• Share
  • Email
  • Embed
  • Like
  • Save
  • Private Content
An experiment on the impact of weather shocks and insurance on risky investment
 

An experiment on the impact of weather shocks and insurance on risky investment

on

  • 1,022 views

 

Statistics

Views

Total Views
1,022
Views on SlideShare
1,021
Embed Views
1

Actions

Likes
0
Downloads
11
Comments
0

1 Embed 1

http://www.slideshare.net 1

Accessibility

Upload Details

Uploaded via as Adobe PDF

Usage Rights

© All Rights Reserved

Report content

Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

Cancel
  • Full Name Full Name Comment goes here.
    Are you sure you want to
    Your message goes here
    Processing…
Post Comment
Edit your comment

    An experiment on the impact of weather shocks and insurance on risky investment An experiment on the impact of weather shocks and insurance on risky investment Presentation Transcript

    • ETHIOPIAN DEVELOPMENT RESEARCH INSTITUTE An experiment on the impact of weather shocks and insurance on risky investment Ruth Vargas Hill and Angelino Viceisza 1
    • Outline 1. Motivation 2. Summary of methodology and results 3. Background on study site 4. Experiment protocol 5. Empirical strategy 6. Results 7. Conclusions and further research 2
    • Motivation: impact of weather shocks • Weather shocks have a significant impact on the Ethiopian economy and on the welfare of rural households • The expectation of shocks also reduces investment in high-return but high risk activities: • Households more susceptible to risk of low consumption are less likely to apply fertilizer (Dercon and Christiaensen 2006) • Encourages production of low-risk buy low return crops (e.g. enset) • Likelihood of investing in non-farm activities is lower for those more susceptible to risk (Rijkers et al 2008) • Existing policy response: • Food aid • Safety nets (PSNP) 3
    • Motivation: new tools for insurance? • Can we do more? In particular can we develop weather insurance to: • Support efficient delivery of public safety nets (e.g. subsidizing premiums) • exist as a market in its own right (commercial insurance purchases) • Innovations in weather-index insurance has encouraged new work on developing weather insurance markets for small-scale farmers • Traditional weather insurance: assessing loss is expensive when farms are small-scale • Weather-based index insurance: pays not by assessing loss in farmers fields, but on information on the quality of the rain from the nearby weather station 4
    • Motivation: new tools for insurance? As elsewhere, a number of pilots in Ethiopia • Pilot in Alaba (SNNPR) with Ethiopian Insurance Company in 2006 (World Bank) • Development of LEAP (Livelihood Early Assessment Protection) software to design drought indexes specifically for the local Ethiopian context was initiated by WFP, World Bank and FAO • Nyala index-insurance pilot for haricot beans in Boset Woreda (SNNPR) in 2009 (WFP) • Index-insurance pilot for teff in Tigray in 2009 (Oxfam) 5
    • Motivation: limits to new tools • Promising new product: • Cheaper to administer, also no moral hazard or adverse selection • But this comes at expense of: • Basis risk (payout does not always coincide with loss) • Difficult to understand (depends on unseen mm, not actual loss) • Also still: • Trust • Liquidity and budget constraints: insurance is still costly • Cost is worth it if insurance provision results in change in behavior that increases average income of farmers • This could be applying fertilizer, or growing more high-risk but high-return crops 6
    • Motivation: limits to new tools • Promising new product: • Cheaper to administer, also no moral hazard or adverse selection • But this comes at expense of: • Basis risk (payout does not always coincide with loss) • Difficult to understand (depends on unseen mm, not actual loss) • Also still: • Trust • Liquidity and budget constraints: insurance is still costly • Cost is worth it if insurance provision results in change in behavior that increases average income of farmers • This could be applying fertilizer, or growing more high-risk but high-return crops 7
    • What we do: summary of methodology • Randomized trials help test impact, but not enough take-up to allow to test impact on behavior (Giné and Yang 2007, Cole et al 2009) • Small scale experimental games can help to explore such impacts • We conduct a experimental game in which farmers make decisions for money in a context familiar to them (whether or not to buy fertilizer) to test hypothesis that insurance provision would enable farmers to take greater, yet more profitable risks 8
    • What we do: summary of methodology • The insurance provided was free of many problems of insurance: • Like index insurance it was based on an independent measure of how good the rains had been: no adverse selection and no moral hazard • Unlike index insurance there was no basis risk: farmers were always paid when they experienced low income • Because it was played in a game there were also no problems with trust or understanding 9
    • What we find: summary of results • Game results provide some evidence that insurance does encourage greater risk taking • more so for risk averse farmers who understood insurance well, and were already inclined to purchase fertilizer • Changes in wealth and farmers experience of fertilizer purchases and weather in previous rounds were also important determinants of fertilizer demand • Provision of real insurance will be the real test 10
    • Study site • Danicho Mukhere kebele (Silte zone) in southern Ethiopia: around 2000 households in 8 villages: • some in the highlands growing enset, barley, potatoes, cabbage • some in the mid-lands growing enset, wheat and chat • mid-land villages have better market and road access 11
    • Study site 12
    • Study site • The average head is 43 years (median), male, has farming as main occupation and little education (median: one year of education) • Nearly all farmers experienced drought in the last 10 years • farmers lose 75% of crop when rain fails (subjective estimate) • expected probability of rain failure in coming season: 0.25 • Good mechanisms to deal with illness and death (iddirs, gifts and transfers between family and friends) • But little insurance against drought: weather shocks result in loss of assets (72%) and reduced consumption (76%) much more than other shocks 13
    • Experiment protocol • Framed game that farmers could relate to their day-to-day decisions: • How much fertilizer will you buy when the return to fertilizer depends on the uncertainty of the weather? • When you have insurance how much will this change? • Farmers were given money • From 2 to 16 Birr • They then played the following game to represent production and consumption 14
    • Experiment protocol Farmer decides how much fertilizer to buy (0, 1 or 2 bags) and pays = Rainfall is drawn = 10 Birr + 2 Birr for each Farmer receives income based on rain and fertilizer = 5 Birr + 0 Birr for each Farmer pays 8 Birr for consumption REPEATED 4 TIMES 15
    • Experiment protocol • The return to fertilizer was varied (2 Birr of additional yield value for each bag or 1.25 Birr of additional yield value for each bag) • The probability of bad rain was varied (1 in 4, or 1 in 5) • To assess how insurance affects risk taking we offered insurance in the last 2 rounds of randomly selected sessions • Either at the actuarially fair price or free • Insurance paid 3 Birr if the weather was bad • Players were requested to buy insurance to increase the power of the test of the impact of insurance on behavior • Compare changes in fertilizer purchases between those were offered insurance and those that were not 16
    • Experiment protocol Farmer decides how much fertilizer to buy (0, 1 or 2 bags) and pays = Rainfall is drawn = 10 Birr + 2 Birr for each Farmer receives income based on rain and fertilizer = 5 Birr + 0 Birr for each Farmer pays 8 Birr for consumption 17
    • Experiment protocol Farmer decides how much fertilizer to buy (0, 1 or 2 bags) and buys insurance = INSURANCE Rainfall is drawn Farmer receives income = 10 Birr + 2 Birr for each based on rain and fertilizer and insurance = 5 Birr + 0 Birr for each + 3 Birr insurance payout Farmer pays 8 Birr for consumption 18
    • Experiment implementation • 241 household heads randomly selected from four villages participated in the game and undertook a household survey: • On average experiments lasted 2.5 hours and paid 27 Birr • Conducted in library in school at centre of village: large room allowed privacy… 19
    • Experiment implementation 20
    • Empirical strategy • Random selection of individuals should result in no differences between individual characteristics with and without insurance • compare changes in fertilizer purchase before (rounds 1 and 2) and after insurance (rounds 3 and 4) between insurance and no insurance sessions • Fixed effect regression: Δ fit = Δ Iit+ Δ ui where Δ fit is change in fertilizer, Iit is an indicator function denoting whether insurance was provided, ui is a time constant individual effect (i= individual, t=round of game) • Few differences in individual characteristics 21
    • Empirical strategy • But, randomization of both weather and insurance across 44 rounds resulted in some important differences in round characteristics… • In sessions in which insurance was provided • Bad weather was less likely ex-poste: changes in weather and wealth were not orthogonal to provision of insurance • Initial fertilizer purchases were higher 23
    • Differences in session characteristics All Insurance No-insurance Round T-stat sessions sessions sessions Proportion of bad weather draws 0.22 0.21 0.23 -1.66* Endowed wealth 7.5 7.7 7.4 0.6 Wealth (Birr on hand) 1 11.3 11.8 10.9 1.33 2 12.9 12.4 12.5 0.94 3 14 16.2 12 3.83*** 4 16.2 17.9 14.9 2.42** Change in wealth (Birr) 1&2 1.6 1.5 1.7 -0.29 2&3 1 2.9 -0.6 9.68*** 3&4 2.3 1.6 2.9 -6.49** Good Weather occurred 1 0.81 0.8 0.82 -0.28 2 0.72 1 0.48 11.12*** 3 0.91 0.81 1 -5.64*** 4 1 1 1 – Fertilizer purchased (bags) 1 1.55 1.71 1.42 4.03*** 2 1.63 1.79 1.5 4.13*** 3 1.55 1.79 1.34 5.46*** 4 1.71 1.79 1.65 2.17** *** p<0.01, ** p<0.05, * p<0.1 24
    • Empirical strategy • After round 2 choice: sessions being provided with insurance universally experienced good weather (ex-poste) • Wealth was much higher for individuals with insurance when insurance was provided • Change in wealth was much higher • Perceived probability of bad weather may also have been less • After round 3 choice: sessions without insurance universally experienced bad weather 25
    • Empirical strategy • Control for commensurate changes by adding Δ wealth and weather in fixed effects regression • Use nearest neighbor matching estimator (Abadie et al 2004) to compare like observations • Is their sufficient overlap (Imbens 2004)? – If outliers in control observations (round 3): artificially precise estimates – caution in interpreting results – If outliers in treatment observations (round 4): biased estimates – omit treatment observations which experienced bad weather – Fixed effects estimation should be ok given multiple observations for each individual allows a more accurate estimate of behavioral response to good and bad weather 26
    • Empirical strategy • Can we assume common time effects across each group, even after controlling for weather? • Higher initial fertilizer purchases for those offered insurance • Limited fertilizer purchases: those already purchasing 2 bags could not increase even if exposure to risk reduced • Time trend of increasing purchases (perhaps because of good weather and wealth increases): confounding effect of insurance in encouraging fertilizer purchases • Opposite effect to Eissa and Liebman (1996) in which higher labor market participation in control caused overestimation of treatment effect (Blundell and Costa-Dias 2002) • Add dummy for those at corner solution in analysis • Perform matching where matching is identical on initial fertilizer purchases 27
    • Results Before doing this we first present the unconditional changes in fertilizer purchases between those who were insured and those who were not: Difference in bags of fertilizer purchased in rounds... 1 and 3 2 and 3 1 and 4 2 and 4 Insurance 0.154* 0.157** -0.152* -0.149** (0.083) (0.073) (0.084) (0.064) Constant -0.0746 -0.157** 0.231*** 0.149*** (0.069) (0.064) (0.058) (0.054) Observations 248 248 248 248 Adjusted R2 0.009 0.013 0.009 0.016 *** p<0.01, ** p<0.05, * p<0.1 Robust standard errors in parentheses 28
    • Results Before controlling for differences in changes in weather and wealth we also present the unconditional matched results using nearest neighbor matching and matching exactly on initial fertilizer purchases (omitting treatment outliers in last 2 columns) Difference in bags of fertilizer purchased in rounds... 1 and 3 2 and 3 1 and 4 2 and 4 Nearest neighbor matching 0.273** -0.061 -0.059 -0.027 (0.113) (0.074) (0.077) (0.061) *** p<0.01, ** p<0.05, * p<0.1 Standard errors in parentheses 29
    • Results Conditional changes in fertilizer purchases between those who were insured and those who were not: Δ insurance 0.090 (0.069) 0.115* (0.060) Δ wealth -0.192*** (0.038) -0.018 (0.031) Square of Δ wealth 0.016* (0.008) 0.035*** (0.007) Good weather in previous round 1.858*** (0.26) Good weather and no fertilizer 1.27*** (0.143) Bad weather and no fertilizer 0.445* (0.266) Bad weather and fertilizer -0.568*** (0.215) Dummy for max fertilizer -1.435*** (0.0901) -1.375*** (0.0769) Constant -0.453** (0.1800) 0.565*** ( 0.1410) Observations 744 744 Number of id 248 248 Adjusted R2 0.3151 0.3601 *** p<0.01, ** p<0.05, * p<0.1, Robust standard errors in parentheses 30
    • Results • Some small impact: under insurance fertilizer purchases were 0.115 higher => average return realized by farmers increased by 5.75% • Are there groups of people for whom the provision of insurance has more of an impact? Impact of insurance High Low for people with Risk aversion 0.125** (0.0571) 0.114 (0.0728) Understanding 0.124** (0.0622) 0.080 (0.0747) C.V. of fertilizer return 0.052 (0.070) 0.174*** (0.066) • Insurance also has a higher impact for those who had previously bought fertilizer: 0.125** versus 0.116 31
    • Conclusion and further work • Conducted a small-scale framed field experiment to explore the impact of insurance • Some evidence that insurance may slightly increase investment in higher return riskier activities. Impact higher for those who: • Understood insurance better • Were more risk averse • Faced a lower C.V. of return • Were inclined to purchase fertilizer in reality • Purchases strongly dependent on weather in previous round: • Partially from changes in wealth, partially from changes in perceptions of cost and benefits => implications for how we understand peoples perceptions of weather and returns to technologies 32
    • Conclusions and further work • Promising new product: • Cheaper to administer, also no moral hazard or adverse selection • But this comes at expense of: • Basis risk, difficult to understand (depends on unseen mm, not actual loss) • Also still: • Trust • Liquidity and budget constraints: insurance is still costly • Cost is worth it if insurance provision results in change in behavior that increases average income of farmers This study provides some evidence insurance increases investment in risky inputs 33
    • Conclusions and further work • How to limit basis risk? Very important – Invest in automated weather stations reduces basis risk for farmers (even if insurance difficult to price) – Combine index with assessment of loss: Nyala’s contracts with seed farmers, research by Daniel Clarke (but moral hazard must not creep back in), future work with Nyala and WFP • How to improve understanding and trust? Index products can be complicated – Does offering simple contracts help? (IFPRI: research on very simple contracts insuring rainfall directly not weighted rainfall) – Work with those used to thinking about insurance—e.g. iddir leaders—who can buy on behalf of others? (Ongoing work by Oxford and IFPRI) • How to address liquidity constraints? – Offering insurance products through iddirs, which have some liquidity 34