The Subjective Well-being Effects of Imperfect Insurance that Doesn’t Pay out

Kibrom HirfrfotChristopher Barrett

Erin LentzBirhanu Taddesse

Cornell UniversityJuly 28, 2014

• Most people are risk averse (Rosenzweig & Binswanger, 1993)

• In theory, actuarially fair insurance improves welfare regardless of payout. • Insurance usually not actuarially fair – loading.• But, there is still demand – offers peace of mind.

• Assessing welfare gains in expected benefit terms is misleading.• There is little evidence that insurance generates welfare gains.

• Low uptake of index insurance suggest no welfare gains from insurance (Gine et al. 2008; Cole et al. 2013; Binswanger-Mkhize, 2012)

• However, Jensen et al. (2014) find index insurance produces net utility benefit in Kenya.

• Yet if one believes in DARA, the poor stand to gain the most from insurance.

• We explore whether an imperfect insurance improves subjective well-being (SWB) for a poor, rural population.• Novel approach: we exploit panel nature of data and no indemnity payout to test

buyers’ remorse.

Does index insurance lead to welfare gains?

• We use data from eight woredas in the Borana zone of southern Ethiopia.• Mobile pastoralism is main means of sustenance.

Study setting and index insurance

• Cyclical movement in search of forage and water is common.

• Indigenous insurance institutions have declined due to recurrent drought.

• Index based livestock insurance (IBLI) was introduced in August 2012.• Uses NDVI to identify premiums and makes indemnity payout based on observable,

exogenous index of rangeland conditions for each woreda.• Payout is made when vegetation falls below a contractually stipulated threshold.

• There are four season in Borana: long rainy, long dry, short rainy, short dry.• IBLI contracts are sold in two sales periods before the rainy seasons.

l

Study setting and index insurance (cont’d)

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb

1 year contract coverage

Whether LRLD NDVI falls below the contract trigger value is announced

Whether SRSD NDVI falls below the contract trigger value is announced

Sale period for LRLD Period of NDVI observation for LRLD season

Sale period for SRSD

Period of NDVI observation for SRSD season

LRLD season coverage SRSD season coverage

• Three rounds of data have been collected

• A random sub-sample of respondents received discount coupons. Others received information extension. Some received both.• Information extension: Poem tape and Comic books

• To date, there have been no indemnity payouts.• Households who bought IBLI are materially worse off, but could be better

off in broader well-being terms. Are they?

Data

March 2012 Baseline 516 hh

August-September 2012 Sales period

January-February 2013 Sales period

March 2013 R2 Survey 508 hh

August-September 2012 Sales period

January-February 2013 Sales period

March 2014 R3 Survey 514 hh

• IBLI purchase (“opt-in”) decisions are likely to be endogenous.• Opt-in decision is likely to depend on subjective assessment of risk.• We use IBLI’s randomized design features to deal with endogeneity.

• Two-step estimation strategy:• First we estimate an uptake (selection) equation as

• All treatments are randomly assigned - can only be correlated with SWB through IBLI uptake.

• We control for a series of covariates that may affect IBLI uptake – herd size and income, gender, age and educational attainment of household head, food aid receipt, household composition, reera fixed effects.

Estimation strategy

• In stage two we estimate the SWB equation as

• There are at least two ways IBLI can influence SWB• Non-monetary (psychological) benefits or costs

• Insurance may reduce stress about adverse outcomes (give peace of mind): (• Insurance could increase stress if basis risk is high, and it’s like lottery: (• Buyer’s remorse:

• 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 .

Estimation strategy (cont’d)

• We therefore estimate the aggregate effect of IBLI on SWB as:

• Where is the TLU equivalent of wealth gained or lost due to IBLI purchase: • Since there wasn’t any payout to date, our estimates are lower

bound of the effects of IBLI on SWB.• In SWB terms the actuarially fair premium is given as:

Estimation strategy (cont’d)

• Ordered logit regression estimates

Results (cont’d)

  Model Model ModelDependent variable: SWB (1) (2) (3) Predicted IBLI uptake 0.753*** 0.601*** 0.828*** (0.251) (0.199) (0.290)Number of TLU insured 0.010 -0.004 -0.004 (0.012) (0.008) (0.009)Purchased IBLI in R2 but not in R3 -0.528*** -0.516*** -0.260* (0.167) (0.138) (0.145)Annual income (‘000 Birr) 0.002 0.002 (0.005) (0.004)Number of TLU owned 0.030*** 0.030*** (0.006) (0.007)Asset Index 0.238*** 0.224*** (0.069) (0.066)Controls No No YesReera fixed effect No No YesObservations 1,529 1,529 1,529Number of groups (households) 550 550 550

Controls : Household head gender, age, age squared, schooling, and household size.

Standard errors clustered at the reera level in parentheses*** p<0.01, ** p<0.05, * p<0.1

• In all 3 models, IBLI has a positive and significant effect on SWB. •Our result is robust to

inclusion of income, wealth, and var. controls.•We find significant buyer’s

remorse effect.• The positive piece of mind

effect is significantly greater than the negative buyer’s remorse.

• Aggregate effect of IBLI on SWB and actuarially fair premium rates

Standard errors clustered at the reera level in parentheses*** p<0.01, ** p<0.05, * p<0.1

• IBLI has a positive and significant aggregate effect on SWB. The point estimate suggests insuring a TLU increases SWB by 0.205.• Insuring 5 TLU moves one from “very bad” to “bad” or “good” to “very

good” on the SWB ladder.• Average TLU owned≈20. Our result suggests full coverage would lift one

from lowest to highest SWB category.

Results (cont’d)

Variables: Coef.Change in SWB 0.205***

(0.072)Actuarially fair premium 51.480

(109.527)

• While we find statistically significant evidence of buyer’s remorse, its magnitude is considerably smaller than the positive effect of having insurance coverage.

• Even an imperfect insurance product that does not pay out can leave a poor rural population better off. • Given IBLI is yet to payout in Borana, the actuarially fair premium for the

sample period was roughly 51 Birr, significantly less than the average premium per TLU ( Birr 698 in R2 and Birr 575 in R3)

• IBLI was welfare enhancing even at a premium rate that was (temporarily and ex post only) an order of magnitude higher than the actuarially fair rate

Conclusions: Insurance, even when it does not pay out, leads to welfare gains

Thank You

• Prior to the August-September 2013 and January-February 2014 sales periods only discount coupons were distributed to 408 hh each.

Treatments

August-September 2012 January-February 2013Discount coupon

Comic book

Poet tape None

Discount coupon

Comic book

Poet tape None

Disc. coupon 412 411(79.84) (79.81)

Comic book 86 108 77 99(16.67) (20.93) (14.95) (19.22)

Poet tape  66 0 86 55 0 71(12.79) (16.67) (10.68) (13.79)

None 62 66(12.02) (12.82)

Sample 516 515

• LPM estimates of IBLI uptake

LPM results

  Model ModelDependent variable: IBLI uptake (1) (2)Discount: SP1 only 0.259*** 0.256*** (0.046) (0.047)Discount: SP2 only 0.267*** 0.270*** (0.046) (0.046)Discount: SP1 & SP2 0.206*** 0.203*** (0.047) (0.046)Value of discount (%) SP1 0.193*** 0.196*** (0.060) (0.059)Poet tape: SP2 only 0.193*** 0.183*** (0.061) (0.057)Comic book: SP2 & SP2 0.164** 0.169** (0.071) (0.069)IBLI premium: SP2 0.169* 0.134 (0.087) (0.119)Iqub membership -0.086** (0.039)Controls No YesReera fixed effect No YesR-squared 0.406 0.418Observations 1,535 1,532Number of unique observations 551 551

Standard errors clustered at the panel round and reera level in parentheses*** p<0.01, ** p<0.05, * p<0.1

Controls : Annual income, TLU owned, expected TLU loss, value of non-livestock assets, household demographic variables, household composition, group membership, and religion dummies

•Randomized instruments have significant effect on IBLI uptake.•Receiving discount coupon and

amount of discount have strong effect.• Income has negative effect on IBLI

uptake. Self insurance?• Sargan & Basmann

overidentification tests confirm instruments are valid. •Wald test rejects null of joint

insignificance of instruments.

•Randomized instruments of discount coupons and information extension (comic books and poet tapes) have significant effect on IBLI uptake.

• Receiving discount coupon and amount of discount have strong effect.• Income has negative effect on IBLI uptake. Self insurance?

•Sargan & Basmann overidentification tests confirm instruments are valid. •Wald test rejects null of joint insignificance of instruments.

• Controls : Annual income, TLU owned, expected TLU loss, value of non-livestock assets, household demographic variables, household composition, group membership, and religion dummies

Estimation results of IBLI uptake