) 72 (all) villages
– Intensive training (4 hours) 35 villages
2. Three new randomly placed reference weather stations
– 2 in Dewas: 16/29 villages
– 1 in Bhopal: 12/30 villages
No. of villages Average distance
Existing weather station 44 10 Km
New weather station 28 5 Km
Exogenous (randomized) treatments
3. Allocation of price discount vouchers
• In Dewas and Bhopal (59 villages)
– Random selection at household level:
– 5 hhs x [ Rs. 45, Rs. 90, Rs 135, Rs 180 ]
– 10 hhs x No discount
– Only sample households received discounts
• In Ujjain (13 villages)
– Random selection at village level (all hhs receive vouchers)
– 2 villages x [ Rs. 30, Rs 60, Rs 90, Rs 120 ]
– 5 villages x No discount
Research results, I
Treated Villages(all households)
Household sample
# of Sales
Acres insured per sale Uptake
Acres insured per purchasing hh
Ujjain 115 0.5 2.5% (10/398) 0.9
Dewas 45 1.5 1.8% (16/881) 2.7
Bhopal 141 0.3 13.6% (123/904) 0.4
Total 301 0.6 6.8% (149/2183) 0.6
• Overall uptake 6.8%• On average, they insured less than an acre and much less than their
total soy land ownership• There are important differences between districts
Research Results, IISummary Of Results: Dependent variable is whether household bought insurance or not (1) (2) (3) (4) (5) (6) (7) (8)
OLS IV OLS IV OLS IV OLS IV
Discount (ratio of price) 0.231*** 0.227*** 0.264*** 0.266***
(0.052) (0.052) (0.061) (0.061)
Distance to ref. Station -0.011* -0.011* -0.010* -0.010*
(0.006) (0.006) (0.006) (0.006)
Additional Training 0.049** 0.038 0.051** 0.050**
(0.024) (0.028) (0.024) (0.024)
District Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes
HH characteristic covariates No No No No Yes Yes Yes Yes
Observations 2,183 2,183 2,183 2,183 2,164 2,164 2,164 2,164
R-squared 0.124 0.061 0.099 0.134 0.005 0.072 0.082
Standard errors adjusted for clustering at village level are in parentheses; *** p<0.01, ** p<0.05, * p<0.1
Research ResultsSummary Of Results: Dependent variable is log of land insured (1) (2) (3) (4) (5) (6) (7) (8)
OLS IV OLS IV OLS IV OLS IV
Log of price of cheaper contract -0.582*** -0.566*** -0.594***
-0.596***
(0.133) (0.136) (0.137) (0.137)
Distance to ref. station -0.029 -0.028 -0.026 -0.027
(0.022) (0.022) (0.022) (0.022)
Additional training 0.167* 0.140 0.178* 0.180**
(0.091) (0.103) (0.090) (0.091)
District Fixed Effects Yes Yes Yes Yes Yes Yes Yes Yes
HH characteristic covariates No No No No Yes Yes Yes Yes
Observations 2,183 2,183 2,183 2,183 2,164 2,164 2,164 2,164
R-squared 0.100 0.043 0.110 0.023 0.052 0.094
Standard errors adjusted for clustering at village level are in parentheses; *** p<0.01, ** p<0.05, * p<0.1
• Distance to weather station has no effect quantity bought, but only on whether household buys or not
Implications policy and practice
Cost and Benefit (uptake) analysis of interventions
• ILT• Cost per‐person $10.40 -> + 5% points take-up• Cost of Increasing take‐up rates by 10% points = $20.80 per-person
• New weather stations • Cost per-person $6.67 -> + 5% points take‐up• Cost of Increasing take‐up rates by 10% points = $13.34 per-person
• Price discounts• To increase take‐up rates by 10% points a discount of 115 Rs ($2.30) per policy
is needed.• In Bhopal and Dewas the amount spent on discounts per-person who was
offered a discount was $0.2 -> increase in take-up by 10% points
• Price discounts is the most cost effective intervention
Discussion
• Marketing efforts are key! We have casual evidence that take-up differences across districts is related to marketing efforts by insurance company
– Second round implementation will pay more attention to incentives to insurance agents
– Research pilots need to encourage permanent presence among treatment group
• Ideal study is on impact on consumption and production decisions (welfare)– We requiere higher take-up rates
• What’s the ideal demand analysis of multiple products?– System of demand equations
– Again we need higher take-up rates