Adverse Selection in Micro Health Insurance file1/21 Adverse Selection in Micro Health Insurance: ... USD 1 = approx. PKR 100, client and spouse alwayscovered under mandatory scheme

Torben Fischer, Markus Frölich, Andreas Landmann 12th International Microinsurance Conference Colombo (Sri Lanka), 17.11.2016

1/21

Adverse Selection in Micro Health Insurance:Evidence from a RCT in Pakistan

Torben Fischer Markus Frölich Andreas LandmannUniversity of Mannheim University of Mannheim and Paris School of Economics,

Center for Evaluation & Development J-PAL Europe and Center for Evaluation & Development

12th International Microinsurance Conference

Colombo (Sri Lanka), 17.11.2016

Presenter

Presentation Notes

departure


2/21

Challenges of Microinsurance

Obstacles specific to low and middle income countries:

- Limited client awareness/demand Cole et al. (2013), Dercon et al. (2011)

requirement for simple products

- Administration costs need to be low Staff qualification limitedManaging different contract types difficult Client risk classification difficult Brau et al. (2011)

Schemes vulnerable to adverse selection

Focus here: Adverse selection in Micro Health Insurance (MHI)

Motivation Setting Demand Adverse Selection Conclusion


3/21

Research Agenda: Adverse Selection in MHI

1. How to identify adverse selection (AS) empirically?

2. Is there evidence for AS in simple pooling contracts?

3. Can contract design (risk pooling) mitigate presence of AS?

4. What are the implied welfare costs of AS?



4/21

Source: 1World Bank Indicators 2014 | 2World Bank (2012) | 3Heltberg and Lund (2009)

Pakistan – in a nutshell


Key Indicators (2015)1

- Population: 189 Mio. (61% rural)

- GDP/capita: $1429 (low-middle inc.)

- Out-of-Pocket Expend.: 87%

Social Protection

- Formal Insurance < 2%2

- Free public facilities, but strongpreference for private care

- Health events are the largestsource of financial risk3


5/21

- Collaborate with National Rural Support Programme (NRSP)- largest rural support program in Pakistan (> 2.5 Mio. HHs)

- Status Quo: mandatory hospitalization insurance- covers expenses of client & spouse only

- Intervention: Offer voluntary insurance for dependents

- Target Group: Credit clients in one rural district in Punjab

- Data & Timing:

Status Quo and Interventions


2014 2015 2016BSL

RolloutPhone Survey


6/21

- Research project covers 502 villages- Insurance for dependents offered in 334 villages

- Include 4283 clients from 1050 credit groups- 3433 clients with 12286 dependents attended meetings

- Average HH characteristics in “innovation villages”:- 53% of clients female, 55% of without education- 5-6 HH members (3-4 dependents)- 23’000 PKR (about 230 USD) total HH income per month- 12% of HH with hospitalization in last year

The sample



7/21

Each client offered 1 out 4 pooling contracts:

Voluntary Hospitalization Insurance for Dependents


Note: USD 1 = approx. PKR 100, client and spouse always covered under mandatory scheme


8/21

- Policies on the village level (334 villages)- Discounts on household level (4’283 households)

Unit of Randomization



9/21

Learning Objectives - Insurance Demand

1. Sensitivity of demand w.r.t. premium?

2. Insurance pattern within household?

3. Differences across policies?



10/21

Sensitivity w.r.t. Premium & Pattern within HH

Partial uptake individual insurance Price decrease of 30% doubles number of insured individuals

0.2

.4.6

.8U

ptak

e (in

%)

70 80 90 100Premium (PKR)

HH Level Dep. Level

Individual (P1)



11/21

Insurance Demand & Enforcement of Eligibility

Eligibility criteria enforced in implementation Less households, but more individuals in P3 and P4

0.2

.4.6

.8U

ptak

e (in

%)


HH Level Dep. Level

Individual (P1)

0.2

.4.6

.8U

ptak

e (in

%)

120 130 140 150Premium (PKR)

HH Level Dep. Level

Individual High (P2)

0.2

.4.6

.8U

ptak

e (in

%)


HH Level Dep. Level

Household (P3)

0.2

.4.6

.8U

ptak

e (in

%)


HH Level Dep. Level

Group (P4)



12/21

Note: Share of individuals with outpatient treatment in last month

Adverse Selection – Using Baseline Health Data

Implement a positive correlation test (here with outpatient history)



13/21

20

40

60

80

100

120

140

Exp

ecte

d C

ost I

ndex

(PK

R)

Not Insured Insured

Individual (P1), N=2979

Note: Share of individuals with outpatient treatment in last month

Adverse Selection – Predict Costs using Ex Post Health Events

Use more indicators and translate into expected costs



14/21

Positive Correlation Test Across Products

Positive correlation b/w health index and insurancestatus mainly in P1 and P2



15/21

Using Predicted Costs & Price Variation

020

4060

8010

012

014

016

018

020

0Ex

pect

ed C

ost I

ndex

(PKR

)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1% Insured

P1: Insured



16/21

020

4060

8010

012

014

016

018

020

0Ex

pect

ed C

ost I

ndex

(PKR

)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1% Insured

P1: Insured

020

4060

8010

012

014

016

018

020

0Ex

pect

ed C

ost I

ndex

(PKR

)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1% Insured

P2: Insured

Individual Policies




17/21

020

4060

8010

012

014

016

018

020

0Ex

pect

ed C

ost I

ndex

(PKR

)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1% Insured

P3: Insured

020

4060

8010

012

014

016

018

020

0Ex

pect

ed C

ost I

ndex

(PKR

)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1% Insured

P4: Insured

Household Policies




18/21

0

40

80

120

160

Prem

ium

, Exp

ecte

d C

ost I

ndex

(PK

R)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Share Insured

Demand Average Costs Marginal Costs

P3

Welfare Analysis: Bundled Policies

Use expected cost points to estimate average cost curve Add demand curve

Household Policy Group Policy

0

40

80

120

160

Prem

ium

, Exp

ecte

d C

ost I

ndex

(PK

R)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Share Insured

Demand (lfit) Average Costs (lfit) Marginal Costs (lfit)

P4



19/21

Individual Policy Individual High Policy

Welfare Analysis: Individual Policies

0

40

80

120

160

Prem

ium

, Exp

ecte

d C

ost I

ndex

(PK

R)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Share Insured

Demand (lfit) Average Costs (lfit) Marginal Costs (lfit)

P2

0

40

80

120

160

Prem

ium

, Exp

ecte

d C

ost I

ndex

(PK

R)

0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1Share Insured

Demand Average Costs Marginal Costs

P1



20/21

Welfare Analysis: Equilibrium and Efficient Allocations


00,10,20,30,40,50,60,70,80,9

1

Individual P1Individual High P2Household P3Group P4

0

20

40

60

80

100

120

0

5

10

15

20

25

30

Equilibrium Equilibrium EquilibriumQuantity Price Welfare


21/21

Rigorous design of pilot allowed considerable learning:

• Demand exists in high-need setting• Substantial AS in individual hospitalization insurance policies• Risk bundling on higher levels mitigates AS• Potentially higher ‘equilibrium welfare’ in bundled products

Conclusion



22/21

Thank youfor your attention

Documents

Adverse Selection in Micro Health Insurance file1/21 Adverse Selection in Micro Health Insurance: ... USD 1 = approx. PKR 100, client and spouse alwayscovered under mandatory scheme