Health Insurance Schemes in India: An Economic Analysis of Demand

Management under Risk Pooling and Adverse Selection

A Thesis Submitted to the University of Mangalore for Award of

the Degree of Doctor of Philosophy in Economics

By

Sukumar Vellakkal

Research Supervision by

Gopal K. Kadekodi Professor and Former Director, ISEC

INSTITUTE FOR SOCIAL AND ECONOMIC CHANGE NAGARABHA VI, BANGALORE-560072, INDIA

December 2007

This small piece of work is dedicated

To

My Beloved Wife Remya

&

To

The illiterate landless agricultural laborers of Indian land

through my beloved parents

• INSTITUTE FOR SOCIAL AND ECONOMIC CHANGE

Nagarbhavi po: BANGALORE-560 072

DECLARA TION

I hereby declare that the present thesis titled 'Health Insurance

Schemes in India: An Economic Analysis of Demand Management

under Risk Pooling and Adverse Selection' is a result of the original

research undertaken and carried out by me under the guidance and

supervision of Prof. Gopal K. Kadekodi, Professor and Former

Director of Institute for Social and Economic Change (ISEC),

Bangalore.

I have properly acknowledged the sources from which I may have

borrowed ideas. I declare that the material of the thesis has not

formed, in any manner, the basis for awarding of any Degree or

Diploma previously of University of Mangalore or any other

University.

~ Date: 31 sl December 2007 Sukumar Vellakkal

(Ph. D. Fellow)

INSTITUTE FOR SOCIAL AND ECONOMIC CHANGE Nagarbhavi po: BANGALORE-560 072

CERTIFICA TE

This is to certify that thesis entitled' Health Insurance Schemes

in India: An Economic Analysis of Demand Management under Risk

Pooling and Adverse Selection' submitted by Mr. Sukumar Vellakkal

for the award of the degree of Doctor of Philosophy in Economics is

based on the candidate's own research work under my guidance and

supervision Juring the period of the study.

It has not been previously formed the basis for the award of any

Degree/Diploma/ Associateship/Fellowship or other similar titles to

any candidate.

Place: Bangalore Date: 31- 1:2- - .2001-

~ .. ,.~ ... , ~,\... . . ~

/"':(,r~'-' ~...., ,", r '\

I / '\ ",

: ',I BAI\CALORE: )':. , . , .... '/ . . I J : ';/ \. , J / \ -. \.. ' /

~~ '-- " '-'.....,.,.". I {-"-",.,. ,

" - [:flC • -' • .;;;;s'_.~ . , _ ". -....o;;,.~. _

Gopal (Ph. D. Supervisor)

Acknowledgement

I sincerely express my heartfelt deep sense of gratitude to Professor

Gopal K.Kadekodi, my mentor and supervisor for the thesis. It is a great privilege to

complete this thesis under his guidance and supervision. He was kind enough to find

sufficient time to guide me although his days were busiest as the Director of ISEC.

His way of guiding me in this thesis work was unique; it gave me enough confidence

to think in depth on my research topic. His moral support, constant encouragement,

and above all, his love and affection have been an instant tonic for me to carry out

this research work. I also express heartfelt gratitude to Mrs. Savitha Kadekodi for

her love and affection, and moral support. I am grateful to Ms. Kamala Aunty for all

her moral support throughout.

I would like to take this opportunity to express my deep sense of

gratitude to Dr. K K. Hari Kurup, Lecturer, Govt College, Kasargod. It was he who

introduced me to the research world and guided me to get admission to ISEC. His

moral support, guidance and constant encouragement are invaluable to me, I am

short of words. I express deep sense of gratitude to Mrs. Deepa Kurup. And also,

respecful thanks and gratitude to Narayanettan.

I cherish the memory of late. Mr.Chandrasekharan, Lecturer, Govt

College, Kasargod. His immature death was a big shock and loss to me. I greatly

acknowledge his support and concern towards me.

I have greatly benefited from my academic association with Prof. David M

Dror, Erasmus University, Netherlands. I use this opportunity to express my deep

sense of gratitude to him. His methods and approaches to research have influenced

me a lot.

I am fortunate enough to work with Prof. Ruth Koren, Tel Aviv University,

Israel. I have learned a lot from her, I express my gratitude to her.

My respectful thanks to Dr. Marion Danis, NIH, USA, for guiding me in

the CHAT tool since for a long time. I greatly acknowledge her moral support and

affection rendered towards me.

I am fortunate enough to meet an eminent personality, a good

academician with a great heart: Prof. J F Wen, University of Calgary, Canada. I have

benefited a lot from the discussion with him. I am fond of Prof. J F Wen and his wife

Gabrille who made my stay at Canada during my PhD works a wonderful one.

I thank Prof. Anil Gumber, senior faculty, Warwick University, UK for his

guidance through out my PhD research. He was kind enough to timely respond to my

queries and also showing much interest on my research work. I use this opportunity

to express my heartfelt gratitude to him.

It was great to learn from Prof. Shashanka Bhide, Professor, NCAER and

former RBI Chair Professor, ISEC. I greatly recall his advice, suggestions and

comments as a Doctoral Committee member of my PhD research.

Back to ISEC, the comments and suggestions of the doctoral committee

members of my PhD research were very constructive and useful for me. lowe to

Prof. Madheswaran for his academic and personal support. Special thanks to Prof.

KNM Raju, former professor of PRC unit, ISEC.

The critical and constructive comments and suggestions of the panel

members at various bi-annual seminars in the institute were very helpful in bringing

the thesis to the present shape. Here, I specially thank Prof James, Head, PRC unit

at ISEC, Prof.lndrani Gupta, Institute of Economic Growth, New Delhi, Prof.

Rajashekhar, Head, Center of Decentralization at ISEC, Dr. Mathiyazhakan and Dr.

Gayathri.

I am grateful to ISEC for selecting me for the PhD programme. I thank the

ISEC fraternity for the support throughout the period.

I am very much fond of Prof. Govinda Rao, the then Director of ISEC. His

ideology has influenced me a lot. I greatly acknowledge his advice and personal

support rendered towards me, it really helped in my PhD research.

The faculty members of the institute were high co-operative and supportive.

My Whole hearted thanks to Prof. K N Ninan, Prof M R Narayana, Prof. Sangeetha,

Dr. Venkatachalam, Prof.Meenakshi Rajeev, Prof. Usha Devi, Dr. Gaythri Devi, Dr. V

P Vani, Dr. G S Shastri, Dr. TV Sekhar, Dr. Madhusree Sekhar, Dr. Sivakami, Prof. R

S Deshpande.

I have greatly benefited from the discussion with Prof D Narayana of CDS, my

respectful thanks to him.

I gratefully recall my teachers during my MA programme at Govt.College,

Kasargod: Prof. Joseph Lopez and Prof. S N Holla. I express my deep sense of

gratitude to them for all guidance and encouragements.

I enjoyed the friendship of Subodh, who made my stay at ISEC very

interesting and lively.

I heartfelt respect and thanks to Dr. Jyothis and Dr. Jeena Jyothis for their

inspirational academic and personal support throughout the period.

I greatly acknowledge the friendship and academic association of Ms. Erikka,

Erasmus University, Netherlands and Alex, University of Cologne, Germany. It was

wonderful to work with them, their collaboration made the CHAT exercise conducted

in various Indian villages and slums very interesting. I also thank Mr. Ralf

Rademacher for his constructive suggestions and comments on CHAT tool.

My special thanks are due to Mrs. Olga and Mr. Hugo for their personal and

moral support through out. The support they extended towards me made my stay in

the Netherlands very much comfortable.

Thanks are due to Mr. K S Narayana, AR (Academic) for his careful and

efficient administrative help, and also for language editing of this thesis. I also thank

Mrs. Margaratte, Accounts section, Mrs. santha, Reception and Mr. srinavasamurthy,

Director Office, IsEC.

I sincerely acknowledge the help and assistance received from Mr. Krishna

Chandran, Mr. Satish Kamath in the computer center and members of Library staff of

ISEC, especially Mr.Kalyanappa.

I take this opportunity to thank my friends at IsEC: Bikas, Badri, Poulomi,

Nisha, Anitha, Somasekar, Bhanumurthy, Ashish Das, Emil, santhosh, Anand Vadi,

Pratheeba, Venu, sathyasiba, Durba,Sabu], Biplab, Manojit, Avinandan, Rajdeep,

Akshay, Jaganath, Sitakantha Sethi, Rishi, Dukhabandha sahoo, Geethu, Pattu,

Sarbhani, Gnadhari, Kalid Wasim, Malini, Smitha, Tunga, Subir, Sachi, Nithin,

Yogeswari, Kannan.

Discussion with my friend Mahesh was very productive; I greatly acknowledge

his support at various stages of my PhD research.

Thanks also to Prashobh for his personal and academic help at various stages

of my PhD work. My special thanks to Naveen, Anantha and Srikant for giving a nice

friendship at IsEC. The friendship of Mainak Majumdar, Lija, sunitha, sarala and

Binitha helped me to ease the pressure and difficulties at various stages of this work,

special thanks to them.

I also thank my friends at CDS: Rajesh Puliyara, Shy jan, rajesh Kommath,

Anil, Abdul, Achan, Hari, syam, Subratho, Harilal. My special thanks to Nirmal Roy,

Krishna and kunhikrishnan for extending their help in data collection and also for the

nice friendship throughout.

I greatly acknowledge the support extended to me by Mr. Sanjeev, PhD

fellow, CMDR Dharward.

I greatly acknowledge the fellowship given by ICSSR for my PhD research. I

also acknowledge SIC!, New Delhi for awarding me the fellowship and giving me an

opportunity to do research in Canada. I thank the staff of SIC!. Further, I thank the

staff of University of Calgary for their support during my stay at Canada. The

friendship of Abdu, Blake, Omar and Julia and John made my life more comfortable

there.

I express my deep sense of gratitude to faculty members of Institute of

Health policy and Management, University of Erasmus, Netherlands, for formally

teaching me the essence of Health economics.

The support received from University of Mangalore is great; I wish to thank

Prof. Joshi and Dr. Jayasheela for their kind help constant encouragement. I also use

this opportunity to express my sincere thanks to Ms. Soni, Ph.D section, for her help

and kind co-operation.

Back to home, I cherish the memory of my father (late) who left this world

during the initial stage of my PhD research. I recall the moral support extended by

my family towards me, my deep sense of gratitude to them: Mohan, Vijayan,

Sureshan, Rameshan, Sara, Sumathi, Divya, Vineetha, Sreeja, Kunhi krishnan,

Nisha, Babu. I also express my gratitude to my Father-in-law and mother-in-law.

And also, special thanks to Renjith and Reshmi.

Last but not least, Remya, my beloved, for all she is to me, I am short of words.

Chapter 1 1.1 1.2

1.3

1.4 1.5 1.6 1.7 1.8

1.3.1 1.3.2 1.3.3

Chapter 2 2.1 2.2 2.3 2.4

2.4.1 2.4.2

2.5 2.6 2.7

2.7.1 2.7.1.1 2.7.1.2 2.7.1.3

2.7.2

Chapter 3

3.1 3.2

3.3 3.4

TABLE OF CONTENTS

Declaration Certificate Acknowledgement Abbreviations List Tables List of Figures

Introduction Motivation for the study Relevance of Health insurance Schemes for India

Taxonomy of Health Insurance in India Public (social) Health Insurance Schemes Micro Health Insurance (MHI) Schemes The Private Health Insurance (PHI) schemes Research Problems and Questions Objecti ves of the stud y Main Research Hypotheses Scope of the study Organisation of Thesis

Concepts, Review of Literature and Methodology Introduction Health Insurance: Basic concepts and principles Health Insurance Market Market Failures in Health Insurance Market Selection Bias Moral Hazard Demand for Health Insurance Some selected study on Health insurance schemes in India Data sources and Methodology of the Study Data sources Primary Data on PHIs ECCP Household data on MHIUs Primary data on Clients Preferences on Health Insurance Benefits (Choosing Healthplans All Together (CHAT-I)) Methodology of present research

Equity Aspects of the Health Insurance Coverage in India Introduction Equity in Health Care and Equity in Health Insurance Coverage Health Insurance Schemes and their target population Inter-income class distribution of health insurance coverage

Page No. 1-14 1 3

7 7 7 8 10 12 12 13 14

15-48 15 15 18 21 21 24 29 34 36 36 37 40 41

42

49-

49 49

52 54

3.5 3.6

3.7 3.8

Chapter 4 4.1 4.2 4.3 4.4 4.5

4.6 4.7

ChapterS

5.1 5.2 5.3

5.4

5.5

5.6

5.7 5.8

5.3.1 5.3.2

5.5.1

Chapter 6 6. 1 6.2 6. 3 6.4

Chapter 7

7.I 7.2 7.3

7.4

Intra-income class analysis of health insurance coverage Econometric Estimation on the probability to have MHI coverage for various income class households Premium Burden on Households Chapter Summary

Factors Determining Micro Health Insurance Coverage Introduction Factors Determining Health Insurance Coverage Educational Profile Household Size Health Insurance Coverage and the role of Self Help Groups (SHGs) Econometric Estimation Chapter Summary

Information Asymmetry, Market Failure and the Health Insurance Coverage Introduction Conceptual and theoretical frame Familiarity of different aspects of Insurance Awareness about the Insurance System Role of 'Insurance Habit' Asymmetric information and Information Dissemination Channel on Health Insurance Coverage

A model of insurance agent's rational choice Insurance Agent and Selection Bias Empirical estimation on the presence of adverse selection: Significance of Health Risk Econometric Estimation Chapter Summary

Selection Bias in Micro Health Insurance Schemes Introduction Adverse selection in MHI schemes Role of SHGs in Adverse Selection Chapter Summary

Preferences for Health Insurance Benefits and Health Insurance Schemes Introduction Analytical Aspects Preferences of the people for different health care benefits without budget constraint Preferences of the people for different health care benefits with budget constraint

57 60

66 69

67·92 71 71 74 77 79

85 92

93·119

91 94 98 98 102 104

106 109 113

liS 119

120·132 120 120 130 131

133·153

133 133 13S

138

7.4.1 CHA T- Decision tool to elicit people's preferences for health care benefits

I Design of the decision exercise 2 Definitions of the various benefit types 3 Determination of Actuarial Costs 4 Selection of the Insurance Premium 5 Survey Population

7.4.2

7.4.3

7.5 7.6 7.6.1

7.6.l.i

7.6.I.ii 7.6.l.iii 7.6.l.iv 7.6.2 7.6.2.i

7.7 Appendix I

Chapter 8 8.1

8.2 8.3 8.4 8.5 8.6 8.7 8.8

Appendix 2

Preferences for various health Insurance benefits at Individual level Choice of Benefits: Some Qualitative Insights during the CHAT Experiment Preferred HI Package and the prevailing HI schemes Discussion Benefits of primary importance Health Insurance Package 1: Benefit package I: OP(b)+IP(b)+ T(b)+D(b) Health Insurance Package 2: IP(b)+ T(b )+D(b) Health Insurance Package 3: OP(b)+T(b)+D(b) Health Insurance Package 4: OP(b)+IP(b)+D(b) Benefits of secondary importance Health Insurance Package 1: Preventives care (P) + Indirect Cost (IC) + Medical Equipment (ME) + Dental care (DC) + Mental care (M) Chapter Summary CHAT Materials

Summary, Policy Suggestions and Conclusions Introduction Main Objectives

Main Research Hypotheses

Data Sources and Methodology Summary of the Main Chapters Main Findings Policy Implications

Scope and limitations of the present study and suggestions for future research

Questionnaires used for the present study 1- Questionnaire for Household Survey, ECCP Project 2- Interview schedule for the voluntarily insured people 3- Interview schedule for the uninsured people in the locations of PHI insured References

138

132 140 141 136 142 142

143

146 148 148 148

ISO ISO lSI lSI 151

153 154-160

161-178 161 162

162

162 166 172 174

177

179-200

179 186 195

201-212

ADB

CGHS

CHAT

ESIS

ECCP

GDP

GIC

HI

IRDA

MHIUs

LlC

MHls

NIC

NCAER

NSSO

OOPS

NIAC

PRIs

PHI

Rural MHls

Urban MHls

SEWA

SHGs

OIC

UIIC

UHI

WHO

WHR

WDR

WTP

Abbreviations

Asian Development Bank

Central Government Health Scheme

Choosing Healthplans All Together

Employees State Insurance Scheme

European Union Cross Cultural Program

Gross Domestic Product

General Insurance Corporation

Health insurance

Insurance Regulatory and Development Authority

Micro Health Insurance Units

Life Insurance Corporation

Micro Health Insurance schemes

National Insurance Corporation

National Council for Applied Economic Research

National Sample Survey Organisation

Out Of Pocket Spending

New India Assurance Company Ltd

Panchayat Raj Institutions

Private Health Insurance

Rural Micro Health Insurance schemes

Urban Micro Health Insurance schemes

Self Employed Women's Association

Self Help Groups

Oriental Insurance Company Ltd

United India Insurance Company Ltd

Universal Health insurance

World Health Organisation

World Health Report

World Development Report

Willingness To Pay

Table No

1.1

l.2

1.3

2.1

2.2

2.3

2.4

2.5

3.1

3.2

3.3

3.4

3.5

3.6

3.7

3.8

3.9

3.10

LIST OF TABLES

Title

Locations and size of membership of some selected MHI

Units

Private Health Insurance sector in India

Selected health insurance coverage in India

A summary of different studies about selection bias

A summary of the reasons for having health insurance in

Ireland

Number of Households from the locations of Micro Health

Insurance Units

Locations of the Study Population

Classification of households in to various income classes and

cut-off pOints

Proportion of the households across different income classes

Proportion of the households across various income classes

in each 'Rural MHI' schemes

Proportion of the households across various income classes

in each 'Urban MHI' scheme

Proportion of the Insured and Non Insured households

across different income classes in Rural MHI. Urban MHI

and PHI schemes

Proportion of the Insured and Non Insured households in

various MHI schemes across different income classes

The mean value of the 'HI ratio' across different income

classes

Definition of variables

Probability to have health insurance coverage- Probit model

results of Rural MHI schemes

Probability to have health insurance coverage- Probit model

results of the Urban MHI schemes

Percapita health insurance premium (in Rs.) paid by different

Page No.

8

9

10

23

32

41

42

45

54

56

56

58

58

59

61

63

64

67

3.11

4.1

4.2

4.3

4.4

income classes

Mediclaim policy premium (in Rs.)

Highest educational qualification among the Insured and

Non Insured households (%)

Mean value of the household size

Defmition and measurement of variables

Probability to have health insurance coverage- Probit model

results of Rural MHI schemes

68

75

77

85

86

4.5 Probability to have health insurance coverage- Probit model 89

results of Urban MHI schemes

4.6 Probability to have health insurance coverage: Marginal 90

effects of the pro bit model of the selected variables

5.1 Knowledge about Insurance providers by the Insured and Non 99

Insured (%)

5.2 Knowledge about some selected insurance products other 100

than health insurance schemes (%).

5.3

5.4

5.5

Knowledge about different types of health insurance products

(%)

Knowledge about health insurance poliCies other than

Mediclaim Policy at the time of joining- insured people (%)

'Other Insurance Enrolment Status' of the Insured and Non

Insured (%)

101

102

104

5.6 Main source of information on health insurance (Mediclaim 104

Policy) scheme for both Insured and Non Insured (%)

5.7 Definitions of variables 116

5.8 Probability to have health insurance coverage- Probit model 117

results of PHI schemes

6.1 Households reporting bad health or bad medical situation 121

(high risk) at least one among the members in the household

in MHI schemes (%)

6.2 Probit model results specific to the probability of the high risk 121

to have health insurance

6.3 Definition and measurement of variables 123

6.4 Probability to have health insurance coverage- Marginal effect 124

of Probit model estimate of Rural MHI schemes

6.5 Probability to have health insurance coverage- Marginal effect 125

of Probit model estimate of Urban MHI schemes

6.6 Probability to have health insurance coverage in MHI 129

scheme- Marginal effect of Probit model estimate of some

selected parameters

7.1

7.2

7.3

7.4

7.5

7.6

7.7

7.8

7.9

7.10

Preferences of the PartiCipants at the individual level (%)

Ranking of preferences and cost of insurance benefits

Preferred Health Insurance Packages

Existing Health Insurance Schemes and Insurers in India, as

on August 2007

Benefits Provided by Critical Illness policy

'Benefits of Secondary Importance' covered by General

Insurers

Sticker cost* of benefits in CHAT exercise

Characteristics of Study Participants in the CHAT exercise

Benefits offered in the CHAT Exercise

Status of different health insurance benefits of both MHI

Units and PHI providers

143

145

146

147

149

151

154

154

156

159

Figure No 2. 1 3.1

3.2 3.3 4.1 4.2

4.3

4.4

4.5

4.6 4.7-a

4.7-b

4.7-c

4.8-a

4.8-b

4.8.c

LIST OF FIGURES

Title page

Classification of households in to different income classes 44 Impact of Equity in health insurance coverage on Equity in 52 Health Target Population of the Health Insurance Schemes 53 Mediclaim premium for various age groups 69 Determinants of Micro Health Insurance Coverage 73 Proportion of the educational qualification among the Insured 76 and the Non Insured in the 'Rural MHI' schemes Proportion of educational qualification among the Insured 76 and the Non Insured in the 'Urban MHI' schemes Household size across different income classes in the case of 78 Rural MHI schemes Household size across different income classes in the case of 79 'Urban MHI' schemes Micro Health Insurance Model 80 The SHGs membership status of the Insured households of 81 the 'Rural MHI' schemes The SHGs membership status of the Non Insured households 81 of the 'Rural MHI' schemes Proportion of Insured and Non Insured among the SHG 82 members in Rural MHI Schemes The SHGs membership status of the Insured households of 83 the 'Urban MHI' schemes The SHGs membership status of the Non Insured households 84 of the 'Urban MHI' schemes Proportion of Insured and Non Insured among the SHG members in 'Urban MHI' Schemes 84

4.9 Probability to have health Insurance coverage for each income 91 with SHG membership

5. 1 PHI model (Partner-Agent model) 105 5.2 Households reporting bad health or bad medical situation at 115

least one among the members in the household in PHI scheme (%j

7.1 Preferences for various health care benefits among those who 136 are willing to pay for health insurance (N=2390j

7.2 Preferences for various health care benefits among those who 137 are willing to pay for health insurance (N=2390j

7.3 CHAT Board 139

CHAPTER ONE

INTRODUCTION

1.1. Motivation for the study

Improvement in health status is vital for the enhancement of human

capabilities. Illness is an important source of deterioration to human

health. Of all the risks facing poor households. health risks pose the

greatest threat to their lives and livelihoods. A health shock adds health

expenditures to the burden of the poor. Even a minor health shock can

cause a major impact on poor persons' ability to work and curtail their

earning capacity. Moreover. given the strong link between health and

income at low income levels. a health shock usually affects the poor the

most (Dror and Jacquier 1999; Cohen and Sebstad 2003b).

Non-availability of necessary finances Is a major obstacle in the health

care attainments of people in many developing countries. including India.

With the continuing resource constraints of the government and

competing sectoral demands. the allocation needed in the health sector

may not increase to adequate level in the near future. Nonetheless. the

present trend of cut in government subsidies as a part of the 'new

economic reforms' is likely to put more pressure on this sector.

It is in this context that many countries are looking forward to the

alternatives to the tax based resource mobilization for health care

finanCing. 1\vo broad methods such as cost containment and cost sharing

method can be proposed as alternatives on resource mobilization for

health care (World Bank. 1987). Privatization and community

participation strategies are proposed for cost containment. Cost sharing

I

methods include User Financing and Health Insurance l • There is a

growing awareness that access to healthcare cannot be free-of-charge,

due to the low level of government spending on health, nor funded mainly

out-of-pocket by care-seekers, due to the regressive effect of this financing

mode [James et al., 2006J. Health Insurance (HI) has emerged as part of

the reform drive in many countries, both as a way of augmenting financial

resources available for care, and as a means of better linking health

demand to the provision of services (Dror and Preker. 2001). HI is

becoming a major policy preoccupation as it can provide risk management

that respects the complexity of the risks and is one of the best fmancial

tools to prevent a situation whereby people with income above the poverty

line would fall under it. Promoting HI is a rational and powerful response

as it serves the insured well even when the insurance is a very humble

local micro health scheme. as evidenced from some of the micro schemes'

increasing access to health care. significantly (Dror et al .• 2005). HI

mechanism is getting more popularity even in developing countries

backed up by the evidence from the successful experience of the

developed countries where HI system is an integral part of the health care

system. Notwithstanding the view that HI is a viable solution [Churchill

2006J. HI is nearly nonexistent among poor communities in rural India.

The HI cQverage (i.e. the number of people covered by HI) in India, in some

form or the other, i.e .. whether in public or private sphere, Is abysmally

low and is only around 3% of the total Indian Population (IRDA. 2004). At

the same time, interest in taking steps to spread insurance coverage is

growing. Private insurance companies are propagating marketing methods

and products which should enhance access to insurance among the

wealthier segments of the population. Commercial companies are also

aiming at selling insurance to people living closer to the poverty line, in

part compliance with the regulations that Impose a quota of "social" and

"rural" contracts. Community organizations and other bodies have also

contributed to the growth of health insurance, notably by supporting the

development of India's micro insurance market among the poorer

I It can be seen that PrivatizatIon. Community participation and User fees involve a kind of out-of-pocket expenditure burden on the households which Is perhaps minimized by the Introduction of Health Insurance as a risk pooling mechanism.

2

segments of population. In this backdrop, the present study is an attempt

to understand both the Private Health Insurance (PHI) and Micro Health

Insurance (MHI) Schemes in India.

1.2. Relevance of Health Insurance for India

Several recent papers and reports have critically reviewed the Indian

health care delivery and financing system. As indicated by the World

Development Report 2003, the total world health expenditure is 9.0

percent of the Gross world income out of which the share of both public

and pIivate sector is 5.3 percent and 3.7 percent. respectively. For

developing countries as a whole, expenditure on health accounts for

about 5 percent of total public expenditure and, on an average, 2 to 4

percent of the GOP (WDR 2003). As against this, the total health

expenditure in India is 5.2 percent of the GOP. and out of this the public

health spending account for less than 20 percent and the rest is the

contribution by the private sector (WDR 2002). In India, the Per capita

total expenditure on health at average exchange rate (US$) though

increased from $ 22 in 1998 to $ 30 in 2002, the Per capita government

expenditure on health at average exchange rate (US$) was $ 6 through out

the period (WHR 2005).

It has reported that 40 percent of the hospitalized having had to borrow

money or sell assets, during the decade 1986-96, there was a doubling in

the number of persons who were unable to seek healthcare due to

financial reasons (NSSO 1996), and almost 24 percent of the hospitalized

Indians fall below poverty line because they are hospitalized (Peters et al

2002). A recent World Bank (2001) study on India concludes that out-of­

pocket medical costs (estimated to be more than 80% of the total medical

expenditure) alone may push 2.2% of the population below the poverty

line each year. Many studies indicate that Indians tend to use health care

services more frequently (Duggal and Amin 1989; Berman 1996).

According to the NSS data (1996), the percentage of ailing persons treated

during 15 days is 83 in rural area and 91 in urban area and among those

who have not sought medical care in spite of their illness, around 24 per

3

cent in rural and 21 per cent in urban areas have cited their lack of

financial capacity as the reason for not seeking treatment.

Recent household-level studies carried out in India. both at national and

regional levels. have indicated that the proportion of patients that pay for

services can be qUite high. ranging from 64 to 90 per cent (Duggal and

Amin. 1989; George. 1997; Sundar. 1992). Out of the total health

expenditure in India. the public health spending accounts for less than 20

percent. and the remaining is the contribution by the private sector (WDR

2002). Peter Berman (1996) revealed that almost all of this private spending

is on curative care - consultations. diagnostics and in-patient care. In

contrast to this. a lion's share of public health expenditure is on preventive

and promotive health care. which is at the expense of curative care (Phadke

1994). Moreover. a slight majority of people who are ill or sick seeks care

from public providers for in-patient care. that is. the most common

outpatient episodes are treated before the private provider2 . Furthermore. it

is important to note that. as revealed by the recent household-level studies

on utilization on health care. even public care is not all that 'free' after all;

there are many incidental expenses that consumers have to bear on their

own (Uplekar and George 1994. Sundar 1995). Sundar (1995) points out

that average spending per out-patient episode at the public facilities is

about 40 percent of the average expenditure on visits to the private sector.

while the public in-patient treatment expenditures average about a quarter

of the private in-patient treatment costs. Dissatisfaction with the quality

and quantity of curative services. under funding and lesser access IS the

limitations of India's public health care system and the majority of the

consumers of the public health system are the weaker sections of the

SOCiety. and there IS a growing preference for health care services being

provided by the private sector. In short. the treatment from both public and

private facilities imposes considerable fmancial burden on individuals in

the form of out-of-pocket expenses. However. approximately 65 percent of

all spending on curative and diagnostic care in India consists of direct out­

of-pocket expenses. which are not reimbursed (Peter Berman. 1996).

2 As already mentioned above. the people used to approach both the public and private sector health care provider for treatment. As far as India's public health care system is concerned. dissatisfaction with the quality and quantity of curative services. under funding and lesser access is the limitations painted out by the studies.

4

The adoption of 'new economic policies' and the subsequent refonns such

as sector refonns escalated the cost of health care further. The present

trend of imposing user charges in the public hospitals in many states and

reduction in public health subsidy may lead to an increase in the health

care burden of the population. Making this issue more vulnerable, 80% of

the public health subsidy goes to the Iicher sections of the society (Mahal,

A, et.al., 2000). The financial burden of health care is, however, unduly

heavy for the households belonging to the infonnal sector indicating a

potential for voluntary comprehensive health insurance schemes for such

sections of the society (Gumber and Kulkarni, 2000). The health care

expenditure is on an increase in India; the annual rate of inflation in the

health sector is estimated to be 31 per cent and 15 percent for inpatient

and outpatient care, respectively (cited in SUJatha Rao, 2004). As a result,

the out of pocket expenditure of the people has increased more than

proportionately duIing this peIiod. Further, we are in an era of the rapid

technological progress, which makes it possible to treat more diseases with

new potential areas for treatment and prolong life expectancy with resulting

increases in need and demand for health services, which will put more

pressure on the health care financing front.

India is at the door front of the transition from the second to third stage of

demographic transition characterized by low birth rate and low death rate.

The fact here is that the life expectancy of the people is increasing and, as

a result, the number of ageing or elderly population too is proportionately

increasing. The basic theory of the inverse relationship between age and

health status states that as people get aged, their health status will

deteIiorate, and has ample empiIical evidence to substantiate giving a

strong message of an overall increase in the health care burden of society

due to high medical care consumption of these groups (Omram, 1971).

It can also be observed that India is moving towards an epidemiological

transition, and some states like Kerala have already begun to expeIience

such a transition. Major epidemiological studies (Jamison et al. 1993;

Murray and Lopez 1996) have documented important changes in the

burden of disease and mortality in developing countIies. As a population

5

undergoes a demographic transition. it also experiences a shift in Its

characteristic patterns of disease. Omram (1971) believed this pattern is

unchanging and labeled It as epidemiological transition and described It

as a shift away from diseases of famine and pestilence to receding

pandemics to an age of generative and man made diseases.

Epidemiological transition Implies a change in the morbidity profile from

acute. infectious. and parasitic diseases (e.g. plague. smallpox. and

cholera) to non-communicable. degenerative. and chronic diseases (e.g.

cardiovascular diseases. cancer. diabetes. and neoplasm). More

speCifically. three fundamental changes In the configuration of a

populatlon's health profile take place during epidemiological transition: (I)

mortality decline due to infectious diseases. injuries. and mental illness:

(ii) shift of the burden of death and diseases from the younger to the older

groups; and (iii) change in health profile from one dominated by death to

one dominated by morbidity. Differing epidemiological patterns between

town and country coexist and continue to widen. In such a situation of

polarization. the danger is that of the scarcity of resources for diagnostic

and curative services avallable to the rural and the urban poor. Thus. the

demographic change In terms of Increase in the number and proportion of

elderly people in the population. and epidemiological transition in the

form of new types of diseases are the actual and potential sources of

higher health expenditure for the people. For appropriate societal

responses to the requirements arising out of the epidemiological transition

is a concomitant health care transition (Caldwell. 1990).

Hence. one of the issue that emerges out of the above discussion is that

the cost of treatment poses severe constraints for both who are seeking

health care and those who are not. Largely. the cost of treatment Is

significantly influencing the health seeking behavior of Indians. Some

sections of the society are able to afford the health care services while

others can ill-afford it. Certainly. this scenario calls for an alternative cost

sharing mechanism where health insurance is conSidered as an efficient

mechanism through pooling of the health care burden between the rich

and the poor. between healthy and unhealthy. and between young and

aged. Recently. there is a growing interest and consensus among policy

makers. community organizations and researchers in India on health

6

insurance as an efficient and equitable social security mechanism to

ensure universal access to high quality health care to all sections of the

society.

1.3. Taxonomy of Health Insurance in India

The health insurance situation in India can be understood under the

following headings:

1.3.1. Public (Social) Health Insurance Schemes

The most prominent an10ng the protective schemes are the Employees'

State Insurance Scheme (ESIS) for workers in the organized private

industrial sector and the Central GoveTI1ment Health Scheme (CGHS) for

its employees. The beneficiaries of the above schemes are the salaried

class who belong to formal sectors. Some "Employer-managed health

facilities" and the "reimbursements of health facilities" are also available

in India which are limited to only a few. The 2003-04 Union budget

proposed introduction of a universal health insurance (UHI) plan for

people below the poverty line in tie-up with Insurance Companies.

1.3.2. Micro Health Insurance (MIll) Schemes

MHI schemes are based on not-for-profit principle and targeted to the

underprivileged sections of the society. In India. currently there are more

than 20 MHI units and many organizations are coming ahead with

various proposals to Introduce HI from getting inspiration from the

successful stories of the existing MHI units.

7

Table: 1.1 Locations and size of membership of some selected MID Units

Name ofMHIs Location Size of Membership (Number of Individuals)

1) ACCORD-1992 Gudallur, Nilgiris (Tamil Nadu) 13070

2) BAlF-2001 Pune District, villages around 1500 U ruli -Kanchan(Maharashtra

3) BULDHANA Buldhana (Maharashtra 175,000 4) DHAN Mayiladumparai block, Theni 19,049

District(Tamil Nadu 5) KARUNA TRUST T.Narasipura taluk, Mysore Dt. 634,581

& Bailhongal taluk, Belgaum Dt (Karnataka)

6) MGIMS Wardha, Maharastra 30,000 HOSPITAL

7)NAVSARGJAN Patan District, North Gujarat -TRUST

8) RAHA Raigarh. Ambikapur. 92.000 Jashpurand Korba ts of districts of ChatUsgarh

9) SEWA 11 districts of Gujarat 1,067,348 10) STUDENTS West Bengal 5,60,0000

HEALTH HOME 11) VHS Chennai, Tamil Nadu 104,247 12) YESHASWlNI Bangalore. Karnataka 25,00,000

TRUST 13) NIDAN Patna. Vaishali. Muzaffarpur. 1.020

Khagria. Nawadah. Begusaria locations in Bihar

14) UPLIFT Pune, Maharastra 10,966

Sources: ECCP data, Devadasan et al, Documents from various MHIs

1.3,3. Private Health Insurance (PHI) Schemes

The private health insurance (PHI) schemes, often called Private Voluntary

Health Insurance schemes (PVHI) , are the schemes offered by insurance

companies in the open market in which enrolment into the scheme is not

determined by legislation. In India. the public and private sector

companies provide the PHI (voluntary). The General Insurance

Corporation (GIC), which comprises of four insurance companies namely

NIC. NIAC, OIC and UIC, is the largest public sector organization of

providing the PHI in India. The various policies introduced by the GICs

8

are Mediclaim Policy (group and individual), Jan Arogya Bima. Personal

Accident Policy. Nagarik Suraksha Policy and Overseas Mediclaim Policies

(employment and study jcorporate frequent travel/business and holiday).

Among these policies. the Mediclaim policy is relatively popular. After the

establishment of Insurance Regulatory and Development Authority

(iRDAl. many private corporates also have entered the HI market. The

Bajaj Allianz. Royal Sundaram. IerCI Lombard. Cholamandalam. Tata and

Reliance are the prominent private insurance companies. An important

peculiarity of these corporations is the tie-up with some health care

provider having super specialty facilities.

Table: 1.2 Private Health Insurance Sector in India

Public jPrlvate Name of the Insurance Title of the health sector companies insurance policies Public sector The Oriental Insurance 1.Mediclaim Policy companies Company Ltd 2.Jan Arogya Bima

Policv The New India Assurance l.Mediclaim Policy Company Ltd 2.Jan Arogya Bima

Policv National Insurance 1.Mediclaim Policy Corporation 2.Jan Arogya Sima

Policy United India Insurance 1.Mediclaim Policy Company Ltd 2.Jan Arogya Bima

Policy Private sector Royal Sundaram Alliance Health Shield companies Insurance Company

Limited Cholamandalam General Basic Health Cover Insurance Company Limited TATA AlG General Tata AIG Health First Insurance company Ltd Bajaj Allianz General Health Guard Insurance Company Ltd Critical Illness ICICI Lombard General Insurance Company Limited. HDFC Chubb General Group Accident Policy. Insurance Company Hospital Cash (Accident Limited only)

Source: Insurance Regulatory and Development AuthOrity (IRDA). 2006

9

The Life Insurance Corporation (LIC) of India introduced a special

insurance programme called 'Ashadeep' which covers medical expenses

for four dreaded diseases namely, Cancer (malignant), Paralytic stroke

resulting in permanent disability, Renal failure of both kidneys or

Coronary artery diseases where by pass surgery has been done, Another

policy by the LIC, called Jeevan Asha Plan, covers many surgical

procedures. But these poliCies are a kind of savings schemes and the

premium is almost equal or more than the insurance amount. in short, do

not follow the principle of insurance (risk pooling) in strict sense of the

term.

1.4. Research Problems and Questions

In India, the coverage of HI in some form or the other, i.e., whether public

or private, is abysmally low and is only around 3%3. Even though there is

no data set to give an accurate figure on India's HI coverage, a rough

estimate is given in the following table.

Table: 1.3 Selected health insurance coverage in India

Sources of coverage Covered lives (in thousands)

Central government Health Scheme 4,276

(CGHS)

Employees State Insurance Scheme (ESIS) 31,050

Mediclaim Policy (voluntary) 10,000

Universal Health Insurance scheme ------*

Government non-life insurance companies 56

Non-government Non-life insurance 13

companies

Community health insurance 215

Sources: Insurance Regulatory and Development AuthOrity Journal, October, 2004, and compiled by the author from different sources .

• Not available.

3 Different estimates were being cited by various authors from different sources on health insurance coverage in India: Peter Berman (2006):10%. Misha Segal (2004); 15%, Susan Mathie, and Kenneth Cahill (2004): 3%, Indrani Gupta (2004); 3%. All these authors raised doubts on the reliability of their estimate; however, the estimate by Indrani Gupta (2004) at 3% seems to be more reliable. The main message of all these citations is that health insurance coverage is very low in India.

10

As mentioned before, the CGHS and ESIS cover the people of the fonnal

sector only by just limiting the coverage to the central government

employees and Industrial workers, respectively. A majority of Indian

population belongs to infonnal sector and do not have any fonnal social

security measures against the illness episodes.

It is evident from the literature reviewed that HI mechanism is a viable

solution in tenns of promoting efficiency and equity in the health care

sector. After liberalization and globalization of Indian Economy, it has

been assumed that market mechanism may meet the requirements of

people, and the State can limit its role as a facilitator. Even though, the

insurance industry is an emerging sector in India and the HI premium

contributes to less than one percentage of their total premium revenue,

expectation is growing among some corners that the voluntary HI market

is one of the options before the public to have health insurance coverage

until and unless the government and other organizations come up with a

concrete policy solution to provide the same to its citizens. Further, it can

be inferred from the policy documents that both the Central and State

governments of India4 are looking towards a strong private health

insurance market to meet the increasing huge health care finanCial

burden of the people. One of the characteristics of these insurance

companies is the presence of branches allover India that are fairly

distributed. But, as the PHI schemes are being offered by market sector,

one can not expect that these schemes will cover the poorer sections of

the society, till such time when specific schemes by these providers to

address the poorer sections are in place. The government and market

sectors have largely failed to develop insurance for the poorer sections of

the nation. Micro Health Insurance Units (MHIUs) created and operated

by local people have been proposed (Dror and Preker 2002) as an

approach to insuring health for the poor. Some of the studies reveal an

impression that the MHI Schemes are getting momentum in India and

they are able cater to the health care requirements of the poor

(Devadasan,2004) .

4 SpeCifically. the Union budgets and State government's budgets and also the IRDA publications have been highlighting the promotion of health insurance with the help of Insurance companies.

11

Thus. two fonns of HI such as PHI and MHI schemes are the options

before the public. Why these schemes have covered only a small fraction

of the Indian population? Many studies have indicated that Indians are

willing to pay for HI (K. Mathiyazaghan. 1998: Dror et at.. 2007). In a low

income country like India where a majority of the people are living in rural

areas and working in the infonnal sectorS, this low level of health

insurance coverage is not justifiable, espeCially in a context where any

kind of catastrophic illness leads to a high cost of treatment and loss of

earnings6 • As already noted. studies show that the cost of treatment in

India is very high in both pubJic7 and private sector hospitals and has

been increasing; and also the quality of care in the public sector health

facilities are very poor. Many people are not able get access to health care

mainly because of financial constraints.

The basic question here is centered on the very low level of HI coverage in

India. The country has no previous experience of having a situation of

high level of coverage and later on falling to the lower level, to give a

satisfactory answer to this question. It is in this context. the present

study is an attempt to address such a low level of health insurance

coverage and issues related to the scale up process of HI with social

welfare objectives. The following questions are raised in this context.

1) Why do many people fail to purchase health insurance in India if it is

so valuable? What are the constraints for the growth of a sound HI

mechanism in India?

2) To what extent both the PHI and MHI schemes have covered the

weaker and poorer sections of the society?

3) Between PHI and MHI systems. which one is more equitable and

adaptable to the Indian situation?

4) To what extent the available HI schemes in India reflect the

preferences of the people?

, It is estimated that about two-firths of India's GDP originates from the informal sector and almost 90%of families depend on this sector for their livelihood. S The studies on the use of health care services show that the poor and other disadvantaged sections lalso. they are in debt trap) such as scheduled castes and tribes are forced to spend a higher proportion of their income on health care than the better off. 7 Recently. in India. many state governments Introduced User fee to thetr district hospitals.

12

1.5. Objectives of the Study

The broad objective of the study is to understand the prospective role of

PHI and MHI as risk pooling health care financing strategies In India and

to point out what are the requisites for the growth of the same to achieve

the goal of a universal and comprehensive HI system.

The following are the specific objectives of the study.

I) To examine the equity aspects of HI coverage in India.

2) To examine the determinants of scale up of MHI and PHI schemes on

an equity basis.

3) To examine the Significance of information asymmetry and adverse

selection as factors influencing the scale up process of HI coverage In

India.

4) To analyze the ability of HI schemes to reflect the preferences of people

for various HI benefits to enhance the scale up process of HI coverage

in India.

1.6. Main Research Hypotheses

The follOwing hypotheses have been considered relevant in this context.

1) MHI schemes are not better than PHI schemes In assuring equity In HI

coverage In India.

2) Information Channels do not have significant roles in the coverage of

both MHI and PHI schemes in India.

3) There is no adverse selection in the HI enrollments In India.

4) The prevailing HI schemes do not reflect the preferences of the people

in India

13

1.7. Scope of the Study

Notwithstanding the view that HI is a viable solution to ensure access to

basic health care services to the masses, the number of people with HI

coverage is very low in India. We do not have a history of spread of HI

coverage in the past to find answer for the low HI coverage. There are

some structural issues with the system. The present study is an attempt

to find the causes for the low HI coverage and to derive necessary pre

conditions for the growth of a sound HI system in India. The study

addresses the scope and relevance of both the Commercial/Private Health

Insurance Schemes and Community/Micro Health Insurance schemes

(MHI) that are organized at the macro and grassroots levels, respectively.

Given the growing interest on the importance of HI. the outcomes of the

present study is considered useful in gUiding policy making and other

stake holders on the scale up process of HI in India.

1.8. Organisation of the Thesis

The study is organized in eight chapters. The introductory chapter

presents the context, relevance, research problem and objectives of the

study. The second chapter deals with the conceptual and theoretical

frame including literature review, data sources and methodological

aspects of the study. The third chapter investigates the nature of HI

coverage by analyzing how equity is assured in the HI enrolment across

both PHI and MHI schemes. The fourth chapter analyses the determinant

of HI coverage in MHI schemes. The fifth chapter discusses the role of

information asymmetry through information dissemination channels and

tests for selection bias in both PHI scheme as factors affecting both the

equity and scale up process of HI coverage. The sixth chapter examines

the selection bias in MHI schemes. The comparison of the preferences of

the people and the prevailing HI schemes is the theme of the seventh

chapter. The last chapter summarises the thesis findings and discusses

policy implications.

14

CHAPTER TWO

CONCEPTS, REVIEW OF LITERATURE AND METHODOLOGY

2.1. Introduction

This chapter explains the concepts used in the study. presents a critical

review of relevant literature. data sources and methodological aspects of

the study. First, the basic concepts of HI are explained. Further. the

review of literature is organized on selected themes with an objective of

elaborating the conceptual. theoretical and analytical frame of the study.

The literature review is organized under the following themes namely,

Health Insurance Market, failures in Health Insurance Market. Demand

for Health Insurance. Willingness to Pay for Health Insurance. and Some

selected study on Health insurance schemes in India. Subsequently, the

data sources and methodological aspects of the study are described at

greater detail.

2.2. Health Insurance: Basic Concepts and Principles

Normally people seem to dislike risk. What is unpredictable to an

individual is predictable to a group of Individuals. Health care expenses

are not only expensive but highly random in nature. Health Insurance

mechanism provides a way by which risk sharing within a society may

take place (Akin. 1987). One of the most efficient ways of providing access

to universal health care is to pool health risks between rich and poor,

young and old. and employed and unemployed. to enable cross

subsidization in the form of health insurance. HI is a mechanism of

pooling fund from its members and paying them when they fall siCk. The

fundamentals of risk pooling and sharing depend on to what extent both

the risks and incomes related cross subsidization. that is. risk solidarity

and income solidarity are prevalent. Cross subsidization can be classified

into two types as vertical and horizontal. Vertical cross subsidization is

15

realized when solidarity is achieved between groups, for example, pooling

the health risk between rich and poor, young and old, and employed and

unemployed etc, while horizontal cross subsidization is within the groups,

that is, the healthy would pay for the illness costs of the sick in the same

group. The vertical cross subsidization will promote an efficient and

equitable risk pool. Moreover, an effiCient insurance system is based on

the law of large numbers, that is, it requires large risk pools.

The literature summarizes the case for HI under the following three

categories, namely i) illness cannot be predicted, ii) hospitalization costs

are lumpy and cannot be planned, iii) the proportions falling ill requiring

hospitalization in any large population is small and, therefore, pennits

risk pooling. These three factors enable a person to cover the risk of

illness at a very small cost, provided an appropriate insurance scheme is

in position (Krishnan 1996). By pooling financial contributions from many

people, insurance plan can cover the hospital expenses of those

experiencing catastrophic events, such as near-fatal illness or injury. The

experts of health care fmancing argue that there is no alternative to

pooling medical risks that provides the same level of protection to its

members. On equity angle, within the risk pool. benefits are provided on

the basis of need rather than by income class. Further, payments go to

the sickest people, and, because lower income and less-educated people

tends to be sicker, they also have the potential of benefiting more from

insurance claims (Me Greevy 1990).

The premium of insurance (cost of insurance) is determined by the value

of actuarially fair Premium (expected payoff) and the price (loading fees) of

insurance. Health insurers detennine actuarial premiums by using either

community or experience rating8 . When an insurance company uses

community rating, the actuarial premium is based on the risk

characteristics of its entire membership and there is no diSCrimination in

actuarial premium calculations on the basis of age, health status, claims

history or other factors (everyone pays the same premium for the same

product, irrespective of their risk or previous claims experience). In

8 The insurance system using such a premium rating is called as Private community-rated health insurance and Private risk-rated health insurance models, respectively.

16

contrast. when actuartal premiums are determined using experience

rating. insurers place individuals. or a group of individuals. into different

risk categories based on vartous identifiable personal characteristics.

such as age (as a health proxy). gender. industrial occupation. and prior

illness and will be charged different premium accordingly. Another way of

fixing actuarial premium is through risk rating. which is a mix of both

community and experience rating. Insurers add loading fees to the

actuartal premium to arrive at a total premium. Loading fees cover the

administrative costs of supplying insurance. specifically. the costs of

marketing. underwriting. management. advertising. and claims

processing. In conventional theory, the loading fee portion of the premium

is considered the "price" of insurance because it represents the cost of

transferring the risk of medical expenditures from the individual

consumer to the insurer. Therefore. the price of insurance is the portion

the premium over and above the payment for the expected medical care

expenditures (Phelps. 1997).

To regulate and monitor the health care utilization of insured people.

insurers normally use the techniques of co payments and indemnity

payments. There are two types of Co-payments: 1) Co-insurance where

the benefiCiary must pay a certain percentage of the medical care

expenditure specified in the health insurance policy: 2) Deductibles where

insured pays a fixed amount of the medical care expenditure. The

indemnity payments are a kind of Payout limits where insurance

company pays no more than an established amount.

Individuals can take up insurance individually. or the cover can come as

parts of a group which are referred to as 'Personal or Individual' and

'Group' health insurance. respectively. The distinction between group

insurance and individual health insurance is important because the

former can bring important social elements into the private cover. Premia

under group insurance are often lower because insurers bear lower

administrative costs and the size of the pool is greater.

17

2.3. Health Insurance Market

Economists generally favour choice in health insurance for the same

reasons they favour choice in other markets; choice allows people to opt

for a plan that is best for them and encourages plans to provide services

efficiently. But choice in HI is a mixed blessing because of adverse

selection due to which people can select their poliCies according to their

health risk and insurers can reduce the degree of adverse selection

problem by offering insurance coverage to the needy after distinguishing

them. However. optimal design of poliCies must make tradeoffs

appropriately between risk sharing, on the one hand, and agency

problems such as moral hazard and supplier induced demand (cost

escalation), on the other.

Michael E. Chemew et al (1999) examined existence of equilibrium in

insurance markets when the number of insurance policy attributes is

increased (i.e., managed care is introduced). Individuals choose an

insurance contract from an endogenous choice set. The introduction of

managed care improves the ability of low risks (from healthy people) to

distinguish themselves from high risks (from unhealthy people). However,

managed care expands the product space in which a pooling policy could

break a separating eqUilibrium.

One of the major problems of an unregulated competitive market for

individual health insurance scheme is the seeming incompatibility of the

equivalence prinCiple and the solidarity (or fairness) prinCiple. The

equivalence principle of a competitive insurance market implies that an

insurer has break even on each insurance contract. The solidarity

prinCiple implies that the high-risk individuals receive a subsidy from low­

risk individuals to access the health insurance coverage.

The two types of solidarity in health insurance market, which may lead to

cross subsidization, are risk solidarity (solidartty between high risk and

low risk individuals) and income solidarity (between high-income and low­

income individuals). In a competitive market a system of cross-subsidies . cannot be sustained because competition minimizes the predictable profit

18

per contract. Consequently, an insurer has break-even on each contract

either by adjusting the premium to the consumer's risk (premium

differentiation) or by adjusting the accepted risks to the premium

(selective underwriting).

Because of the importance of health insurance coverage in meeting the

huge medical bill in the uncertain illness episode, access to health

insurance coverage is an important aspect in the HI literature. Premium

rate restrictions are often conSidered a tool to increase access to coverage

for high-risk individuals in such a market. The study by Wynand et al

(2000) analyzed 3 strategies to increase a high-risk individual's access to

coverage in a competitive individual HI market: 1) Premium rate

restrictions9 for specified HI coverage, 2) Risk-adjusted premium

subsidies, and 3) A combination of both. By considering the risk-adjusted

premium subsidies, the study inferred that subsidy approach is the

preferred strategy to increase access to coverage for high-risk individuals.

According to this study, a competitive market for individual HI tends to

risk-adjust the premiums. It can be also seen that, now-a-days, the

premium per person may be related to age, gender, family size, region,

occupation, length of contract period, individual or group contract period,

the level of deductible, the sum insured, health status at the time of

enrolment and health habits such as smoking, drinking, exercising (Abel­

Smith 1992).

When markets cannot charge premium that accurately reflect the

individual's risk of using covered services, competition is not efficient; low

risks are not able to purchase comprehensive coverage in such markets

(Rothschild, M and Stiglitz 1976). Pauly (1974) and Stiglitz (1994) have

argued that compulsory pools are a Pareto-improvement over competitive

markets, which is an indication for the need of government intervention in

the HI market. Pools can make all consumers better off because the low

risk's gain from purchasing more complete coverage is greater than the

subsidy required for the high risks. Because government is the only social

organisation with compulsory membership, it alone has the power to

9 There are two types of premium rate restrictions- Community rating (by class). and a ban on certain rating factors. or rate banding (by class).

19

enforce a welfare-enhancing pool of low and high-rtsk individuals. Dahlby

(1981) has pOinted out that this conclusion is valid only if Nash

equilibrium does not exist in the competitive insurance market. Wilson

(1977) has shown that compulsory partial pooling which permits private

insurers to sell supplementary coverage may represent a Pareto­

improvement over the competitive equilibrium.

If premium for HI are not rtsk related. there exists a consumer

information surplus that may result in adverse selection. The study by

Wynand et al (1995) revealed that insurers can greatly reduce this

surplus by risk-adjusting premium. They concluded that there need not

be any substantial unavoidable consumer information surplus if

consumers can choose whether to take a deductible for a one or two-year

HI contract with otherwise identical benefits. Therefore. adverse selection

need not be a problem in a competitive insurance market with rtsk­

adjusted premiums or vouchers and with such a consumer choice of

health plan. For example. in Chile. private insurance does not face

adverse selection. pOSSibly due to the design of insurance plans. However.

this design does not prevent over-utilisation.

According to Randall P. Ellis (1998), reimbursement incentives influence

both the intenSity of services and who is treated when patients differ in

severtty of illness. He compared the social optimum to the prtvate

Coumot-Nash solution for three provider strategies: creaming - over­

provision of services to low severity patients; skimming - under-provision

of services to high severtty patients; and dumping - the explicit avoidance

of high severity patients. Cost-based reimbursement results in over

provisioning of services (creaming) to all types of patients. Prospectively

paid providers cream low severtty patients and skim high severtty ones. If

there is dumping of high severtty patients. then there will also be

skimming.

It can be seen that the above discussed review deals with the issues like

the importance of choice in the health market in order to deal with the

selection issues, changes in premium rate as a way to increase access for

different category of risk people. and market eqUilibrium. In this context,

it can be observed that the Indian health insurance market, even though

20

at the infant stage. is characterized by varieties of HI products offered by

11 insurance companies. comprising of both public sector and private

sectors. By a preliminary look up on each policy. it can be inferred that

the insurance poliCies of each company is different from the rest in its

design and characteristics. Moreover. the premium rate for each

insurance policy varies according to disease. age. gender. group/single

policy and total insured amount. The present study will analyse how each

policy will address the issues like the high health risk and low health risk

people. and their access. cream skimming. the over-utilization etc.

Recently. the introduction of managed care in the Indian health insurance

market by the Insurance Regulatory and Development Authority (IRDA) in

the form of Third Party Administrators (TPAs) may be expected to

influence the features of the insurance poliCies.

2.4. Market Failures in Health Insurance Market

There are mainly 2 forms of market failures in HI market. namely

Selection Bias (adverse selection) and Moral Hazard.

2.4.1. Selection Bias

Asymmetric information about potential demand for medical care creates

another analytical problem for insurance markets. Individuals themselves

know much about their health condition than the insurance companies.

Selection can be described as actions by insurers and consumers to

exploit un-priced risk heterogeneity and break pooling arrangement

(Akerlof. 1970). It can be classified into two types - Adverse Selection and

Cream Selection (skimming). The problem of adverse selection is present

in all lines of insurance due to the hidden infomlation. the people

Insuring themselves are those who are increasingly certain that they will

need the insurance (Akerlof 1970). "Adverse selection" arises because

individuals face different risks. Customers who know themselves to be at

high risks are motivated to buy more insurance and are likely to use it

(WDR 1993). In a population of individuals whose underlying health risks

21

are heterogeneous. more or less healthy people will demand different

insurance policies. which will be an "adverse selection" for the insurance

companies. Defensive efforts to obtain valuable information about risks

add to the cost of insured health care without improving health outcomes.

Adverse selection presents a serious problem for risks existing at the time

when insurance is taken up but an even and more complex problem

arises from the fact that an initially low risk person becomes high risk

later in life (WDR. 1993). Neither solution is easy to implement because of

the extreme uncertainty: insurance can cover known risks but not

uncertain risks.

A straight forward method of preventing an extreme form of adverse

selection - that is. one in which low-risk individuals do not buy the

specified HI coverage and thereby do not cross-subsidize the high risk

individuals - is to mandate that everyone buys the specified HI coverage.

On the contrary. a straightforward method of preventing an extreme form

of cream skimming - that is. one in which insurers refuse to (renew a)

contract with relatively high-risk individuals - is to require open

enrolment. The cream skimming may result in the exclusion of the

population group consisting of aged. poor. women and high health risks lO •

10 A detailed discussion on various aspects of adverse selection is attempted in one of the subsequent chapters.

22

Table 2.1 A summary of different studies about selection bias

Paper Data Empirical Highlights of the Selection Methods results

Juba. Lave, Carnegie- Maximum Lower family self- Adverse and Shaddy Mellon likelihood logit reported health (1980) University estimates of status results in

employees' detenni-nants significantly less health of plan choice chance of selecting insurance HMO enrollment enrollment and survey

Farley and 1977 OLS and 2SLS Ambulatory care Ambigu-Monheit National estimation of expenditures have ous (1985) Medical Care health an insignificant

expenditure insurance impact on health survey purchases insurance

purchases Wrightson, Disenrollees Comparison of Disenrollees have Adverse Genuardi, from 7 plans costs and lower inpatient and offering disenrollment costs and occupy Stephens different rates for less risky (1987) types of insures demographic

managed groups than care continuing

enrollees Cardon and National Tobit-style Individuals who Adverse Hendel Medical model are younger, male, (1996) Expenditure insurance or in "excellent"

Survey chOice self-reported health are significantly less likely to become insured

Ellis (1985) 1982-83 Logit estimates Age and worse Adverse employee of health plan previous years' health plan chOice health expenses enrollment are associated and expense with chOice of records of a more generous large firm health coverage for

the next year Marquis Plan Comparison of 73% more Adverse (1992) selection of plan choices individuals in high

families in with age/sex Iisk quartile Rand Health adjustments choose most insurance under vaIious generous plan Experiment group-rating than those in low

regimes risk quartile, even with age/sex or expeIience rating

Van de Ven Survey and RegreSSion of Age-and sex- Adverse and Van claims data risk factors on composition of Vlenit (1995) from 20,000 prediction plans explain 40

families error of % of error in insured by difference in predicted cost largest Dutch costs between differential insurer, members of between plans Zilveren high and low Kreins cost plans

23

2.4. 2. Moral Hazard

One of the limits, which have been much stressed in insurance literature.

is the effect of insurance on incentives (Arrow 1965). It is frequently

observed that widespread medical insurance increases the demand for

medical care. Moral hazard (hidden action) refers to the likely malfeasance

of an individual making purchases that are partly or fully paid by others

(Arrow, 1965; Pauly, 1968 and 1974; Zeckhauser. 1970; Kotowitz. 1987).

It occurs when members of a HI plan use services more frequently than

they would have had they not been the members. In short. the moral

hazard occurs when insurance contracts are written on the basis of

endogenously incurred expenses and not on the basis of exogenous health

needs (William, 1999). Moral hazard is a concern because it conflicts with

risk spreading goals. HI involves a fundamental trade off between risk

spreading and appropriate incentives.

Increasing the generosity of insurance spreads risk more broadly but also

leads to increased losses for the insurance companies because individuals

choose more care (moral hazard) and providers supply more care

(principal-agent problems). Pauly (1987) attributes incomplete HI coverage

to "moral hazard". That is, for insurance markets to produce an optimal

allocation of resources, the method of insurance must be neutral with

respect to the demand for medical care. Neutrality is absent with

insurance because the price of medical care to the insured is below

marginal cost, leading to inefficient usage of medical care resources.

Moral hazard also results from patients making less effort to search for

low cost providers,

Moral hazard arises because medical needs are not fully mOnitorable. and

different people with similar condition have different optimal

expenditures, at least as best as the insurance company can determine ll .

If people have "too much" health insurance, they may have an incentive to

II There is some moral hazard in the markets for house and vehicle insurance. But unlike the consumption of too much health care, these actions are climes, with penalties that may greatly exceed the value of the asset. It is harder to attribute them to behavioral chOices. There is no market value for the human body and no possibility of abandoning one that is worn out and acquiIing a new one.

24

use "too much" health care at too Wgh prices. Unfortunately the difficulty

of judging health care risks and the impossibility of placing a value on a

living body makes it impossible to determine how much is "too much" in

health care and HI (WDR. 1993). Two types of control are typically used to

minimize moral hazard such as demand side and supply side policies:

Demand side measures attempt to control a patient's demand through

financial penalty in the event of an insurance claim by using Co­

insurance, Deductible, Payout limits and No-claims bonus l2 . Insurers

also use non-fmancial control techniques to verifY that treatment given is

appropriate, effective and cost-effectiveI 3 . Supply side policies are mostly

used as state control on the overall level of expenditure and access to

services through limitations on the number of doctors, hospital beds and

medical technology using regulatory and bureaucratic controls. But these

measures are not in the best interest of health care.

Significant effects of insurance on the demand for health services (moral

hazard) have been found in the literature (Newhouse, 1993; Cameron et

al., 1998; Bertranou, 1998). The usual finding is that those who

voluntarily purchase HI have a higher health risk than an average

individual in the population, and consume more health care services than

if they were not insured. Empirical evidences show that both moral

hazard and demand inducement are quantitatively important.

The RAND Health Insurance Study was one of several social science

experiments conducted under federal government auspices in the 1970s

to learn more about how insurance affects demand for health care

(Newhouse, 1993). This study basically followed standard "laboratory"

experimental design methods where a total for 5,809 enrollees was chosen

from four US cities and two rural sites. By giving different option of health

12 Co-insurance - A percentage of the charges for medical care specified in the policy, that the benefiCiary must pay. Deductibles - insured pays a fIXed amount of the cost. Pay oUllimits •• insurance company pays no more than an established amount. No-claims bonus - no (or small number of) claims in a year results a reduction on the cost of next year's policy. l"These techniques are generally known collectively as 'managed carc' and encompass both financial incentives for providers - such as capitation funding for institutions and individual physician bonuses/penalties - and management of clinical activity - for example utilization review, physician audit and drug formulations.

25

insurance plan with different copayments for the selected people. the

study inferred that HI coverage leads to over-utilisation of health care at

various degrees as according to the types of health care goods. Emmett B.

Keeler and John E. Rolph (1988) analyze claims data from the RAND

Health Insurance Experiment, which were grouped into episodes of

treatment. The insurance plans in the experiment have coinsurance and a

cap on out-of-pocket spending. Using new statistical techniques to adjust

for the increased sickliness of those who exceed the cap. the effects of

coinsurance on cost per episode and number of episodes are estimated.

Cost sharing reduced the number of episodes but had little effect on cost

per episode. People in the experiment responded myopically as their

current insurance status changed through the year. The price elasticity of

spending was about -0.2 throughout the range of coinsurance studied.

Cameron and Trivedi (1991) within the framework of inter-temporal (two­

period) utility maximization under uncertainty. pOinted out that in

Australia. in the initial periods. individuals (or family groups) choose the

health insurance plan. without knowledge of the health status. which will

determine their demand for services during the period to follow.

Individuals choose a health plan and make utilization decisions so as to

maximize expected utility.

To estimate the over-utilization. ClaudiO SapeUi and Bernardita Vial

(2003) compared the utilization of health care services with public and

PHI by the dependent and independent workers with and without

insurance in Chile. There is a difference in premiums in public and

private insurance poliCies of the Chilean health insurance system. Public

insurance sets premiums as a percentage of income. while private

insurance uses risk rating. In Chile. law allows private insurance

institutions. to adjust premiums according to age. sex. and number of

dependents. Based on the 1996 CASEN survey. the study found that

utilization of services by the non-insured is almost always lower than

utilization by the insured. The study inferred that this might be due to

differences in costs. moral hazard. or self-selection based on observable or

non-observable characteristics. As expected. utilization increases with

illness. There is no clear relationship between income and utilization of

26

medical services; utilization among insured workers decrease with

income. but among non-insured workers it increases.

In the above-mentioned Chilean study. the dependent vaIiable is the

number of services (physician visits and days of hospitalization)

consumed by the household head during the three months preceding the

survey. The choice of HI is made by the household head. and is assumed

to have been made prior to the 3-month period in question. To estimate

the utilization equation. the study used count data model. since the

dependent variables are discrete. The empirical model allowed for different

marginal effects in different population segments; and corrects for self­

selection of individuals in health insurance decisions. since the choice of

health plan is endogenous. The model to explain the quantity of services

consumed in the insurance plan j takes the form:

Where yt is the number of services consumed by the family unit i.

dji the dummy variable with a value of one if insurance plan j is

purchased. Xi the vector of characteristics of family unit i. and £i is

the heterogeneity component in the count equation. The expected

value of over-utilization associated with the purchase of health

insurance is the difference between (i) the expected number of

services consumed by individual i after purchasing the insurance.

and (ii) the expected number of services which this individual would

consume if he/she had not purchased health insurance:

Moral hazard (insurance) = E (yt/DI =1) - E (YN/DI = 0)

Where YI is the utilisation with insurance. and YN the utilisation if non­

insured; DI = 1 for individuals who had purchased health insurance (and

DI=O for uninsured individuals).

The probability of purchasing health insurance is greater for faInilies with

higher income. young children. larger household size. and more

27

education; and when the household head is older, female, and contributes

to a pension saving account. A higher income, younger age, smaller

number of dependents, residence in urban area, higher educational level,

and employment in a larger company, all these increase the probability of

choosing private insurance, The finding that older age and more

dependents positively affect affiliation to the public insurance indicates

the presence of self selection against public insurance based on

observable risk variables. All these results agree with previous work in

the area (Sapelli and Torche, 1998). The study revealed that insured

workers consume more than twice the quantity consumed by non-insured

workers. While considering private and public insurance beneficiaries

separately, it was found out that moral hazard is larger in case of public

insurance. This result is consistent with the fact that independent

workers who purchase public insurance have access to almost complete

coverage in physician visits, but in the private insurance sector co­

payments are usually different from zero. For hospitalization days moral

hazard is not significantly different from zero, a result consistent with

much lower price elasticity of demand for hospitalization than for

physician visits. In short, there is no over-utilisation in the case of

hospitalization, for either public or private insurance.

Bertranou (1998) studies the relationship between utilisation of

outpatient health care and HI in Argentina, His results are similar to the

Chilean study. For working people without mandatory insurance he finds

higher utilisation among the insured (45% above average utilization) using

an OLS regression. When he uses two stage least squares to account for

the endogeneity of the dummy variable for HI, he finds an even higher

utilisation among the insured.

With rare exceptions, the provision of actuarially fair HI tends to

substantially increase the demand for medical care by redistributing

income from the healthy to the sick (de Meza, David, 1983), This suggests

that previous studies, which attribute all the extra demand for medical

care to moral hazard effects, may overestimate the efficiency costs of HI.

28

Stephen H. Long (1998) in his paper provided a test of the hypothesis that

people shift their consumption of health services to time periods when

they have more generous insurance coverage. in order to take advantage

of third-party payment. Data from the Survey of Income and Program

Participation is used to compare utilization rates for people in transition

between being insured and being uninsured to those who are

continuously insured and continuously uninsured. The study found little

support for the hypothesis that people anticipate changes in their

insurance status and arrange their health care consumption accordingly.

It has been observed that choosing optimal HI coverage involves a trade­

off between the gain from risk reduction and the deadweight loss from

moral hazard. Willard G. Manning and M. Susan Marquis (1996)

examined this trade-off empirically by estimating both the demand for HI

and the demand for health services. The study relies on data from a

randomized controlled trial of cost-sharing effects on the use of health

services and on the health status for a general. non-elderly population.

Using the Egyptian Household Health Utilization and Expenditure Survey

(1995). Winnie Yip and Peter Berman (2001) had shown that the School

Health Insurance Program (SHIP) of Egypt significantly improved access

by increasing visiting rates and reducing fmancial burden of use (out of

pocket expenditure). With regard to the success of targeting the poor.

conditional up on being covered. the SHIP reduced the differentials in visit

rates between the highest and lowest income children. However. only the

middle-income children benefited from reduced finanCial burden (within

group equity).

2.5. Demand for Health Insurance

The motivation behind people going for HI has been explained. from time

to time. oy the scholars in various ways. Daniel Bernoulli in 1938

postulated that an individual derives different levels of satisfaction or

utility from different levels of income (or wealth). Given a specific concave

utility function and a specific insurance problem. he derived the

conditions under which the utility level achieved after paying the

29

insurance premium exceeded the expected utility level from remaining

uninsured, and suggested that insurance was purchased because people

were maximizing expected utility. The cardinal measurement of utility

function by John Von Neumann and Oskar Morgenstern in 1944 made it

possible to measure the shape of the expected utility function and to

predict how individuals with variously shaped utility functions would

respond to the opportunity to purchase insurance. Shortly thereafter,

Milton Friedman and L.J Savage (1948) stated that consumers purchase

insurance because they prefer a certain loss to an uncertain loss of the

same expected magnitude, and concluded that the consumer is "choosing

certainty in preference to uncertainty". Kahneman and 1\rersky (1981,

1986, and 1988) through the prospect theory with the support of

empirical evidences argue that the opposite is true namely, that

consumers actually prefer an uncertain loss to a certain loss of the same

magnitude. John Pratt (1964) and Kenneth Arrow (1965) separately

developed a statistic of curvature of the consumer's utility function, a

measure at each level of income, y, is r(Y) = -U"(Y)/U'(Y), that became the

measure of the relative risk averseness of the individual consumer. Risk

aversion and uncertainty about future health creates a demand for HI

(Arrow, 1963; Phelps, 1975). Kenneth Arrow (1963) argued that if

consumers were 'rational expected utility maximizers', and 'risk averse',

and charged actuarially fair premiums, the case for health insurance is

"overwhelming."

Since the advent of Mark Pauly's (1968) influential article, almost many

health insurance economists believed in a theory that implies that the

voluntary purchase of HI makes the consumer worse off. Empirical

calculations based on this theory have borne this out. These studies have

implicated that consumers are worse off with HI contract with the features

of coinsurance rates, deductibles, and limits on out of pocket spending.

As against the conventional theories of demand for HI. a recent theory by

John A Nyman (2003) suggests that consumers who voluntarily purchase

unsubsidized HI are better off. And also, this theory rejects as

unnecessary the decidedly unintuitive approach to understanding the

purchase of insurance based on the utility function developed in 1944 by

John Von Neumann and Oskar Morgenstern. HI transfers income from

30

those who purchase insurance and remain healthy. to those who

purchase insurance and become ill. Thus. the decision to purchase

insurance is essentially a comparison of 1) the expected utility lost from

paying premium when healthy. and 2) the expected utility gained from the

income transfer if ill (Nyman. 2003). While trying to understand the

motivations behind buying HI. one caution has to be made here that all

these theories are adaptable to developed countries context.

The access value of insurance as a motivation for purchasing HI has gone

largely unrecognized in conventional theory. Accordingly. the access value

theory of health insurance proposed (Nyman. 2003) the main motivations

for purchasing insurance is that desire to gain access to those health care

services that would otherwise be unaffordable. For example. although a

U5$300.000 procedure is unaffordable to a person with U5$50.000 in net

worth. access is possible through insurance because the annual premium

is only a fraction of the procedure's cost. The value of insurance for

coverage of unaffordable care is derived from the value of the medical care

that insurance makes accessible.

In many developing countries. specifically in India. where access to health

care services is restricted due to low level of ability to pay. the access

value theory of HI has more policy relevance. Indeed. in India. HI has

begun to accept as a mechanism to extend health care security to the

poor. and reaching consensus to the principle of "access to health care

through access to Hr'.

Consumers differ in terms of the amounts and types of HI coverage they

buy. and these differences are reflected in such items as deductible

amount. the coinsurance rate. and the number of sickness events

covered. Some consumers purchase HI plans that offer first-dollar

coverage for all types of medical services. including routine care. Others

purchase HI plans with large deductibles and co-payments that cover only

catastrophic illnesses. Difference in health care coverage can be explained

by a host of factors. including the price of obtaining health insurance. the

individual's degree of risk aversion. the perceived magnitude of the loss

31

relative to income, and information concerning the likelihood that an

illness will accurately occur.

Through a regular telephone survey in Ireland, responses were obtained

from 2620 individuals randomly selected from the Electoral Register, as

according to the study by C. Harmon and B. Nolan (2001)14, reason for

having insurance, almost everyone with insurance regards 'being sure of

getting into hospital' and 'fear of large medical or hospital bill' as either

very or quite important. However, 'being sure of getting good treatment in

hospital' is also regarded as very or quite important by almost all the

insured.

Table 2.2 A summary of the reasons for having health insurance in Ireland

% Saying Reason very

important

Being able to have a private or semi-private 27.8

room in hospital

Being able to choose your own consultant

Being sure of getting in to hospital quickly

Being sure of getting good treatment in

hospital

Being able to get in to private hospitals

Being sure of getting consultant care

Being able to arrange hospital treatment for

when it suits you

Fear of large medical or hospital bills

52.7

86.4

77.4

27.2

67.5

68.7

88.5

% Saying very or quite important

65.2

88.9

98.6

95.9

63.3

96.0

95.7

98.4

Further, the study analyzed with a probit model, the choice of the

individual whether to buy private insurance or not. The explanatory

variables include a number of individual characteristics, such as age, sex

14 In Ireland. those with insurance generally received 'private' care in private or semi-private accommodation. and choose their own consultant. but much of this private care is delivered in public hospitals. In Ireland. the health care delivery system comprises of both public hospital and private hospital. which is somehow similar to Indian situation.

32

and marital status; household composition in tenns of the number of

children. adults and elderly persons; family income levels and health

status variables. The summary of the results of the study is as follows.

Specification-A includes only personal and household characteIistics. The

overall predicted probability of being pIivately insured on the basis of the

mean values of the data is 35% - very close to the actual figure in the

sample. The expected probability of choosing private health insurance is

dramatically different between the below median income (17%) and above

median income (48%) groups. The individual characteIistics such as

higher levels of education and higher levels of attainment are highly

significant in the probability of being insured. Both age and gender are

statistically significant determinants of the demand for insurance - older

respondents have lower probability of choosing private insurance. while

women appear to be more likely to be insured than men. In Specification­

B. the disposable income of the household is included as an additional

explanatory Variable - all of the principal results in the specification-A

remain. whereby higher income is associated with an increased

probability of choosing private insurance. Under Specification-C.

additional explanatory variable is the self-reported health status Le., very

good. good. fair, bad. and very bad. The poorer self-reported health levels

are all statistically significant and lower the probability of choosing private

insurance. Specification D includes Medical card status as an additional

independent variable. This is strongly negatively associated with the

demand for HI.

In Chile. the 1996 CASEN survey reports that 60% of the total

populations are beneficiaries of public HI, 25% purchase private

insurance, 11% have no insurance and the remaining individuals have

special coverage schemes. The study by Claudio Sapelli and Bemardita

Vial (2003) found that among independent workers. the percentage who

purchase insurance increases with income, age and household size, and it

is larger for women and for workers who live in urban areas. Among the

dependent workers who purchase private insurance increases with

income, and decreases with age and household size.

33

M. Susan Marquis (1995) examined decisions to purchase individual

insurance by workers who do not have employment-based insurance.

Using data from the Current Population Survey and the Survey of Income

and Program Participation. coupled with prices for a standard insurance

product in different market areas. the study estimated a price elastiCity of

-0.3 to -0.4 and an income elasticity of 0.15 and raised doubts that even

substantial subsidies to the working uninsured would induce many of

them to purchase coverage voluntarily.

2.6. Some Selected Studies on Health Insurance Schemes in India.

Another aspect of the demand for HI is the willingness to pay (WrP) for HI

by the people. In WfP for HI studies, respondents were presented with

well described but hypothetical situations of buying HI. Several studies in

India and abroad reveal that WfP for HI will be affected by sex, age, and

years of schooling, income, residence, and health status.

A study by K Mathiyazhagan (1998) about the willingness to pay for a rural HI

scheme through people's participation in rural Karnataka suggests that most of

the people are willing to join and pay. However, the probability of willingness to

join was found to be greater than the probability of willingness to pay. Further,

the study reveals that socio-economic factors and physical accessibility to

quality health services are Significant determinants of willingness to join and

pay for such a scheme.

Dror eLal. (2007) provides evidence on Willingness to pay (WfP), gathered

through a unidirectional (descending) bidding game among 3024 households

(HH) In seven locations where MHI units were in operation in India. Insured

persons reported slightly higher WfP values than uninsured. About two-thirds

of the sample agreed to pay at least 1 %; about half the sample was willing to pay

at least 1.35%; 30% was willing to pay about 2.0% of annual HH income as HI

premium. Nominal WfP correlates positively with income but relative WfP

(expressed as percent of HH income) correlates negatively. The correlation

between WfP and education is secondary to that of WfP with HH Income.

34

Household composition did not affect WTP. However. HHs that experienced a

high-cost health event and male respondents reported slightly higher WTP. The

observed nominal levels ofWTP are higher than has been estimated hitherto.

Sodani P R (2001) investigates the community's preferences on various

aspects of health insurance. According to the study. quality of care and

cost are the two important factors identified by the community as the

factors affecting their decision to subscribe to any new health insurance

plan. An integrated provider and insurer system is preferred. irrespective of

public or private-based management. Hospitalization and maternity

sen1ces are preferred among the given choices for benefits to be included

under the plan. The results also suggest that there is high level of

willingness to join a health insurance plan in future if designed carefully

for the infonnal sector.

Another study by Gumber and Kulkarni. (2000) conducted in Gujarat State

of India explored the availability of health insurance coverage for the poor.

especially women. their needs and expectations from a health insurance

system and likely constraints in extending current health insurance

benefits to workers in the infonnal sector. The study made a comparative

analysis of different fonns of health insurance i.e. the ESIS. Mediclaim

policy and SEWA, in the infonnal sector of Gujarat State. They analyzed the

comparative advantage of different fonns of health insurance in meeting

the health care burden of the people in the infonnal sector and also

estimated the demand for and willingness to pay for the health insurance.

The households subscribing to Medi-claim generally belonged to the higher

income strata and their average annual income was twice that of the

households enrolled with SEWA and ESIS as well as that in the non­

insured category. The literacy rate is very close to 100 percent for both

male and female medici aim households. They further pointed out that over

92 percent of the non-insured households in both rural and urban areas

had no awareness about the existing health insurance schemes. Further.

only a miniscule number of households were aware of other insurance

plans available in the market.

35

The lIMA study (1987) for an ADB seminar. based on Maharastra and

West Bengal. reviewed various health insurance schemes. namely. the

Sewagram experience in Maharastra. the Seba co-operative health society

in West Bengal. the government owned GIC schemes. the ESIS. and

CGHS. The moral hazard and adverse selection are the threat faced by the

Sewagram and Seba. as revealed by the study. The study highlighted the

poor performance of the mediclaim policy due to 1) even those who

can afford the premium are not typically insurance conscious; and 2) the

insurance companies have very low priority to the HI business since the

HI premium forms a meager portion of their total premium income and

hence they would not have followed aggressive marketing strategies.

Bhatt (2000) by studying the Mediclaim policies of GIC's at the

Ahmedabad City revealed that 64 percent of the claimant suffered from

non-communicable diseases where communicable diseases still account

for 50 percent of the mortality in India and also the average age of the

claimant was 29.45 and 43.08 for both the communicable and non­

communicable diseases. respectively. The study pOints out that there is

an increase in both the enrolment and claims. and a third of the increases

in claims are due to the problem of adverse selection and supplier

induced demand.

2.7. Data Sources and Methodology of the present Study

The present study looks into the available HI schemes before the public as

a remedy to reduce their health care burden. As it has been noted before

there are mainly 3 types of HI schemes. namely. 1) Public (Social) health

insurance schemes like CGHS and ESIS. 2) PHI schemes. and 3) MHI

schemes. But the Social Health Insurance (SH!) schemes are limited only

to some selected people in the formal sector. Thus. the present study

considers both the Private Health Insurance (PHI) and Micro Health

Insurance (MH!) Schemes as the viable HI schemes for the public in India.

The PHI schemes are the HI schemes offered by insurance companies

(irrespec4ve of whether the private sector and public sector companies) in

the open market in which affiliation into the scheme is not determined by

legislation. The MHI schemes are schemes provided by civil society and

36

other community organization at the grass root level in some selected

locations.

2.7.1. Data Sources

There is no systematic data base in India on HI schemes. The secondary

information from the providers of both PHI and MHI schemes is not

sufficient enough to address the research questions posed in the present

study. but it can be used to substantiate some of the research issues.

Thus. the information supplied by insurance companies and MHI units is

used as the secondary data source for the study. In addition to this. the

present study utilized 3 major data sets. namely.

I) Primary Data on Private Health Insurance Scheme (PHI)

2) Household Data on Micro Health Insurance Schemes (MHl)

3) Primary Data on Clients Preferences on Health Insurance Benefits

CHAT-I

2.7.1.1. Primary Data on Private Health Insurance (PHI)

Through the primary survey information from both insured and

uninsured people were collected. The study has used household (family)

as the sample unit because {as argued by the WIiters like Ngui, Burrow.

and Brown (1990)} the health of one family member (income unit) may

affect the utility of others. As income is likely to be shared between all

members of the family. the financial costs associated with one member

seeking treatment will also be shared. As such. the utility of all members

of the family may decline In the event of one of the members falling ill.

A multi-stage stratified random sampling technique has been used to

select the households for the primary survey. In the first stage. Kerala is

used to identifY the households. In the second stage. two districts were

selected out of 14 districts from Kerala. In the third stage. branches of the

four public sector insurance companies have been selected from each of

the two districts to identifY the insured people.

37

In the first stage. the state of Kerala had been selected to understand the

ground level realities of the nature of voluntary HI in India. In taking one

or more state or geographical area as a sample to represent India is a

difficult task with its endless diversity in terms of population. climate.

topology. religious beliefs. languages. and socio-economic and cultural

settings. All these diverSified facts have their own role in the health status

achievement and the related decision-makings by the population. The

Insurance companies providing health insurance schemes have been

fairly spread across all Indian States and Territories. and they sell similar

insurance policies across the States. and there is no evidence of any kind

of sharp and significant difference among all Indian states on the nature

of Voluntary HI coverage. The research questions raised and the research

objectives stated in the present study allow taking any State or

geographical area for selecting the insured people. from the list provided

by insurance companies. Moreover. given the time and resource

constraints. and the absence of a sound secondary data set for the

present study. purposively the state of Kerala has been selected.

In the second stage. two districts were selected out of 14 districts of the

state. On the basis of development. the 14 districts of Kerala can be

classified as less developed and developed districts. The study covers two

districts selected randomly - the first one from among the less developed

districts and the second one from among the developed districts. Thus.

the two districts. one from the Northern Kerala. i.e.. the district of

Kasargod and another from Southern Kerala. i.e .. Trivandrum have been

selected for the household survey. Both the districts are characterized by

the presence of urban. semi-urban and rural areas.

In the third stage. samples of the health insured and uninsured

households were selected. The insured people were selected from the list

provided by the branches of insurance companies in these two districts. It

can be seen that the Non-life insurance companies comprising of four

public sector and six private sector companies are supplying Voluntary HI

in India. These companies are providing a large number of policies in

Indian HI market. Among these. the four public sector companies. viz,

38

NIC. VIC. NIAC. and UIC. occupy a Iion's share in the HI market (around

90% of the total share). As against this. the private sector companies are

the new entrants to the market and have a very small market share. To

add. both the public and private sector companies are under the purview

of Insurance Regulatory and Development AuthOrity (IRDA) and these

companies have to follow the norms prescribed by the IRDA while

developing HI policies. Thus. it can be expected that there will be some

uniformity among the various HI policies of all insurance companies.

Thus. the study will take the sample of Insured people having coverage

from the public sector companies. The HI policies (schemes) of the four

public sector companies are similar in nature by both title and features.

The Mediclaim Insurance Policy (individual and group) of these four

companies reflects many features of a standard scheme and many of the

standard HI poliCies of the Private General Insurance Companies follow

these features of Medlclaim policy. Thus. the HI schemes of the private

health insurance companies such as Health Shield. Basic Health Cover.

Tata AIG Health First. and Health Guard are similar to the Medlclaim

policy with only slight differences. Moreover. the HI in India is generally

equated with the Medlclaim policy and also one of comprehensive one as

compared to all other poliCies. Introduced in 1987. this policy underwent

some modification in the subsequent years especially in the upper limit of

the Insured amount. Furthermore. out of the total HI coverage the

Mediclaim policy occupies more than 80% of the market share (lRDA.

2003). Hence. the present study has used Mediclaim policy as the sample

space to understand more about Indian HI and people who bought the

Mediclaim policy were identified as the Insured people. and were

interviewed. The list of insured people selected in the fmanclal year of

2003-04 with the aim of getting information from the insured by

incorporating the recent developments in the Indian HI market. Moreover.

as the HI schemes are renewable in nature on an annual basis. taking the

year 2003-04 will automatically cover the people enrolled into the

insurance schemes In the previous years also. The sample size of the

insured people is 200 that were selected from the Mediclaim Policy

enrolment list of the branches of four public sector companies from the

two districts. Thus. total 100 insured households from the district of

Kasargod and total 100 insured households from the districts of

39

Trivandrum with 25 households from the list of the Mediclaim Policy of

each of the four Insurance Companies were selected as sample of insured

households. The sample size of uninsured is 200 households and it was

randomly selected from the locations where insured households have

been selected. In short. the total sample size is 400 households consisting

of un-insured and insured.

1\vo separate questionnaires were administered for both insured and

uninsured. and household heads (decision-makers) were interviewed.

Data collection was carried out in August to December 2004. Similarly.

peer group discussions were organized with insurance offiCials with the

help of a semi-structured interview schedule.

2.7.1.2. ECCP Household Data on Micro Health Insurance (MIll)

The European Union Cross Cultural Program (ECCP) Project on

"Strengthening Micro Health Insurance Units for the Poor in India" has

conducted a household survey In seven locations allover India. where

micro health insurance schemes are already in place. One of the core

objectives of the ECCP Project is capacity building through promoting

research on HI in India. Hence. the present study utilized the Household

Survey Data provided by the ECCP project. The survey was conducted in

the locations of 7 MHI units operating in 4 Indian States. viz. Maharastra.

Karnataka. Tamilnadu. and Bihar and collected information from both the

Health Insured (through MHIUs) and Health Uninsured. hereafter called

as Insured and NonInsured. The non insured households are selected

from the same location where the insured households are. The total

sample size is 4931 households consisting of 2453 (49.7%) insured from

MHIUs and 2478 (50.3%) uninsured from the locations where insured

people were selected. However. the present study has considered only 5

MHI. they are: 1) YESHASWINI TRUST-Karnataka. 2) KARUNA

TRUST-Karnataka. 3) DHAN-Tamil Nadu. 4) UPLIFT-Maharashtra. 5)

VHS-Taml! Nadu. Thus. the total sample size is 3523. consisting of the

1755 insured households and 1768 non insured households.

40

Table 2.3 Number of Households from the locations of Micro Health Insurance Units

MHI Units Insured Non Insured Total Households Households Households

Karuna Trust 352 348 700 (Karnataka) Yeshaswini Trust 346 354 700 (karnataka) o HAN (Tamil 360 362 722 Nadu) UPLIFT 347 353 700 (Maharastra) VHS (Tamil 350 351 701 Nadu) Total 1755 1768 3523 Source: ECCP Data

Further. the present study classified the MHI schemes as Rural MHI and

Urban MHI for convenience of making analysis and interprets the results.

The Karuna Trust. Yeshaswini Trust. DHAN are located in rural areas

and labeled as Rural MHI units and the UPLIFT and VHS are located in

the urban areas and labeled as Urban MHI units. However, such a

classification has limitations too. These schemes have different mode and

structure of operations. Nonetheless. they differ in the benefit package,

level of premium subsidy and the target populations. So Interpretations of

the results are done as Rural MHI and Urban MHI and also in terms of

each of 5 MHI units.

2.7.1.3 Primary Data on Clients Preferences on Health Insurance Benefits (Choosing Healthplans All Together (CHAT-I)

A deCision tool called Choosing Health-plans All Together (CHAT-I) was

experimented in various locations in India to measure the preferences of

the people for HI benefits, which Is a modification of the CHAT exercise

that was previously designed to allow groups to choose HI packages In the

United States. Groups of village members were recruited from the

following locations of two Indian states: Karnataka and Maharashtra.

41

Table 2.4 Locations of the Study Population

STATES LOCATIONS NUMBEROF HOUSEHOLDS

V.P. Hundi 16 Madavadi 24 Kiragasur 26

KARNATAKA Banave 12 Hiriyuru 12 Beedanahalli 24 Kempalahnahundl 24 Gokhale Nagar 12 Karve Nagar 22 Janata Vasahat 25

MAHARASHTRA Kashewadi 23 Ter 17 Wakerwadi 12 Dhanegaon 13 Chata 13 Boranlani 14 Musti 13

A total of 302 members participated in 24 sessions during November­

December 2005. These participants consist of both Micro Health Insured

and Noninsured.

2.7.2 Methodology of Present Research

The study makes an attempt to understand the HI mainly from a demand

side perspective although considerations have given to the supply side.

Thus. the study has used a sample of the insured people from both MHI

and PHI schemes. Moreover. a sample of uninsured people also has been

drawn from the same locations where the sample of Insured people was

drawn from. In the study, the insured people of both MHI and PHI are

called as 'MHI insured' and 'PHI insured', respectively. In the present

study, the uninsured people from the locations of MHI schemes are called

'MHI uninsured' even though they do not have any relation to the MHI

Units other than living In the same geographical locations where MHI

insured also live. Similarly, the uninsured people from the locations of

PHI Insured are called as 'PHI nonlnsured' (or uninsured) people.

42

The main issue addressed in the study is the scaling up process of HI on

an equitable basis and identification of various contributory factors for

such an outcome. On a welfare point of view. scale up of HI should also

ensure coverage for the poor people too, which may increase the access to

health care for the weaker sections of the society. First of all. we analyze

the equity aspects of HI coverage. In this regard. the study examines the

nature of PHI and MHI coverage to understand whether these schemes

have given coverage to the weaker and disadvantaged sections. Therefore,

the representations of weaker and disadvantaged sections in various HI

schemes are investigated. As the weaker and disadvantaged sections of

Indian society are heterogeneous groups. it is difficult to put them under

one entity for the sake of analysis. However, they live at their subsistence

and belong to the informal sector. One of the common features of these

populations is their low level of material well-being. The material well­

being of a household can be measured by its annual per capita household

income. Hence, in the present study. we define the weaker sections and

disadvantaged sections of Indian society as those who have very low level

of income. In this context. equity in HI coverage is defined as the situation

of having HI for the low-income people.

To understand the equity aspects of HI coverage, the analysis is

performed in terms of various groups. The households are classified in to

6 categories of income groups on an income scale starting from 'Destitute'

to 'Wealthy' based on their annual per capita household incomes. These

income classes are again merged in to 3 categories as 'Low Income Class

Households', 'Middle Income Class Households' and 'High Income Class

Households' for further analysis.

43

Figure 2. 1 Classification of households into different income classes

1111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111111

... ... ... ... ...... Destitute Extreme Poor Moderate Poor

"

Vulnerable Non-Poor

Low Income Class Middle Income Class

Non-poor Wealthy

High Income Class

Taking a cut off point of income for the claSSification of the households

into the above six categories is a tricky issue. The official estimates (GOI.

2006) state that around 26% of the Indian population lives below poverty

line and the cut off point of income has been defined. from time to time. in

terms of calortes of food. As poverty is understood as a multi-dimensional

problem than simply a matter of starvation. the offiCial poverty lines - Rs.

368 and Rs. 559 per person per month for rural and urban areas - are

highly unsatisfactory. A recent study by Guruswami and Abraham (2006)

estimated the necessary income to lead a normal life and argued that the

poverty line in India should be Rs.840 per cap1ta per month. At this

expenditure level. nearly 69% of the total and 84% of the rural population

lives below the poverty line. However. we do not go for any estimate on

defining poverty line for the present analysis, but adopt a simple

classification of the population into a range of low income to high income

class.

Therefore, for the present analysis. the household data set is re-arranged

by sorting the households of the MHI locations in an ascending order on

the basis of their per capita household incomes. which has resulted in the

ordertng of the households from the lowest income class to highest

income class. Accordingly, the first 16.7% from the income households

are called as "Destitute'. and, next 16.7% of the households as 'Extreme

poor'. Similarly. the remaining households are also arranged with

intervals of 16.7%. The last 16.6% of the households who belong to the

44

highest income households are namt'd as 'Wealthy'. These lncomt' !'(roups

are used for the present analysis. To recall. this exercise is perlormed on

the selected household data set only, and no way related to the

population, The outcome from the abo\'e classification based from the MHI

locations are also applied for sorting the PHI households Into dIfferent

income classes. The following table summarizes the methods of orderinf,!

of the households into different income groups.

Table 2,5 Classification of households in to various income classes and cut-off points

Income Class Percentage Income group of Households (in Ra)

Destitute 16.7 0-4800 Extreme Poor 16.7 4801-7200 Moderate Poor 16.7 7201-9000 Vulnerable Non-Poor 16.7 9001-12000 Non-Poor 16.7 12001- 1 7200 Wealthy 16,6 I 720 1 and abo\'e Total 100 -

The above table shows the income group for each category of populatJon

with an inter;al of 16,7% sample population, starting from 'Oeslltute- to

'Wealthy'. Accordingly, for example, the 'Destitute' who constitutes the

first 16,7% of the populaUon belong to the income group of Rs 0-4800,

and '\Vealthy' who constitutes the last 16.6"() belong to the incolllt' group

of Rs. 17201 and above. In the present analysis, we use the above income

c1assificaUon to define various income groups from 'Destitute' to 'Wealthy',

The extent of equity in HI cO\'erage is examined by comparing the

insurance cO\'erage status of !\1HI clients \\ith the unlnsured people of the

same locations of each MHI units, and the PHI cHents to the uninsured

people from the locations from where PHI dit'nts art' recrutted. And

Simultaneously, compartson has also been mack between \'arious income

groups of MHI insured and PHI insured. The HI cow rage by lncome class

is anal\'zed In terms of both lnter Income class and lntra lncome classes.

After Investigating the representation of \'aJ1ous lncome groups In the risk

pool by considering the insured households alone. the analysis is

extended to understand how the insured populations differ from the non­

insured population of the respective location of the HI schemes. It is done

by taking the case of both the Insured and the Non-insured households.

The mean income of both the Insured and the Non-insured within each

income class is compared to understand the HI coverage within each

income class (that is. intra-class). Further. the premium burden of MHI

and PHI schemes on households are examined.

Further. the determining factors resulting in equity and inequity in MHI

and PHI schemes are examined and analyzed also as factors contributing to

the scale up of HI. The determinants of HI coverage (and therefore of the

determinants of equity) in MHI schemes are analysed. The significance of

various factors such as Level of Education. Household Size. membership

status in Self Help Groups/Other Community Organizations (broadly

termed as SHGs) membership status and Health Risk on the probability to

have HI coverage by each income class in each of the selected MHI schemes

are examined. The above mentioned factors are analyzed under the frame

of push and pull factors. To distinguish the pull and push factor aspects of

household income. the different income classes and their membership

status in SHGs are analyzed; first a comparison of the SHGs membership

status of the Insured across different income class is done. and secondly.

comparisons of the SHGs membership between the Insured and the Non

Insured within and across different income classes are performed. The role

of education and household size on the decision to buy HI is examined after

classitying the education and household size under various income classes.

Binary Probit models are used to estimate the significance of both the push

and pull factors on the probability to buy HI of households.

After analyzing the determinants of HI coverage in MHI schemes. the

analysis is extended to investigate the determinants HI coverage in PHI

schemes (and therefore of the determinants of equity). Adverse Selection is

an important constraint for Insurers to sell HI schemes. It acts as a factor

adversely affecting the scale up process of HI. First the issue of selection

bias is analyzed in PHI schemes. In the scale up process of PHI scheme.

the income maximizing behavior of insurance agents and its effects on

46

equity aspects of HI coverage and selection bias in PHI schemes is

presented in the form of a simple theoretical model. The analysis is

performed in a theoretical frame of asymmetric information with special

reference to PHI schemes. The factors contributing to asymmetric

information on HI poliCies are examined by analyzing the level of

familiarity of HI and other forms of insurance. In this context, a

comparison of different aspects of awareness on different types of

insurance in general and HI in particular between the Insured and the

Non Insured of the PHI schemes is done. Further, the familiarity of the

people with different forms of insurance is measured by comparing the

'Other Insurance enrolment statuses' of both the PHI Insured and the Non

Insured. Insurance Agents are the main information dissemination

channel for HI schemes. We analyze the role of information asymmetry on

HI schemes between the Client and Insurer on the selection bias through

his role. In PHI schemes, the role of 'Insurance Agents' in the scaling up of

HI and 'Selection Bias' is examined by analyzing utility maximizing

behavior of the 'Insurance Agents'. The theoretical model so developed is

empirically tested. Both bivariate and multivariate analyses including the

Probit regression model are applied to estimate the significance of various

factors including 'Selection Bias' affecting the decision to buy HI.

After the investigation of the determinants of PHI coverage. including

testing of 'Selection Bias', the analysis is extended to examine the

presence of 'Selection Bias' in MHI schemes also. Firstly, it is investigated

on whether there is any Adverse Selection or not in each MHI. Thereafter.

presence of adverse selection by each income class in each of the MHI

scheme is investigated. Moreover, the role of SHGs in 'Selection Bias' is

also examined. Binary Probit regression models are used to examine the

factors affecting the probability to opt MHI by the people.

Finally, the question whether the prevailing HI schemes are attractive or

not has been examined. If the prevailing HI schemes do not reflect the

preferences and requirements of the clients, it seems to be unattractive

and would lead to low level of health insurance coverage. Therefore,

whether the prevailing HI schemes match the preferences of the people or

not are investigated. The preferences of the people for various health care

47

benefits are measured under two situations: 1) Without Budget

constraint, and 2) With Budget constraint. The preferences without

budget constraint are measured from the responses of those who are

willing to pay for health insurance, for which we use the ECCP data. And,

the preferences with budget constraint are measured with the help of a

decision tool called CHAT exercise (Choosing Healthplans All Together).

CHAT is a simulation exercise designed to allow persons to define their

own benefit package within constraints of limited resources. Thereafter,

we analyze whether the preferences of the people match with various

benefits on offer with the available HI schemes in India.

48

CHAPTER THREE

THE EQUITY ASPECTS OF THE HEALTH INSURANCE COVERAGE IN INDIA

3. I. Introduction

The main objective of this chapter is to analyze the equity aspects of HI

coverage of various forms of HI schemes in India. To what extent the HI

schemes have fairness in providing coverage to all sections of the society,

especially the weaker and disadvantaged sections? The equity aspects of

the HI coverage in both the MHI and PHI schemes are analyzed

separately. The HI coverage in terms of income class is analyzed in terms

of both inter income class and intra income class. The mean income of

both the Insured and the Non Insured within each income class is

compared to understand the health insurance coverage within each

income class (that is. intra-class). After investigating the representation of

various income groups in the risk pool by considering the insured

households alone. the analysis is extended to understand how the

Insured populations differ from the Non Insured populations of the

respective location of the health insurance schemes.

3.2. Equity in Health Care and Equity in Health Insurance Coverage

Equity here is meant on social justice or fairness; it is an ethical concept

grounded in principles of distributive justice. Equity in health implies that

ideally everyone irrespective of gender. income. age. race. religion etc.,

should have a fair opportunity to attain their full health potential and.

more pragmatically. that no one should be disadvantaged from achieving

this potential. if it can be avoided (WHO. 1985). A health disparity is

inequitable if it is systematically associated with social disadvantage in a

way that puts an already disadvantaged social group at further

disadvantage (Starfield. 2001: P Braveman and S Gruskin. 2003). Equity

in health care is a multidimensional phenomenon and is defmed in terms

49

of equality of access to health care, equality of utilization of health care.

distribution according to need of health care, and equality of health

(Whitehead, 1992),

Both macro and micro studies on the use of health care services show

that the poor and disadvantaged sections such as trtbal populations are

forced to spend higher portions of their income on health care than the

better off (Visalia and Gumber, 1994; Gumber, 1997). In theory,

government provision of health care should cover the poor, but in practice

it often does not. About 80% of the public health subSidy goes to the

richer sections of the society (Mahal Ajay, 2002). The weakest and

disadvantaged sections of Indian societylS consist mainly of daily wage

workers including agricultural labourers, construction workers and

domestic workers, farmers. tribal populations etc., especially women,

children and the elderly in the family of these population groups. The poor

suffer from far higher levels of ill health, mortality, and malnutrition than

do the better off. They are more susceptible to ill health and. are

particularly vulnerable in regard to health status and health care. They

live and work in unhygienic conditions and have poor nutrition levels all

of which make them susceptible to both infectious and chronic diseases,

In India. as elsewhere, the poor die earlier and have higher level of

morbidity than the better off (World Bank, 2003). Women and the aged

populations are also more vulnerable in regard to health status and their

problems are aggravated, especially when they belong to the poorest

segment of the society. Poor women are most vulnerable to diseases and

ill health due to unhygienic living conditions. heavy burden of child

bearing. low emphasis on their own health care needs and severe

constraints in seeking health care for themselves. Similarly, the aged

population who are more prone to illness also comes under the above

category. The demand for medical care is likely to increase with age. at

least beyond the middle years to compensate depreciation in the overall

stock of health due to ageing. Apparently, the elderly people of the low-

15 Examples of more and less advantaged social groups include socioe-conomic groups (typically defined by measures of income, economic assets, occupational class, and/or educationallevell. racial/ethnic or religious groups, or groups defined by gender, geography, age, disability. sexual orientation, and other characteristics relevant to the particular setting.

50

income households will be more deprived of health care than their high­

income counter partners. The health of the poor must thus be a matter of

major concern for everyone committed to equitable development, from

policy makers to service providers.

The financial constraint is the main barrier for access to health care lS and

perhaps one of the leading factors for inequity in health care in India. A

majority of weaker sections of the society earns its livelihood from the

informal sector. They neither have fixed employer-employee relationship

nor do they obtain statutory social security benefits. which imply that

they do not get health care benefits. paid leave for illness. mateITlity

benefits. insurance. old age pension and other social benefits. As the

health care burden in the form of out-of-pocket expenses fall more on

those least able to pay. HI is seen as the viable solution (Churchill 2006):

we can expect that equity (fairness) in HI coverage in the form of

enrollment to HI schemes to all sections of the society irrespective of

gender. income. age. race. religion etc .. would increase equitable access to

health care and thereby enhance the equity objectives in health care

system in India. Some of the studies in developing countries context

(Juetting. Joharmes. 2001) have revealed that health insurance coverage

has enhanced access to health care and financial protection in rural

areas. Moreover. on an equity angle. within the risk pool. benefits are

provided on the basis of need rather than income class. Payments go to

the sickest people. and. because lower income and less-educated people

tend to be Sicker. they also have the potential of benefiting more from

insurance claims (Mc Greevy. 1990).

16 40 percent of the hospitalized having had to borrow money or sell assets. during the decade 1986-96, there was a doubling of persons unable to seek healthcare due to finanCial reasons (NSSO 1996). and almost 24 percent of the hospitalized Indians fall below poverty line because they are hospitalized (Peters et al 2002). Furthermore, around 24 per cent in rural and 21 per cent in urban areas have cited their lack of fmancial capacity as the reason for not seeking treatment ([NSSO 1996).

51

Figure 3.1 Impact of Equity in health insurance coverage on Equity in Health

Equity in Health Insurance Coverage

Equity in Access to health care

Equity in health

3.3. Health Insurance Schemes and their Target Population

In the following section. we figure out the target population of both MHI

and PHI schemes. We have noted in the previous chapter that there are

around 20 MHI schemes operating in some specific regions of India

including both rural. semi urban and urban locations. Similarly. around 11

General Insurance Companies comprising of both the public and private

sectors offer different HI schemes through their branches. which are

operational allover India. The figure below shows the target population of

both PHI and the selected MHI schemes.

52

Figure 3.2 Target Population of the Health Insurance Schemes

l Health Insurance Schemes I ~

MHI schemes I PHI schemes I ~

+ + Public Prlvate sector [Tata A1G. sector Royalsundaram.Cholam

[NIAC. NIC. an-dalam MS. ICICI

OIC. UICI Lombard. and BaJaj A1lianz

1) Low income + .. Households 1) All 1) All

sections of sections of 2) Farmers the society the society

3) Women 2) Both 2) Both members in urban and urban and SHGs and their rural rural family members Population Population

It is obvious from the above figure that most of the MHI schemes aim at low­

income populations, especially farmers and woman members of Self Help Groups

(SHGs). But the PHI schemes that are being offered by both public and private

sectors do not put any restriction on their target population. In other words, their

policies are open to everybody who has the ability to pay the premium.

Let us now examine whom they have actually covered. In the following

sections we investigate the nature of HI coverage with the help of the

empirical estimates by testing to what extent low-income classes are

covered by various HI schemes. Both inter-income and intra income class

distribution of HI coverage has been examined. In Inter income class

analysis. a comparison of the proportionate representation of various

income clF\sses ranging from 'Destitute' to 'Wealthy' within Insured and

Non Insured and between Insured and Non Insured is performed, In Intra

53

income analysis, each income class is considered and examined whether

the lower bottom of each income group is covered or not as compared to

the Non Insured.

3.4. Inter-income Class Distribution of Health Insurance Coverage

Let us examine to what extent health insurance schemes have covered

vartous income classes, The Table 3.1 explains the proportion of different

income groups in the HI coverage of each scheme. We have presented both

the Insured and the Non Insured households: first we compare the relative

proportion of the Insured in each risk pool and thereafter the proportion of

the Insured to the Non Insured of the same locations of each HI scheme.

Table 3.1: Proportion of the households across different income Classes

Income Class

Rural MHI Urban MHI schemes schemes PHI schemes

Non Non Non Insured Insured Insured Insured Insured Insured

Destitute 9.26 8.93 30.13 25.57 0.00 0.00 Extreme Poor 14.93 17.67 8.90 13.64 3.50 0.00 Moderate Poor 14.27 19.45 16.21 17.05 6.50 0.00 Vulnerable Non-Poor 18.81 18.98 15.35 12.22 21.00 0.00 Non-Poor 18.90 16.54 19.23 20.17 28.50 3.50 Wealthv 23.82 18.42 10.19 11.36 40.50 96.50 Total 100 100 100 100 100 100

( 1058) ( 1064) (697) (704) (200) (200)

Source: Field study and ECCP data Figures in the parentheses show the number of observation.

It can be seen from the above table that the PHI is highly biased towards

the Higher Income Classes. It has completely excluded three income classes

of the bottom, and highly represented by the 'wealthy" people. It is a clear

case of horizontal income solidarity among the high income classes where

the healthy rich pay for the health care costs of the unhealthy of the same

54

group. Nonetheless, it was observed during the field study that around 93%

of the PHI clients are from the urban locations. Therefore, it can be inferred

that the PHI schemes are not only biased towards the urban population but

also pick the relatively high income classes in the urban areas. The main

message from this trend is that the PHI scheme has failed to act as an

equitable financing strategy to give access to health care for the poorer

sections of the SOCiety. As against this, compared to the PHI scheme, the

MHI schemes have ensured fairness in giving access to the low-income

people. Among the two types of MHI schemes that we have selected for the

present analysis, the 'Rural MHI' has covered less proportion of the

'Destitute' and 'Extreme Poor' (around 25%) as compared to the 'Wealthy'

and 'Non-poor' (around 43%) where the 'Urban MHI' schemes have covered

more 'Destitute' and 'Extreme Poor' (around 39%) as compared to the

'Wealthy' and 'Non-poor' (around 29%). And also, both schemes have a fair

representation of middle-income class consisting of 'Moderate Poor' and

'Vulnerable Non-Poor' in its risk pool. These are the situations when we

consider only the insured people. Looking into the proportion of the non

insured population in the respective locations of each MHI schemes, it can

be seen that the 'Rural MHI' schemes have covered relatively less number

of 'Low Income' households and have covered relatively more number of

'High Income' class households, and the 'Urban MHI' schemes have covered

relatively more proportion of 'Low Income' class households and covered

relatively less proportion of 'High Income' class households.

As we have merged the MHI schemes like 'Rural MHI' and 'Urban MHI'

schemes and examined the representation of various income classes, let

us also examine the representation of various income classes within each

MHI unit.

55

Table 3.2: Proportion of the households across various income classes in each 'Rural MIll' schemes

Rural MHI schemes Income Karuna trust Yeshaswini trust Dhan Class Non Non

Insured Insured Insured Insured Insured Destitute 13 8 5 8 9 Extreme Poor 23 17 8 19 14 Moderate Poor 14 19 12 20 16 Vulnerable Non-Poor 19 17 19 20 18 Non-Poor 14 19 25 18 18 Wealthy 16 20 31 15 25 Total 100 100 100 100 100

Source: ECCP data

Non Insured

10

17

19

20 13 21

100

Table 3.3: Proportion of the households across various income classes in each 'Urban MIll' scheme

Urban MHI Schemes UPLIFT VHS

Non Non Income Class Insured Insured Insured Insured Destitute 13 14 47 38 Extreme Poor 6 14 12 13 Moderate Poor 19 18 13 16 Vulnerable Non-Poor 22 14 9 10 Non-Poor 26 25 12 15 Wealthy 14 14 7 9 Total 100 100 100 100

Source: ECCP data

Among the three 'Rural MHI' schemes, Karuna Trust has covered relatively

more of 'Low Income Class' (36%) consisting of the 'Destitute' and the

'Extreme Poor', and also relatively more proportion of 'Low Income Class' as

compared to the Non Insured population of their locations. Similarly,

among the 'Urban MHI' schemes, the VHS has covered more proportion

(59%) of "Low Income Class'.

56

So far we have discussed the HI coverage across various income classes,

which was an inter-income class comparison within the Insured and

between the Insured and the Non Insured households, An equally

important issue is to examine whether any disparity in the HI coverage

within each income class, that is intra-class disparity, exists or not. The

next section deals with this issue.

3.5. Intra-income Class analysis of Health Insurance Coverage

One of the main findings that stems from the above discussion is that the

PHI schemes are more biased to the higher income classes, and the MHI

schemes more or less have covered the low income population where the

'Urban MHI' schemes have absorbed more proportion of the low income

population as compared to the 'Rural MHI' schemes. Therefore, our next

question is whether relatively low income or high-income households

within each income class are covered or not. Accordingly, the following

hypothesis is posed and tested: "The insured population in each income

class does not belong to the lowest income strata of the same income

class".

In the tables below, the proportion of insured and non-insured within each

income class is presented. In Table 3.4, we consider Rural MHI. Urban MHI

and PHI schemes. First. let us have a look at PHI scheme. Within each

income class, PHI scheme excluded the low and middle incomes

completely; this rmding reconfirms our previous rmding. In the MHI

schemes, when we consider the very income group, namely 'Destitute', the

proportion of insured households are 51 % and 54% in Rural MHI and

Urban MHI schemes, respectively. But the same trend is not visible in the

next top income class "Extreme Poor'.

57

Table 3.4 Proportion of the Insured and Non Insured households across different income classes in Rural MIll, Urban MIll and PHI schemes

Rural MIll Urban MIll PHI Scheme schemes schemes N= 400 (Karuna. (UPLIFT. VHS) Yeshaswini. Dhanl N= 2122 N=1401 Non- Insured Non- Insured Non- Insured insured insured insured

Destitute 49 51 46 54

100 a Extreme 100 0 Poor 54 46 61 39 Moderate 100 0 Poor

58 42 52 48 Vulnerable 100 a Non-Poor 50 50 45 55 Non-Poor 47 53 51 49 89 11

Wealthy 44 56 53 47 30 70

Source: Field study and ECCP data

In the table below. the proportion of insured and non-insured within each

income class for each MHI unit is presented.

Table 3.5 Proportion of the Insured and Non Insured households in various MHI schemes across different income classes

Karuna Yeshaswinl Dhan UPLIFT VHS trust trust N=722 N=700 N=701

N-700 N-700 Non Ilsured Non- nsured !'Ion nsured Non· nsured ~on- nsured

nsured nsured nsured nsured nsured

pesUtute 38 62 63 38 52 48 52 48 44 56

~xtreme Poor

42 58 72 28 54 46 71 29 52 48

Moderate Poor

57 43 63 38 55 45 49 51 55 45

flulnerabl, ~on-Poor

46 54 52 48 53 47 40 60 53 47

~on-Poor

57 43 42 58 43 57 50 50 55 45

r-vealthy 55 45 33 67 45 55 51 49 57 43

Source: ECCP data

58

In the case of 'Destitute', the Karuna Trust and VHS have covered more

than proportionately, that is 62% and 56%, respectively. In the case of

'Extreme Poor' it is 58% and 48%, respectively. Although other MHI

schemes have less proportion of 'Destitute' and 'Extreme Poor' in its risk

pool. it is not that much significant, especially when we compare these with

PHI scheme.

In the discussion above, we have compared the health insurance coverage

across different income classes by taking household as the unit. We do not

know whether all members of the household are covered or not. Equal to

having HI coverage for the low-income households, it is quite important to

have such coverage for all the household members. Is there any disparity in

the proportion of insured people to their household size across different

income classes? For this, we track the proportion of the insured people in

each household across different income classes and make a comparative

analysis of the HI coverage by household member of the low-Income class

to the high-income class.

The proportion of insured members to the household size is denoted by

the 'Household Size-Insured ratio' (HSI ratio) and measured as,

'HSI ratio' = Household Size I number of insured people

The following table presents the mean value of the 'HI ratio' by each

income class of different HI schemes.

Table 3.6 The mean value of the 'HI ratio' across different income classes

Income Rural Urban PHI Class MHI Karuna Yesbasvinl MHI

Trust Trust Dban UPLIFT VHS Destitute 0.62 0.82 0.40 0.44 0.80 0.76 0.81 0.0

Extreme Poor 0.66 0.82 0.47 0.50 0.67 0.63 0.68 0.0

Moderate Poor 0.64 0.83 0.56 0.52 0.79 0.84 0.72 0.0

Vulnerable 0.63 0.76 0.54 0.57 0.82 0.84 0.77 0.00

Non-Poor Non-Poor 0.67 0.83 0.59 0.66 0.84 0.90 0.71 0.82

Wealthy 0.68 0.89 0.57 0.68 0.77 0.81 0.69 0.83

Source: Field study and ECCP data

59

From the above table we can see that mean value of the proportion of

Insured members across different income classes vary considerably. In the

'Rural MHI' schemes. all the three MHI units have relatively low value of 'HI

ratio' for the 'Low Income Class' as compared to the 'High Income Class'.

which means that the proportion of Insured member in each 'Low Income

Class' household is less than that of 'High Income Class'. In the case of the

'Urban MHI' schemes. the 'HI ratios' of the 'Low Income Class' households

are almost equal or higher than that of the 'High Income Class'. However.

on the whole. we can see that there is not a significant difference on the 'HI

ratio' between the 'Low Income Class' and 'High Income Class' across all

the MHI schemes. Thus. if we consider the 'HI ratio' as an indicator to

assess the equity aspects of health insurance coverage. the results are at

par with the previous findings.

3.6. Econometric Estimation on the probability to have MIll coverage for various income class households

To reconfirm the above findings and to estimate the probability of each

income class in the risk pools. an econometric procedure was adopted. In

order to analyze factors associated with insurance coverage in a

multivariate context. families with health insurance coverage were

distinguished from families with no health insurance coverage by defining a

dichotomous variable by assigning value '1' for families with health

insurance and '0' for families with no health insurance. The empirical

analysis made use of a binary Probit regreSSion model. With the Maximum

likelihood function. the probability for each income class to have health

insurance is estimated,

In the follOwing model. the likelihood of having HI by each income class in

each MHI unit is estimated. However. the estimation of probability between

various income classes in PHI scheme is not attempted here maInly

because we do not have enough representation of various income groups as

the PHI scheme is highly biased to the high income class. The table below

explains the description of the variables. We have classified the households

into six income groups staring from the 'Destitute' to the 'Wealthy'.

60

according to their annual per capita household income and recoded into six

dummy vartables. To avoid the problem of dummy vartable trap. one

dummy vartable is omitted in the estimation. namely the 'Wealthy' and

used as the reference category. In addition to the six income classes. we

have included some other relevant vartables also in the model such as

household size. highest education attained by the household. highest

education attained by the household head. health risk of the household.

and SHG membership. but we do not use these variables for interpretation

in this section. The purpose of including these vartables in addition to the

six income classes Is peculiar to the feature of nonlinear regreSSion models.

In nonlinear models the effect of a change in a variable depends on the

values of all variables in the model and is no longer simply equal to one of

the parameters in the model.

Table 3.7 Def'mition of variables

Variables Description Health insurance 1 if the household has health insurance;

o Otherwise Hhsize Household size High education Highest education attained by any family member.

measured in terms of years of schooling Head education Highest education attained by the head of the family.

measured in terms of years of schooling Health risk At least one member of the households is with high

health risk =1; o otherwise

SHG membership Household is member of Self Help Groups= 1; o otherwise

Destitute Household belongs to the category of Destitute= 1; o otherwise

Extreme poor Household belongs to the category of Extreme poor =1; o otherwise

Moderate poor Household belongs to the category of Moderate Poor=l; o otherwise

Vulnerable nonpoor Household belongs to the category of Vulnerable non poor = 1; o otherwise

Nonpoor Household belongs to the category of Non poor -1; o otherwise

Wealthy Household belongs to the category of Wealthy =1; o otherwise

61

The table below (Table 3.8) presents the result of the Probit model, the

probability to have health insurance in the case of Rural MHI schemes. In

the table, first we present the Probit result of 3 rural MHI schemes

together, and thereafter the Probit results for each Rural MHI schemes are

presented. The models explain the likelihood of each income class to have

HI as compared to the 'Wealthy' class. As mentioned before, we have

included some more other factors as independent variables in the model.

We have tested for the multi-collinearity problem (which may result in

inconsistent estimate) in the model between two variables such as the

highest education attained by any member of the household and the

highest education attained by the head of households by estimating the

correlation coefficient between these two variables and found that there is

no multi-collinearity.

It can be seen from the results of the Probit model of the 'Rural MHI'

schemes that there is no statistical significance for the 'Destitute' to have

HI as compared to the ·Wealthy'. It means that the 'Destitute' has no

significantly different probability to have HI as compared to the 'Wealthy'.

Further, it can be seen from the results of the Probit model of the 'Rural

MHI' schemes that the 'Extreme Poor' and the 'Moderate Poor' are 11 %

and 14%, respectively, are less probable to have HI as compared to the

'Wealthy' class, However, when we estimate the Probit model for the three

Rural MHI schemes separately, it can be seen that the 'Destitute' are 17%

and 27% more likely to have HI as compared to the 'Wealthy' in the case

of the Karuna Trust and the Yeshaswini Trust. respectively, but there is

no statistically significant difference between the probability to have HI

between the 'Destitute' and the 'Wealthy', which means that being a

'Destitute' in the location of DHAN MHI scheme has an equal probability

to have HI to that of the 'Wealthy'. Similarly, being a member of the

'Extreme poor' and the 'Moderate Poor' increases the likelihood to have HI

as compared to the 'Wealthy' by about 12% and 2%, respectively, in the

case of Karuna Trust, but reduces by 35% and 25%, respectively in the

case of the Yeshaswini trust, and 16% and 13%, respectively, in the case

of DHAN as compared to the 'Wealthy'.

62

Table 3.8 Probability to have health insurance coverage- marginal effect of Probit model results of Rural MIll schemes Dependent variable: 1 if the household has health insurance; o Otherwise

Rural MHI Karuna Schemes Trust

Variables (Karuna trust, Yeshaswini Trust and DHAN) Marginal Marginal Effect Effect

Hhsize .0011 -.0192 (.17) (-1.55)

Higheducation .0139 .0105 (5. I I)" (2.51)-

Headeducation -.0105 -.0119 (-4.00)- (-2.53)-

Healthrlsk .0998 .0444 (2.98)- (0.72)

SHGmembership -.0000 .0128 (-0.280) (0.33)

Destitute -.0659 .1734 (-1.51) (2.45)-

Extremepoor -.1115 .1251 (-3.10)' (2.02)--

Moderatepoor -.1412 -.0291 (-4.06)- (-0.45)

Vulnerablenonpoor -.0686 .0738 (-1.98)- (1. 16)

Nonpoor -.0410 -.0189 (-1.16) (-0.29)

Y = Pr(lnsured) .4983 .5024 (predict) Log likelihood -1432.77 -471.32

Likelihood Ratio (LR) 65.04 27.74 chi2 ( 10) Pseudo R2 0.0222 0.0286 Number of 2114 700 observations Reference Category: The Wealthy Class Level of statistical significance: - = 1%; •• =5%; _.- ; 10% Values In the parentheses refer to the 'Z' statistics

Yeshaswini DHAN Trsut

Marginal Marginal Effect Effect

.0102 .0368 (0.91) (2.11)--.0254 .0127 (4.57)- (2.26)--.0042 -.0293 (0.92) (-5.57)-.1153 .1483 ( 1.98)- (2.58)-.1538 -.0001 (3.42)- (-0.52) -.2736 -.1144 (-3.93)- (-1.55) -.3541 -.1611 (-6.82)- (-2.58)--.2462 -.1311 (-4.25)- (-

2.16)---.1597 -.0997 (-2.73)- (-

1.69)----.0952 .0041

(-1.60--' (0.06) .4914 .4999

-433.04 -473.49

98.63 49.76

0.1022 0.0499 696 719

The next table presents the Probit model results for the 'Urban MHI'

schemes such as UPLIFT and VHS together in the first instances. and also

separately for the two Urban MHI schemes.

63

Table 3.9. Probability to have health insurance coverage- Probit model results of the Urban MHI schemes

Dependent variable: 1 if the household has health insurance; o Otherwise

Variables Urban MHI UPLIFT VHS Schemes (UPLIFT and VHS) Marginal Marginal Effect Marginal Effect

Effect

Hhsize -.0305 -.0049 -.0119 (-3.41)' (-1.92)" (-0.97)

Higheducation .0069 .0014 .0043 (1. 56) (1. 72)'" (0.64)

Headed ucation -.0016 .0003 -.0124 (-0.41) (0.66) (-2.12)"

Healthrisk .1827 .0109 .2190245 (5.73)- (1.79)'-- (5.17)'

SHGmembership .0000 .0506 .0001 (0.4 7) (2.54)- (0.48)

Destitute .1017 .0039 .1425 (2.06)" (0.65) ( 1.85)'"

Extremepoor -.0497 -.0134 .0420 (-0.84) (-0.91) (0.46)

Moderatepoor .0561 .0054 .0691 ( 1.05) (0.98) (0.80)

Vulnerablenonpoor .1210 .0125 .0536 (2.23)" (1.79)'" (0.57)

Nonpoor .0249 .0040 .0498 (0.49) (0.74) (0.57)

Constant -.0163 -.3726 -.110 1 (-0.10) (-1.49) (-0.43)

Y = Pr(insured) (predict) .4984 .9869 .5031 Log likelihood -939.65 -382.88 -459.14

Likelihood Ratio (LR) 54.55 187.98 36.86

chl2( 10)

Pseudo R2 0.0282 0.1971 0.0386

Number of observations 1395 688 689

Reference Category: The 'Wealthy' Class Values In the parentheses refer to the 'Z' statistics; Level of statistical significance: ' = 1%; ,. =5%; ••• = 10%

In the case of the two 'Urban MHI' schemes taken together, the 'Destitute'

are 10% more likely to have HI as compared to the 'Wealthy". In the case of

the UPLIFT MHI scheme, being a member of the category of the 'Destitute'

does not alter the likelihood to have HI as compared to the wealthy. But in

64

the case of the VHS MHI scheme, being a member of the 'Destitute'

category increases the likelihood to have HI by 14% as compared to the

'Wealthy'.

The main inference that stems from the above discussion is that even

though each MHI scheme has considerable differential bias in giving access

to HI coverage for the low income class, the MHI schemes have a fair

representation of low income class in their risk pool. But the PHI schemes

have excluded the low income class which is a clear indication that PHI has

failed to attain the equity of objectives of the health system. A detailed

discussion of the factors resulting in such an outcome in the case of the

PHI coverage will be done in the next chapter. At present we will investigate

the significance of other factors resulting in equity in the case of the MHI

coverage.

To recall, the equity objectives in a health system is also related to the

income solidarity, that is to the presence of the vertical income cross

subsidization. In such a context, the low income class must be benefiting

from the high income class through the risk pool. When we consider the

case of PHI, there is no such question of such cross subsidization from the

rich to the poor because it has completely excluded the poor from the

coverage. It is a clear case of horizontal cross subsidization among the rich,

so equity objective is not fulfilled. But in the case of the MHI schemes, all

income groups are members of the schemes. However, based on this fact

we can not come to conclusion that there Is vertical income solidarity in the

form of subsidizing the health care expenditure of the poor by the rich,

Apparently, an empirical estimation has to be performed to find out the

level of cross subsidization. In the present study, there is some limitation to

go for such estimation because many of the MHI schemes have not attained

finanCial and system sustainability as they are supported by external

funding, 80 the actuarial premium is perhaps greater than the actual

premium collected from the clients. However, we can see whether the poor

are benefiting through HI or not by comparing the health care

reimbursement from the risk pool between the poor and the rich, but such

an analysis is beyond the scope of the present study. In fact, the MHI

schemes offer a fixed HI benefit package which may not vary according to

65

the ability to pay and willingness to pay the premium by the clients. So

having HI coverage for the low income population itself is an indication of

their access to health care. Moreover. HI is considered as a protection

against the unforeseen financial contingencies due to illness. So having

entitlement for HI coverage for the low income class itself is welfare

promoting as long as such an assurance is there.

3.7. Premium Burden on Households

Similar to the above discussion, another issue which has to be discussed in

this context is the out of pocket payment by the households in the form of

premium to the risk pool. It is important to see whether the low income

classes are getting premium subsidy as compared to the high income

classes. As each MHI scheme is providing a fixed package of benefits with

their schemes, a comparison of per head HI premium paid by each income

class will reveal whether the low income class is getting premium subSidy

or not. The table below presents the percapita annual HI premium paid by

each income class across various HI schemes. We have not included the

case of Kamna Trust in the table because 95.4% of the insured households

pay no premium for the coverage as the scheme is sponsored by the UNDP.

Out of this, the proportion of the 'Low Income Class' consisting of the

'Destitute' and the 'Extreme Poor' are 13.1% and 22.5%, respectively, and

the share of the High income class consisting of the 'Non Poor' and the

'Wealthy' are 15% and 16%, respectively.

66

Table 3.10 Percapita health insurance premium (in Rs.) paid by different income classes

Income Yeshasvini

Class Trust Dhan UPLIFT VHS PHI

Destitute 76.11 85.48 64.46 62.46 0

Extreme 66.19 9l.18 77.78 74.12 0

Poor

Moderate 72.26 77.19 61.78 92.91 0

Poor

Vulnerable 74.77 84.20 65.38 9l.32 420

Non-Poor

Non-Poor 70.65 80.36 54.72 95.36 435

Wealthy 90.68 88.46 67.91 88.41 465

Source: Field study and ECCP data

From the above table we do not fmd any substantial difference between the

HI premiums paid by different income classes. The premium per person

may be related to age. gender. family size, region. occupation. length of

contract period. individual or group contract period. the level of deductible.

the sum insured. and health status at the time of enrolment and health

habits such as smoking. drinking, exercising (Abel-Smith, 1992). But the

MHI schemes charge premium only on the basis of the principle of

'Community Rating' Instead of 'Experience Rating', which means that they

charge same premium rate to all members of the society Irrespective of

their income or health. As against this, the PHI schemes charge premium

on the basiS of risk rating where risk is measured in terms of the age of the

clientsl7. The table below shows how do premium changes as the insurance

amount and the age of clients in PHI schemes.

17 Even though the insurance companies say that the schemes are open till the age of 80, in practice they give enrollment for males till the age of 60 and for females till the age of 55 years, and the people above these age group are entitled to renewal provided that they do not have any claim history.

67

Table 3.11 Mediclaim policy premium (in Rs.) table

Insurance Age Age 35- Age Age Age Age Age amount 0-35 45 45- 55- 65-70 70-75 75-80 (in Rs.) 55 65 15000 213 232 331 379 427 427 551

(1.42) (1.55) (2.21) (2.53) (2.85) (2.85) (3.67) 50000 676 736 1055 1199 1344 1441 1778

(1.35) (1.4 7) (2.11) (2.40) (2.69) (2.88) (3.56) 100000 1310 1425 2039 2322 2598 2784 3445

( 1.31) (1.43) (2.04) (2.32) (2.60) (2.78) (3.45) 150000 1920 2087 3004 3425 3838 4162 5305

(1.28) (1.39) (2.00) (2.28) (2.56) (2.77) (3.54) 200000 2469 2683 3900 4458 5010 5471 7097

(1.23) (1.34) (1.95) (2.23) (2.51) (2.74) (3.55) 300000 3444 3743 5553 6388 7214 7951 10542

(1.15) (1.25) (1.85) (2.13) (2.40) (2.65) (3.51) 400000 4297 4670 7069 8179 9281 10294 13849

(1.07) (1.17) (1.77) (2.04) (2.32) (2.57) (3.46) 500000 5151 5598 8585 9970 11348 12636 17156

( 1.03) ( 1.12) (I. 72) (1.99) (2.27) (2.53) (3.43)

'Recalculated from the original premium table of the MedicJaim policy. The figures in the parentheses show the proportion of the premium to the insurance amount.

One of the main messages from the above premium table of the Mediclaim

policy (PHI) is that premium increases with the age of the clients and the

insurance amount to be chosen by the clients. The lowest insurance

amount. which can be bought by a client is Rs. 15,000 and the maximum

amount is Rs. 5,00,000. Therefore, if we consider only the income of the

household as factor determining the demand for HI coverage, we can infer

that a low income client may buy a low amount of HI coverage as compared

to a high income client who may buy a higher amount of HI coverage. In

this context, looking into the table, a client (in the age group of 0-35)

buying HI coverage for Rs.15,OOO has to pay 1.42% of the insurance

amount in the form premium whereas a client buying the coverage for

Rs. 5,00,000 has to pay only 1.03% of the insurance amount as premium.

Similarly. a client in the age of group of 55-65 has to pay 2.53% of the

insurance amount to buy Rs.15, 000 whereas a client in the same age

group has to pay only 1.99% of the insurance amount as premium. In

short. the clients who are going for higher levels of insurance amount have

to pay the premium at a regressive rate. Even though PHI schemes are not

explicitly making any discrimination on premium on the basiS of income,

but impliCitly they are. The figure below substantiates the above argument.

68

Figure 3.3 Mediclaim premium for various age groups

3 .s E 2.5

-------::::J - • E c: • ::::J • Q) 2 • .... 0 • Q.e --+- Age 0-35 ~ <a c: 8 1.5 .. , --Age 55-651 ~ c: , ::::J <a

, , Age 65-70 1/1 .... 1 - • • c: ::::J •

- 1/1 - c: 0-0 0.5

'';::: <a ....

0 --- - -T- ------------,

0 100000 200000 300000 400000 500000 600000

Insurance Amount (Rs)

Therefore. we can infer from the above discussion that in assuring equity

in access to health care for the low income population. PHI schemes are

not a solution in the current form.

3.8. Chapter Summary

From the above discussion. it is clearly revealed that MHI schemes are

more eqUitable than PHI schemes in the form of giving coverage to the low

income and middle income households along with the high income

households. From a supply side perspective. the MHI schemes have

emerged as a solution to the high health care burden of the low income

people. Many of the MHI units are giving health insurance schemes at a

premium (partially subsidized) below the actuartally fair premium; and

they are able to do so with the financial support from some funding

agencies. negotiations with insurance companies and hospitals and also

surplus from its other kind of services. However. one of the important

objectives of these MHI units is to gradually attain financial sustainability.

Although the MHI schemes are targeting mainly the poor people. one way

to achieve financial sustainability is to include the high income people

69

also in to the risk pool to make it a sound risk pool. This may be one

reason for having both low income and high income people as members of

MHI schemes in the form of vertical income cross subsidization. In

contrast to this. the PHI schemes are market products and open to

everybody which may target mainly high income people even though

insurance companies are claiming it as social security schemes. As the

premium of PHI schemes are relatively high. it is perhaps one of the

reasons why do only the high income people with their ability to pay are

buying for PHI schemes l8 . From this perspective. we can see that the low

income people can not afford the actuarially fair premium and one of the

reasons for the low income people not buying PHI is fuelled by their low

ability to pay of premium.

An important question that arises here is "if income represents the ability

to pay of the people for HI premium. how did the low income people buy

MHI?" There are two possible answers; one is that the MHI schemes target

the low income people so that they are entitled to HI coverage by being

poor. and second is that the MHI schemes offer HI coverage at partially

subsidized premium rate. As against this. PHI schemes are not specially

targeted and designed for the poor people. and also do not provide any

subsidy as such. In short. in PHI schemes the household income is a push

factor but in the case of MHI schemes it may serve as both push and pull

factor for the high income and low income people. respectively.

Now let us move to another issue: If poor households are the target

population of the MHI schemes. one would expect that the schemes would

cover all low income households in the locations where they are operating.

However. as we have seen in the preceding pages. it is - coverage - not the

case: many low income households in the locations do not have MHI

coverage. What are the factors determining MHI coverage for some

households and exclusion for others? The next chapter discusses this issue

by analyzing the factors determining the MHI coverage.

18 PHI schemes are commercial schemes. therefore no substantial premium subsidy is granted to the clients.

70

CHAPTER FOUR

FACTORS DETERMINING MICRO HEALTH

INSURANCE COVERAGE

4.1. Introduction

In the previous chapter we have seen that In tenus of HI coverage, MHI

schemes are more equitable than the PHI schemes. However, we have also

observed that some of the MHI schemes have not only excluded some low

income households but also included high income households in their

coverage. This chapter examines the factors determining HI coverage in

MHI schemes. The significance of various factors such as Level of

Education. Household Size, SHG membership status and Health Risk. on

the probability to have HI coverage in each of the selected MHI schemes is

being examined.

4.2. Factors Determining Health Insurance Coverage

An important issue emerging out of the previous discussion is that if the

MHI schemes target the poor and thereby they are entitled to have HI

coverage, why all the poor in their operational area are not covered by MHI

schemes? Several possible reasons can be given to the question as:

1) The MHI units may be at the process of scaling up of HI and may

perhaps take time to cover the entire lower income households in their

operational area.

2) To attain a long term financial sustainability, MHI schemes would be

aiming at achieving both income solidarity and risk solidarity to cross

subsidize the health care expenditure of the poor from the rich people.

In such a situation, HI schemes must be a sound pool with all income

group~ under risk solidarity. We could not observe any of the MHI units

having an over representation of high income people in the risk pool.

71

Nonetheless, the 'Urban MHI' unit has relatively more low income

people than of high income people.

3 Another reason may be that low income people remain non insured

mainly because they do not have the motivation to buy HI even though

they can afford the premium. It is true especially in the case of high

income people who remain uninsured even though they have the

enabling factor in the form of income to afford a premium. There are

other enabling factors such as education and access to information on

HI schemes, access to HI through the membership in Self Help Groups

(SHGs) etc. Similar to the enabling factors, there may be some

motivating factors in the form of household size and health risk. In

summary, there may be several other factors affecting the health

insurance coverage rather than the income factor alone, we will examine

the significance of these factors by keeping in mind the following two

questions.

I) How does the Low Income Insured population of the MHI locations differ

from the Low Income Non Insured Population?

2) What are the significances of the other factors, which enabled the low

income population to become members of the risk pool as compared to

the Non Insured Low Income population and High Income Insured

Population?

We classitY the factors determining MHI coverage into push factors and pull

factors for the households to buy HI. The push factors can be divided in to

'Enabling Factors' and 'Motivating Factors' making the people to go for HI.

The 'Enabling Factors' include income, regularity in income receipt,

education and access to information on HI, and the 'Motivating Factors'

include health risk, age, income irregularities, household size etc. On the

other hand, the pull factors make the people eligible for enrollment into the

HI schemes. For example, being in the category of 'Destitute' makes a

household eligible for membership in a MHI scheme because many

72

schemes target mainly such groups19. Similarly. being a member of Self

Help Groups (SHGs) and Co-operative societies is another qualification to

be a member of the community MHI schemes because they (MHI schemes)

offer HI mainly through such institutions. Even though these schemes

target the vulnerable sections. they also want to include the relatively

better off and high income population from the coverage. as the

sustainability of any HI schemes requires a risk pool of high income to020 •

Figure 4.1.Determinants of Micro Health Insurance Coverage

Likelihood to have Health insurance

Push factors

+ Enabling factors

+ 1) Income. 2) Regularity in income receipt 3) Education

+ Motivating factors

+ 1) Health risk 2) Age. 3) Income irregularities. 4) Household size

Pull factors

1) Low Income 2) SHG membership

art'a! remium subsidy in the already subSidized premium. As 19 Some MHI schemes g: ~ se~tio~s of the society it can be expected that the vol~nteers of theshe sChhemes tar

lg) ebt we oree interested to give more counseling to make the relatively very

suc sc emes Wl em. weakest witj"l thin the weaker sections of the society.

. n funding from some external agencies. but these funding 20 Some of the MHI umts are getU ~tainlng finanCial sustainability is one of the long term can get stopped at any time. so a objectives of MHI units.

73

To distinguish the pull and push factor aspects of household income. the

different income classes and their membership status of Self Help

Groups/other community organizations (broadly termed as SHGs) are

analyzed. First. a comparison of the SHGs membership status within the

Insured across different income class is done. and secondly. comparisons

of the SHGs membership between the Insured and the Non Insured within

and across different income classes are performed. The role of education

and household size on the deCision to buy HI is examined after classitying

SHG memberships for their education and household sizes under various

income classes. Binary Probit models are used to estimate the significance

of both the push and pull factors among the households on the probability

to buy HI.

In the following section we examine the relevance of education of the

households. household size on the likelihood to have HI. and the role Self

Help Groups (SHGs). The health Iisk of the household as a motivating

factor to buy HI will be discussed in detail in one of the next chapters.

4.3. Educational Profile

The level of Education of the people and awareness on the importance of

health care has a positive relationship. An educated person will be more

likely to know how to and where from to get HI and will be more motivated

to go for HI. As the present study has used household as a unit of analysis.

we assume that the highest level of education attained by any family

member will have a positive externality on the decision making process of

that family. So we will use the highest level of education attained by any

household member in the family for the present analysiS. Here. the level of

education is measured in terms of the years of schooling. Similarly. the

highest education attained by the head of the household is also taken;

however. we will include this variable only in the econometric estimation.

74

Table 4.1 Highest educational qualification among the Insured and Non Insured households (%)

Rural MHI Urban MHI Private Voluntary schemes schemes Health Insurance

Years of Non Non Non schoolinl! Insured Insured Insured Insured Insured Insured Illiterate 11.25 15.33 1.88 3.05 0.00 0.00 1st-4th 0.00 0.00 vear 6.52 S.IS 2.S9 4.94 5th-8th vear 23.16 24.93 25.25 20.49 1.00 10.00 9th - 10th vear 27.22 2S.13 35.93 37.50 1S.00 37.00 II-12th vear 17.01 14.77 17.03 16.86 23.00 14.00 12th -l51h vear 11.44 6.40 12.55 11.19 30.00 26.00 15th and above 3.40 2.26 4.47 5.96 2S.00 13.00 Total 100 100 100 100 100 100

(1058) ( 1064) (697) (704) (200) (200)

Pearson Chi-Square 29.748 (P<O.OOO) IS.511[p<0.010) 32.324(P<0.000}

Sources: Field survey and ECCP Data Values In the parentheses refer to the number of households

It is noted that even though the Insured households hold relatively higher

levels of education as compared to their Non Insured counter parts in each

scheme, the differences are very low, based on such observations we can

not make a conclusion that educational qualification of the households is

really a highly Significant 'enabling factor' for the households to go for HI.

Therefore, we will consider the educational qualification of different income

class and analyze in terms of each type of MHI schemes. For this purpose,

the educational years are again merged in to 3 categories: I} Low level of

education: TIll 4 th years of schooling. 2) Medium level of education: 5 th

years to IOu, years of schooling, and 3) High level of education: 11 u, years

and above schooling.

75

Figure 4.2 Proportion of the educational qualification among the Insured and the Non Insured in the 'Rural MHI' schemes

60

51 50 47 48

C. 40 !II '0 '0 30 ~

:x '" 0 20 :z:

10

o "-,J.J.J..L_

Non-Insured Insured Non 'ns did N ' -I ure nsure on-Insured Insured

Low Income Middle Income High Income

Source: ECCP data

ID Low le..el Educatio

[J Medium le..el Educatio

Figure 4.3 Proportion of educational qualification among the Insured and the Non Insured in the 'Urban MHI' schemes

70

60 55

'i 50 -!II '0 40 '0 ~

:x 30 :::I 0 20 14 :z:

10

0 Non-insured Insured Non-insured Insured Non-insured Insured

Low Income Middle Income High Income

Source: ECCP data

CD Low le..el Education

o Medium le..el Education

• High le..el Education

The major finding from the above figures is that the Insured households

hold relatively higher levels of educational qualifications as compared to

the Non Insured households irrespective of which Income class they

belong to. For example. if we consider the 'Low level of education'. the

proportion of the Non Insured households falling under this category is

very high. This is true for both the 'Rural MHI' and 'Urban MHI' schemes.

76

4.4. Household Size

The recent trend of the transformation of Joint family system to nuclear

family has increased the finanCial insecurity among families in case of

uncertainties (Abel-Smith. 1992). The joint family system is a kind of

informal insurance system for its members. In India a large number of

families are converging into a nuclear family system. In this context. there

is more scope for the increase in demand for HI. But in a situation where

insurance awareness is very low among the Indian SOCiety. the validity of

this kind of theoretical prediction is questionable. To make an empirical

estimation. we do not have data on both Joint family and nuclear family,

and it is beyond the scope of the present study. but based on HH sizes, the

existing data can be sorted into Macro and Micro families. From an angle of

health care expenditure. higher household size means higher health care

expenditure and hence high demand for HI. Considering the Micro and

Macro families. higher household size means higher informal insurance

among the family members and hence low demand for HI. In this context. it

Is very difficult to predict the impact of household size on the level of HI

coverage. The empirical evidence presented in the table below also gives a

mixed picture on the impact of household Size on the HI coverage; that is,

there is no impact of Household size In the case of 'Rural MHI' and a

negative Impact in the case of 'Urban MHI' schemes on HI coverage.

Table 4.2 Mean value of the household size

MHI schemes Non Insured Insured 't'value

Rural MHI 4.29 4.37 -1.075(P<0.283)

Urban MHI 4.81 4.60 2.391(P<0.017)

Source: ECCP data

In the subsequent sections we consider the household size of the various

income classes to examine the correlation between the household size and

the HI coverage. If we take the case of low income people. there are two

aspects to be conSidered. one is. a low income household with lower and

higher family size will be equally motivated to buy health insurance due to

77

the access value of health insurance. Similar is true for high income

household also because higher family size will increase their degree of

financial insecurity due to the uncertain illnesses. As we do not have any

strong theoretical support to argue on the impact of household size on the

HI coverage, let us see what the data speak. The household size Is coded as

a binary variable as household size with a maximum of 4 is coded as one

category and household size 5 and above is coded as another category. The

two figures below show the household size of both the Insured and Non

Insured across different income classes in case of both the 'Rural MHI' and

'Urban MHI' schemes.

Figure 4.4. Household size across different income classes in the case of Rural MHI schemes

------

80

70 69 66

61 60 - 60 53 54 ~

- 46 • 47 '050 '0

~ '0 40 ~ .,

CI Household

~ 39 40 34 :-:

size below

~:-: 31 5 til 30 " 0 :: 20 I

10 I·-~ Ii! Household

- ~ ---- size 5 and 0 above

Non-Insured Insure d Non-insured Insured Non-Insured Insured

Low Income Middle Income High Income

78

Figure 4.5. Household size across different income classes in the r-_____ c_a_se of 'Urban MHI' schemes

80 i 'u

70 61 62 65 ~

- 60 ;l!. D -.. 50 'C "0 40 ~

ll! 30 :::J 0 J: 20

10

0

...,....,. 54 ~. ~ o Househol

~'46 51 49 d size

39 ..... .,-,-below 5 I' 38

~ ..

35 I' ••••• 1', ..

:;: r-:--: 27 • IliI Househol

I ::: .. d size 5

8 .. " S ~ S .... S S ..

~ ! abo..e . ,

Non-insured Insured Non-insured Insured Non-insured Insured

Low Income Middle Income High Income

We observe no correlation between the household size of the Insured and

the Non Insured households within and between income classes across

both types of MHI schemes. Whatever relationship we have observed here is

just incidental, not at all consistent between both types of MHI schemes_

One possible reason for this is the insurance awareness. especially on HI.

Is very low In India, so the decision to go for HI will be determined by many

other structural factors within the system and society. and thus. will be

least determined by the factors like household size.

4.5. Health Insurance Coverage and the role of Self Help Groups

The MHI units offer HI to clients through Self Help Groups directly. The

SHGs may serve two purposes for the MHI units; first. to identify the low

Income people to sell the HI schemes. and second. to act as information

dissemination channels for the HI schemes21 • It is expected that SHGs have

more role In information dissemination on MHI schemes. and, hence. the

people who are members of these groups. are more probable to get

information about MHI schemes than those of who are not members of the

same.

11 The role of SHGs as an information dissemination channel is analyzed in the next chapter.

79

Figure. 4.6 Micro Health Insurance Model

MHI units

Premium

Insurance

Premium

Reimbursement

Insurerl Insurer plus health care provider

SHGs

Product sale& Product servicing

Premium

Product sale& Product servicin~

Reimbursement

Premium

Health care provider

Clients

Hence. being a member of SHGs in the operational area of MHI units work

as a pull factor for having HI coverage for the households. This inference

has more relevance for the poorest households. Thus. a simple question

that arises here Is whether SHGs membership of the household members

has any Impact on distinguishing the low income insured from the low

income Non Insured. the high income Insured and the high income Non

Insured? It Is in this context. the following hypothesis is tested: "Being a

member of SHGs does not increase the probability to have HI for the low

income class population as compared to the high income class population".

First of all. we will examine the membership status of the Insured and the

Non Insured households across different income class in both the 'Rural'

and 'Urban' MHI separately. Figures 7.a and 7.b present the membership

status of the Insured and the Non Insured of Rural MHI schemes.

80

Figure 4.7-a: The SHGs membership status of the Insured households of the 'Rural MH!' schemes

80~--------~~ ________________________ --,

70 67 68

- 60 "­ur 50 "C

] 40

~ 30

~ 20

10

o "--''--Destllute Extreme Moderate Vulnerable Non-Poor Wealthy

Poor Poor Non-Poor

Income Class

Pearson Chi-Square: 8.52 (p< 0.129)

Source: ECCP data

OSHG Non­Member

.SHG Member

Figure 4.7-b: The SHGs membership status of the Non Insured households of the 'Rural MHI' schemes

90 :r-~~--~~~--7~9=--=-~~7;8=======7~9~-==========-======~~------~ 80 70

~ 60 "C -0 50

! -5; 40 fIl

'" ~ 30

20

10

o ~-Destitute Extreme

Poor

69 64

Moderate Vulnerable Non-Poor Wealthy

Poor Non·Poor

Income Class

Pearson Chi-Square: 23.89 (p<O.OOO)

Source: ECCP data

D SHG Non­Member

.SHG Member

81

Figure 4.7 -c: Proportion of Insured and Non members in Rural MHI Schemes

Insured among the SHG

90r--'~--------------______________ ~ 80

~ 70

~ 60 :E 50 o J:. ~ 40

6 30 :z::

20 10

o +'-'------

Destitute Extreme Poor

Moderate Vulnerable Non-Poor Poor Non-Poor

Income Class

Pearson Chi-Square: 9.64(p<O.086)

Source: ECCP data

Wealthy

CJ Non Insured

I. Insured

Let us consider the case of 'Rural MHI', Looking at the lowest two income

classes consisting of the 'Destitute' and 'Extreme Poor', it can be seen that

among the Insured households that a majoIity of them are SHG members

(67% and 68%, respectively). But when it comes to the Non Insured

households of the same income strata, only 16% of the 'Destitute' and 21%

of the 'Extreme Poor' are SHG members, Similar pattern is observed in the

case of the 'Middle Income' class consisting of 'Moderate Poor' and

'Vulnerable Non-Poor' where 55% and 61% of the Insured households and

only 22% and 21 % of the Non Insured households are SHG members. In

the case of the 'High Income' class, 58% of the 'Non Poor' and 61 % of the

'Wealthy' among the Insured households are SHG members while 31% of

the 'Non Poor' and 36% of the 'Wealthy' among the Non Insured households

are SHG members.

The major observations from the above discussion can be summarized as

follows: 1) The proportion of the households With SHG membership is

higher among the Insured Households as compared to the Non Insured

households; 2) Among the Insured households the proportion of

households with SHG membership gradually declines when households

82

move from the lowest income class to the highest income class whereas.

among the Non Insured households the proportion of SHG membership

gradually increases. The main inferences that stem from these observations

are that the SHGs membership increases the likelihood to have HI. and the

SHGs membership helps the relatively low income class to obtain HI. In

other words. the income as a pull factor works mainly when the low income

class has SHGs membership.

A similar analysis holds for the 'Urban MHI' schemes too. but with a little

difference in the findings. As compared the 'Rural MHI' schemes. the

proportion of SHGs membership among the Insured and Non Insured

households are relatively low. may be because the presence of SHGs in

urban slums might be very low. On the whole. the importance of SHGs

membership in having HI for the households and working as a pull factor

for the low income households are low as compared to the 'Rural MHI'

schemes. But. as compared to the Non Insured households. the

importance of SHGs membership on having HI is very high for the Insured

households.

Figure 4.B-a: The SHGs membership status of the Insured households of the 'Urban MHI' schemes

90~----------------------------------------

80

70 ~ ~ 60 "C '0 ~

3! " ~

50 40

30

20

10

o

T7 74 72 70 68 68

32 32

Destitute Extreme Poor

Moderate Vulnerable Non-Poor Wealthy Poor Non·Poor

Income Class

Pearson Chi-Square: 2.453 (p<O.783)

Source: ECCP data

oSHG Non­Member

.SHG Member

83

Figure 4.8-b:

120

100 88 ->!! • 80 -III 'tI '0 60 'i ::I 40 0

:l:

20 12

0 Destitute

The SHGs membership status of the households of the 'Urban MHI' h sc emes

93

Extreme Poor

91 97 94

3 6

Moderate Vulnerable Non-Poor Poor Non-Poor

Income Class

92

8

Wealthy

Pearson Chi-Square: 7.233 (p<O.203)

Source: ECCP data

Non Insured

lJ SHG Non-I Member ,

Member

Figure 4.8-c: Proportion of Insured and Non Insured among the SHG members in 'Urban MHI' Schemes

1'00 90 80

;f- 70 'iii

60 'C '0 50 J: OIl

'" 40 ::I 0 30 J:

20 10 0

Destitute Extreme Moderate Vulnerable Non-Poor Wealthy

Poor

Pearson ChI-Square: 7.05 (p<0.216) Source: ECCP data

Poor Non-Poor

Income Class

I:l Non Insured

• Insured

From the above figures, it is obvious that SHG membership status and MHI

coverage has a positive relationship. In other words. SHG membership

enables the households to have HI coverage. The significance of SHG

84

membership to have HI coverage is calculated With a Probit model below.

Results for each unit of MHI schemes are also presented.

4.6. Econometric Estimation

A maximum likelihood estimate of the binary Probit model is used to

measure the Impact of SHGs membership on the likelihood to have HI.

The six income classes are further reduced in to three income classes to

make the model simple and easy to interpret. The three income classes

with SHG membership Is coded in to six dummy variables. Other

variables included in the model are the household size. the highest

education by any family member in the household. education attained by

the head of the household. health Iisk and the daily wage. The daily wage

is a proxy to measure the irregularity in income as people who are ean1ing

dally wage will be having more uncertainty and irregularities in income

receipt.

Table 4.3 Definition and measurement of variables

Variables Description Health Insurance I if the household has health insurance; 0

OtheIWise Hhsize Household size Lower_edu Highest Educational qualification of the

household is till 4th years of schooling =1; OtherWise-O

Medium_edu Highest Educational qualification of the household is 5 th to 10th years of schooling = I; OtherWise-O

Higher_edu (Reference Highest Educational qualification of the category) household is I I UI and above years of schooling

1; OtherWise-O Headeducation Highest education attained by the head o~ the

famil~ measured in terms of years of schoolm.e; HealthIisk Health status of at least one member of the

households is very poor= I; 0 otherWise Dailywage Main source of the household income is daily

wage= 1; OtherWise=O

85

Variables Description

Lowinc_sh~memb Low income household with SHG membership = 1: Otherwise-O

Lowinc_sh~nonmemb Low income household with SHG non-membership -1; Otherwlse-O

Midlnc_sh~memb Middle income household with SHG membership -1; Otherwise-O

Midinc_sh~nonmemb Middle income household with SHG non-membership = 1; Otherwise-O

Hlglnc _ sh~memb High Income household with SHG membership = I; Otherwlse=O ,

Higinc_ High income household with SHG non-sh~ nonmemb membership = 1; Otherwise=O

In the table below. the Pro bit model results of three 'Rural MHI' schemes

are presented In together and by unit wise.

Table 4.4 Probability to have health insurance coverage - Probit model results of Rural MID schemes

Rural MHI Rural MHI Karuna Yeshaswini Dhan (Karuna Trust Trust trust, Yeshaswini trust, and DHAN)

Variables Marginal Marginal Marginal Marginal Effect Effect Effect Effect

Hhsize -.00012 -.01770 .01322 .03333 (-0.02) (- 1.43) ( 1.16) (0.98)

Lower_edu -.18094 -.15195 -.24542 -.03383 (-5.17)- (-2.76)- (-3.66)- (-0.27)

Medium_edu -.11364 -.11072 -.13891 .00271 (-4.05)- (-2.16)-- (-3.12)' (0.03)

HeadeducaUon -.00509 -.01442 .00581 -.01868 ( -1.83)**' (-2.95)- 0.26) (-

1.85)-"

Healthrisk .10984 .06482 .13603 .25579 (3.11)' ( 1.04) (2.29)-' (2.25)**

Dailywage .01945 -.12843 .09875 -.08876 (0.50) (-1.19) (1.72)"- (-0.79)

Lowinc_shg...memb .33321 .15844 -.01362 .65893 (8.16)- (2.21)'- (-0.15) (9.87)-

Lowlnc_sh~nonmemb -.16621 .01205 -.38885 -.03133 (-4.23)" (0.17) (-6.56)" (-0.24)

86

RUl'al MHI RUl'al MHI Karuna Yeshaswini Dhan (Karuna Trust Trust trust, Yeshaswini trust, and DHAN)

Midinc shg...memb .27534 -.03316 -.03174 .71880 (7.25)' (-0.47) (-0.39) (11.33)·

Midinc_shg...nonmemb -.11136 -.04997 -.19949 -.11395 (-3.22)* (-0.72) (-3.96)* (-0.86)

Higinc _ shg...memb .25797 -.11340 .03844 .79756 (7.34)' (- (0.59) (11.61)*

1.69)'*' Y Pr(insured) .49835 .50233 .48906 .49531 (predict) Lo.e: likelihood -1288.62 -467.72 -431.48 -104.51 LR chi2 (11) 354.72 34.94 101.74 789.09 Pseudo R2 0.1210 0.0360 0.1055 0.7906 Number of 2115 700 696 720 observations

Reference categories: 1) Higher income with no SHG membership. 2) Higher levels ofeducaUon Values in the parentheses refer to the "Z" statistics; Level of statistical significance: • = 1%; •• =5%; ••• = 10%

By considering the model of all the three 'Rural MHI' schemes together,

household size does not seem to have any impact on the likelihood to have

HI.

Higher the level of education higher will be the probability to have HI. As

compared to the households with 'High level of education' (reference

category). households with 'Lower level of education' and 'Medium level of

education' are 18% and 11%. respectively. less likely to go for HI. The

highest education attained by any of the family member has a positive

impact on the likelihood to have HI. and it is true especially in the case of

the 'Kamna Trust' and the 'Yeshaswini trust'. When we consider the

education of the head of the household we found a negative impact on the

probability to have HI. which is against the theoretical expectation. One

possible reason for this may be that these MHI work in the rural areas and

the education level is very low (perhaps below some threshold level) among

the old generation whereas the household head normally belongs to an old

generation consisting of the middle aged.

87

As compared to the 'High income households without SHG membership'

(reference category). the 'Low income households with SHG membership'

are 33% more likely to have HI and the 'Low income households without

SHG membership' are 17% less likely to have HI, which means that SHG

membership is a pull factor for any household to have HI. Similarly.

compared to the 'High income households without SHG membership'

(reference category). being a 'Middle income household with SHG

membership' Increases the likelihood to have HI by 28%. and being a

'Middle income household without SHG membership' reduces the likelihood

to have HI by 11 %, The 'High income households with SHG membership'

are 26% more likely to have HI as compared to the same income group

without SHG membership, In summary, it can be seen that SHG

membership is a positively significant factor to have HI for all income

classes,

Among the 'Rural MHI' schemes, the Kanma Trust's health insurance

coverage Is not significantly determined by the SHG membership except for

the 'Low Income Class',

In the case of the 'Urban MHI' schemes. household size reduces the

probability to have HI by 4%, As compared to the households with 'High

level of education' (reference category), households with 'Lower level of

education' are 23% less likely to go for HI. but 'Middle level of education'

does not have any Impact on the likelihood to have HI as compared to the

'Higher level of education',

As compared to the 'High Income households without SHG membership'

(reference category), the 'Low income households with SHG membership'

are 29% more to have HI and the 'Low income households without SHG

membership' do not have a significant impact on the likelihood to have HI

as compared to the reference category. Similarly, compared to the 'High

income households without SHG membership' (reference category). being

a 'Middle income household with SHG membership' increases the

likelihood to have HI by 38%, and being a 'Middle income household

88

Table 4.5. Probability to have health insurance coverage - Probit model results of Urban MID schemes

Dependent variable: 1 if the household has health insurance; o Otherwise

Urban MID Urban UPLIFT VHS MHI

(UPLIFT.

VHS)

Variables Marginal Marginal Marginal

Effect Effect Effect

Hhsize -.03743 -.05062 -.01288 (-4.10)* (-3.43)* (-1.05)

Lower_edu -.23233 -.29388 -.09514 (-3.88)* (-3.24)* (-0.97)

Medium_edu -.02982 -.00523 -.04937 (-0.93) (-0.11) ( 1.11)

Headeducation -.00420 .00696 -.01327 (-1.02) (1. 13) ( 2.26)**

Healthrtsk .18430 .16559 .20640 (5.42)" (2.82)" (4.65)*

Dailywage .04889 .02981 .06385 ( 1.57) (0.54) 0.56)

Lowinc_shg",memb .29252 .42395 .12820 (5.12)" (5.21)* ( 1.54)

Lowinc_sh/Lnonmemb .05025 -.09131 .04108 ( 1.30) (-1.38) (0.75)

Mldlnc_sh/Lmemb .42067 .49826 -.17374 (7.22)* (7.08)* ( -1.04)

Mldinc_shg",nonmemb .05637 .07088 .01921 (1.43) ( 1.31) (0.32)

Hlglnc _ sh/Lmemb .37404 .44200 -.23717 (6.04)" (6.48)* ( -1.02)

Y = Pr(insured) .50327 .51270 .50320 (predict) Log I1kel1hood -886.58 -384.70 -459.69

LR chi2 (11) 162.07 189.85 41.31

Pseudo R2 0.0837 0.1979 0.0430

Number of 1396 692 693

observations Reference categories: 1) Higher income WIth no SHG membershIp. 2) Higher levels of education. Values in the parentheses refer to the 'Z' statistics; Level of statistical significance: * = 1%; ** =5%; *** = 10%

without SHG membership' does not make any difference from the

reference category. The 'High income households with SHG membership'

are 37% more I1kely to have HI as compared to the same income group

89

without SHe membership. The main inference that stems out of the above

discussion is that in the case of 'Urban MHI' schemes, SHe membership

is a pull factor not only for the 'Low Income class', but also for the 'High

Income Class'. Among the 'Urban MHI' schemes, the HI membership in

the 'VHS' is not significantly determined by the SHe membership status

of the households.

Let us also discuss the significance of SHe membership for each income

class as compared to same income class without SHe membership. The

table below presents selected parameters after re-estimating (3 times) the

above model by simply changing the reference categories. As mentioned

before. the re-estimatlon of the model would not affect the properties of

the model.

Table 4,6. Probability to have health insurance coverage: Marginal effects of the probit model of the selected variables

Variable and Rural Karuna Yesha· OHAN Urban UPLIFT VHS Reference MHI swini MHI cate.!l~

LO\lI1nc_sh~ .442 .141 .371 .658 .254 .460 .100

memb 116.69)' 12.11)" 15.69)' 117.59)' 15.15)' 111.26)' (1.37)

(Reference category: Lowinc_sh~

nonmembl Midinc_sh~ .364 .011 .165 .726 .387 .468 -.175

memb: 112.57)' 10.18) 12.16)" 111.71)' (9.01)' 111.50)' 1-1.14)

(Reference category: Midinc_sh~

nonmembl -.252 .258 -.112 .039 .795 .370 .441 Higinc_

17.95)' 1-1.68)'" 10.611 115.05)' 18.43)' (10.36)' 1-1.27) sh~memb (Reference category: Higinc_ sh~nonmem

bl

The above table is presented in the form of a histogram for a better

interpretation of the results of the model.

90

Figure 4.9: Probability to have health insurance

income with SH(; membership coverage for each

s: <II > co .r. ~ ,.. .'.::

.D co .D 0 ~

~

P~obability to have HI coverage for the households of each Income class with SHG membership as compared to the

households wi~hout. SHG memership of the same income class, by various Income class across MHllocations

100°0 .. . '

80°0

60°,

40°0

20°'0

O~O

·20°0

-40°0

Yeshaswini DHAN

M HI Locations

UPLIFT

• Low Income

• Middle Income

o High income

It can be seen that as compared to the households without SHG

membership of the same income class except the VHS. the households

\'lith SHG membership has significant influence on having HI for low

income class. As compared to low income households without SHG

membership, low income households with SHG membership in the

locations of Karuna Tnlst, Yeshaswtni Tnlst. DHAN and UPLIFT have

14%. 37%. 66% and 46% more probability to have HI coverage,

respectively. Similarly. the SHG membership increases the probability to

have HI for the high income households, especially in the MHI locations of

Yeshaswini. DHAN and UPLIFT. Interestingly. in the locations of Karuna

and VHS the high Income class with SHG membership has less

probability to have HI as compared to households without SHG

membership. However. there is no statistical significance on such

relationship; therefore. any further explanation of this result is

meaningless.

91

4.7. Chapter Summary

We have found that there is a crucial role for Education and 5HGs in

giving access to health insurance for the low income households along

with their counter parts. Nonetheless, these factors have significant

impacts on increasing the size of the risk pool in the form of scaling up of

MHI that will enhance the equity objectives of health insurance coverage

in the form of giving access to large number of low-income people. It is an

indication that the grassroots level organizations enable the outreach HI

coverage among the poor communities. One implication of this finding is

that in order to enhance the scale up process of HI it is important to

ensure the role of these of organizations. The MHI units and other stake

holders also can make use of the Panchayati Raj Institutions (PRIs) that

are wide spread in India along with the SHGs.

So far in this chapter we confined our discussion to factors determining

an equity based scale up process of HI. We are equally interested in

knowing the factors adversely affecting equity based scale up process of

PHI schemes, the next chapter addresses these issues.

92

CHAPTER FIVE

INFORMATION ASYMMETRY, MARKET FAILURE

AND THE HEALTH INSURANCE COVERAGE

5.1. Introduction

In the previous two chapters we carried out a comparative analysis of the

equity aspects of the HI coverage of both the MHI and PHI schemes and

found that MHI schemes are more equitable than the PHI schemes.

Moreover, we have examined some of the determinants of equity in HI

coverage with special reference to MHI schemes. Now. let us examine the

determinants leading to inequity in PHI coverage and its relation to the

scaling up process of PHI. The main focus of this chapter is to analyse the

determinants of the scaling up process of HI leading to inequity in HI

coverage with special reference to PHI. The factors contributing to the

asymmetric infonnatlon on HI poliCies ARE analysed by comparing different

aspects of the awareness on different types of insurance in general and HI

in particular between the Insured and the Non Insured of the PHI schemes.

Further, the familiarity of the people with different fonns of insurance is

measured by comparing the 'Other Insurance enrolment statuses' of both

the PHI Insured and the Non Insured. We analyze the role of infonnation

asymmetry on HI schemes between the Client and Insurer on 'selection

bias'. Insurance Agents are the main channels for infonnation

dissemination on HI schemes. In PHI schemes, the role of 'Insurance

Agents' in the scaling up of HI and emerging 'Selection Bias' are examined

by analyzing utility maximizing behavior of the 'Insurance Agents'. Both

bivariate analyses and multivariate analyses including a Probit regression

model are applied to estimate the significance of various factors affecting

the decision to buy HI.

93

5.2. Conceptual and Theoretical Frame

We have already noted that the HI coverage is very low in India; the

disequilibrium in risk solidarity in the form of 'Selection Bias' might be

cited as one reason for such a market failure. There are two types of

solidarity in a HI system such as Income solidarity and Risk solidarity.

Income solidarity is defined as cross subsidization of the health care costs

of the poor by the rich in a risk pool. whereas risk solidarity is defined as

cross subsidization of the health care costs of high risk (health risk) by the

low risk people. Disequilibria in risk solidarity will lead to 'Selection Bias' in

the form of either more of high risk or low risk people in risk pool where an

ideal risk pool must be characterized by the absence of selection bias. A

risk pool with selection bias may be featured by the presence of either

adverse selection or cream selection (skimming) due to asymmetric

information. Asymmetric information arises when one agent has relatively

better information than the other agent about some parameters that are

relevant for the relationship (Akerlof, 1970).

Adverse selection is perceived to be a major source of market failure in

insurance markets. The problem of adverse selection is present in all lines

of insurance due to hidden information. People who insure themselves are

those who are increasingly certain that they will need the insurance

(Akerlof. 1970). Adverse Selection in HI market is defmed as a situation of

over representation of high-risk people in the risk pool. Individuals

themselves know much about their health condition than the insurance

companies (Rothschild and Stiglitz. 1976). Customers who know

themselves to be at high risks are motivated to buy more insurance and are

likely to use it (woR. 1993). In a population of individuals whose

underlying health risks are heterogeneous. more and less healthy people

will demand more HI. In short. adverse selection arises because individuals

have more information about expected medical expenditures than

insurance providers. In most theoretical models. the asymmetry is relative

to the risk: the client is assumed to know it better about her accident

probability and the (conditional) distribution of losses likely to be incurred

in case of accident (Akerlof. 1970; Miyazaki. 1977; Wilson. 1977). The

94

ability of prospective insurance customers to conceal their true risks can

result in some insurance groups having a disproportionate number of high

users. This situation will lead to higher than average premiums for the

group and create an incentive for low risk individuals to drop out of the

group in search of lower cost coverage elsewhere. Adverse selection

presents a serious problem for risks existing at the time insurance is taken

out: but an even more complex problem arises from the fact that an

Initially low risk person may become high risky later in life (WDR. 1993). A

straightforward way of preventing an extreme form of adverse selection is to

mandate that everyone buys the specified HI coverage. But. such a

compulsion is not possible in a voluntary HI market. Therefore. a profit

motivated Insurer may adopt the strategy of cream selection. another form

of selection bias. The cream selection will result in the exclUSion of the

high-risk people from the risk pool. Cream selection can occur when

insurers are able to identifY subgroups of the population with different

expected medical costs. Insurers may have an incentive to seek out the low

risk population subgroups and sell insurance to them at reduced

premiums as compared to high-risk sub groups. Cream skimming is the

result of regulation in the insurance industry. not competition (Pauly.

1984). Without an efficient mechanism of risk-adjusted premium

differentials. the likelihood of cream skimming exists. In the process of

cream selection strategy by Insurers. there is high probability on the

exclusion of the population group of high risk consisting of aged. poor.

women etc .. from the HI coverage.

The theoretical works of Akerlof (1970), Rothschild and Stiglitz (1976),

Miyazaki (1977). and Wilson (1977) describe separating eqUilibrium. where

high risk consumers purchase policy with higher coverage than the policy

that is purchased by low risk consumers. Rothschild and Stiglitz's (1976)

examined the market eqUilibrium with and without full information on

consumer health risks by Insurers. In their model, when Insurers have full

information about consumer risk characteristics. all risk-averse consumers

are offered to purchase full insurance at actuarially fair prices. This

outcome is not possible when consumers have private information about

their risk status. Instead. insurers may offer different poliCies that induce

consumers to reveal their true risk type by the poliCies they select. In the

95

resulting "separating" equilibrium. market will be segmented with low risk

consumers purchasing a lesser quantity of insurance than high-risk

consumers. Within this framework. imposing restrictions on insurer

underwriting practices is analogous to moving from a situation of full

information to one in which insurers have an information deficit. To the

extent that these new constraints bind. the forced pooling of lower and

higher risks will lead to higher premiums for low risks than would occur in

an unregulated market.

Miyazaki (1977) extends the separating model to allow cross-policy

subsidization. resulting in a wealth transfer from low risks to high risks. In

addition to a separating eqUilibrium. Wilson (1977) describes a pooling

model where high and low risks purchase same policy so that low risks

actually subsidize the insurance purchases of high risks.

In spite of the extensive theoretical interest. there is little empirical

evidence on the extent of the problem. There is as yet little direct evidence

on whether or not adverse selection is or must be an important problem in

HI market. largely because it is difficult to define any strong test (Pauly.

1986). However. the studies by Juba. Lave. and Shaddy (1980). Ellis

(1985), Wrightson. Genuardi. and Stephens (1987), Marquis and Phelps

(1987), Cardon and Hendel (1996). MarquiS (1992), Browne and

Doerpinhaus (1993) and Van de Ven and Van VJenit (1995) found evidences

on the presence of adverse selection in HI market. Cameron et.al. (1988)

find adverse selection in an Australian sample. Phelps (1976) found no

systematic relation between predicted illness of individuals and insurance

choice. Dowd et al. (1991) also find no evidence of evidence of adverse

selection in their Minnesota sample. Farley and Monheit (1985) revealed a

kind of 'ambiguous' selection bias in the choice of HI.

One of the important features of these studies is that unknOwingly or

knowingly they assume people are aware of the HI schemes available in the

market. Moreover. these studies presume the existence of many types of HI

poliCies with different levels of co-insurance and benefit levels. In this

context. we consider the HI coverage in an imperfect market. characterized

by absence of perfect Information. with only a few products that are not

96

perfect substitutes. Further. the present study takes into account the

impacts of the aWareness and information dissemination about the

importance and features of HI policies among the people. and also the role

of insurance agents on selection bias. Most often. the literatures on HI

speak about asymmetry of information between Insurer and Client

(insured) about the health status of the latter. In addition to such an

information asymmetry problem of Insurer on the health risk of the clients.

the present study takes into account the Client's information problems in

the form of asymmetric information between Client and Insurer about the

HI schemes.

In Indian voluntary health insurance market. even though both private and

public sectors insurance companies are offering HI products. the Mediclaim

Insurance Policy of the four public sector companies (NIA. NIC. OIC. and

UIC) corner a large share (84%) of the total HI coverage. and the private

sector companies' products are somehow similar to the Mediclaim POlicy. In

short. there is no distinction between more generous plan and less

generous plan22 as such. Thus. it can be assumed that there is only one

health insurance policy in the market which is addressed to the average

risk people. Under this situation. people can deCide either to join the policy

or not to join and also how much. but they do not have the option of

selecting plans as according to their own health status and morbidity

conditions. Since all agents face the same (unit) price. high-risk

individuals are de facto subsidized. whereas low risk agents are taxed. The

latter are likely to buy less insurance. or even leave the market. A first

prediction of the theory is precisely that. in the presence adverse selection.

the market typically shrinks. and the high-risk agents are over-represented

among buyers. In addition. purchased quantities should be positively

correlated with risk I.e .. high risk agents should. everything being equal.

buy more (amount) insurance. Accordingly. the following hypotheses are

being questioned and tested in this chapter. "There is no Adverse Selection

in the Indian Voluntary Health Insurance Market".

Itl f the Insurance plans are determined on the basis of the types of 22 The generos es 0 1 diseases and amount of health care expenditure covered by the pans.

97

5.3. Familiarity with Various Aspects of Insurance

Rational or purposeful choice among consumers Is possible depending

upon their disposable income and full knowledge about their own

preferences. When consumers have trouble In gathering and understanding

information on preferences, the ability to make Informed decision Is

compromised (Rice. 1998). A key element for the smooth functioning of HI

markets is the premise that all consumers have access to the same

information as do providers and purchasers of health care Insurance, and

understand them. If this condition Is satisfied, Individuals will be able to

judge for themselves the value of the products offered In the HI market.

Some examples are: a HI package, its price, its quality and related

customer service. Thus, a well-informed consumer can make wise deCisions

in the market - whether to buy or not, what to buy, from whom to buy and

how much to buy etc. Accordingly, we test the hypothesis that "higher the

familiarity of people with different forms of Insurance, higher will be the

probability to join the HI". We test the hypothesis by examining the level of

awareness about the Insurance schemes and the 'Insurance habit' of

people.

5.3.1. Awareness about the Insurance System

Here, an attempt Is made to test the level of awareness about any

insurance in general and HI in particular among the Insured of the PHI

schemes and Non Insured of the same geographical locations. Accordingly,

we Investigate whether the Insured are well-informed consumers or not.

The structure of the Indian insurance market can be classified Into the

General Insurance and the non-general Insurance offered by both public

sector and private sector companies. Examples of general Insurance

products are fire insurance, marine insurance, motor Insurance and HI;

and the non-general insurance comprise mainly Ufe Insurance poliCies. It

can be seen that the Insurance policies of the non-general insurance

companies in India are viewed with having both saving and risk

components, whereas the general insurance companies are having only the

rI k t The HI products are featured by a pure risk component s componen.

98

~riSk pooling HI) alone. Therefore, the level of awareness of both types of

Insurance products will h . fl ave m uence on the people about risk pooling HI products. And also thos h . , e w 0 are haVIng awareness about general

insurance are more likely to know about HI as compared to those who are

aware of only non-general insurance.

Table 5.1. Knowledge about Insurance providers by the Insured and Non Insured (%)

fLnsured Irion Insured (N-200) (N-200)

Heard about general 100 29.5 nsurance companies (any of) Heard about non-general 100 99 insurance companies (any of)

Heard about general public sector 100 29.5 insurance companies (any of)

Heard about general private sector 13 2.5 Insurance companies (any of) Heard about non-general public 100 99 sector Insurance companies (any of) lHeard about non-general privat~ 12 2 sector Insurance companies (anv of)

Source: Primary Survey N= Total number of observations

From the above table. It can be seen that when we classi!y the insurance

market into General and Non-general insurance, about 29.5% of the Non

Insured people have heard about general insurance while all Insured

people have heard about general insurance. When it comes to non-general

insurance. no significant difference exists between Non Insured and

Insured. Both categories of people are familiar with non-general insurance

(Le., life insurance). When we further classiJY the people about their

knowledge on the general and non-general insurance schemes on the basis

of the ownership as public and private, there is a trend that people in both

categories heard about public sector insurance rather than private sector.

One probable explanation for this is that till the year 2000 the HI market

comprising both general and non-general in India was under a monopoly of

the public sector. Another explanation is that besides the fact that the

99

private sector insurance business is a new entry, they are also not spread

allover India and are just concentrating in the urban areas. In summary,

we can see that although the Insured people know about the General

Insurance companies, a majority of them have not heard about the private

sector general insurance companies which suggest that they are not fully aware of the insurance market.

Now we extend our discussion on awareness of the people on different

fom1s of Insurance products. Table 5.2 lists out some examples from both

the general and non-general Insurance products. The examples for the

general insurance products are Motor (vehicle) insurance, shop insurance,

fire and house Insurances and examples for the non-general HI are life

insurance policies and pension schemes. It can be seen that 87% of the

Insured have heard about motor insurance while 78% of the Non Insured

people also have heard about it. of shop insurance it is 51 % and 34%, and

of fire or house Insurance it is 32% and 14%, respectively, for Insured and

Non Insured. IntereStingly, there is not much significant difference between

insured and uninsured on life insurance (non-general policies), that is, a

maj ority of the Insured and Non Insured have heard of it.

Table 5.2 Knowledge about some selected insurance products other than health insurance schemes (%).

~eard about I£nsured IN on Insured (N=200) (N=200)

I) General Insurance products

ia) Motor (vehicle) insurance 87 78

ib) Shop insurance 51 34

c) Fire, house insurance 32 14

2) Non-general insurance product(s)

a) Life insurance policies 100 92

~) Pension schemes 67 21

Source: Primary Survey N= Total number of observations

100

It would be interesting to know about the access to infonnation.

particularly. about vartous types of HI products. We have classified the

health insurance schemes into four categories- 1) HI by general insurance

companies. 2) HI by non-general insurance companies. 3) HI provided by

community organizations. and 4) HI provided by hospitals. as shown the

Table 5.3. In the first category of health insurance provided by general

insurance companies. only 14% of the Non Insured have heard about

Mediclaim Policy while 100% of the insured are familiar with it (the selected

sample of Insured people are the Mediclaim Policy clients). First of all. we

can see that a majority of the Non Insured people are not aware of HI

products available in the market. As insurance is not an intuitive concept

for most of the people in a country like India. we can infer that absence of

enough knowledge about the importance and availability of HI products is

one of the main reasons for the low level of HI coverage. It is interesting to

note that among the Insured people (who are having Mediclaim policy) only

a few are familiar with other fonns of general HI policy. For example. only

15% of the insured have heard about Jan Arogya policy and 22% of the

insured have heard about Universal Health insurance policy. From this it

can be inferred that those who have been enrolled to the Mediclaim policy

are also not aware of other HI poliCies available in the market. in short.

they are not well-infonned consumers.

Table 5.3 Knowledge about different types of health insurance products (%)

Heard about jInsured (N=200) !uninsured (N=200)

I) HI by general insurance companies

I) MedicIaim Policy 100 14

2) Jan Arogya policy 15 2

3) Universal health insurance 22 12

2) HI by non-general insurance 7 4 companies

~) HI prOVided by community 2 1.5

organizations

4) HI provided by hospitals 11 8

Source: Primary Survey N= Total number of observations

101

Though there are many insurance companies providing HI policies, as can

be seen from the table 5.3, about 28.5% of the insured population knew

about other insurance companies prOviding HI other than the one they

bought at the time of joining while 71.5% of the insured population did not

know at the time of joining among which 41% of the population came to

know about other insurance companies providing HI after enrolling in to

the present provider. Further, among the insured people, 79% did not know

about the HI poliCies other than Mediclaim Policy at the time of joining.

Table 5.4 Knowledge about health insurance policies other than Mediclaim Policy at the time of joining- insured people (%)

Type of Knowledge Percent (N=200)

Did not know 79

Knew 21

Source: Primary Survey N= Total number of observations

One of the inferences that stems from the above discussion is that those

who have HI are also not aware of other forms of HI and HI providers. It

means that there is an asymmetry of information between the HI schemes

between the insurance providers and the people. Hence, the Indian HI

market is an imperfect market characterized by absence of proper

information.

5.3.2. Role of 'Insurance Habit'

So far, we have discussed about the information structure on the

information mainly in terms of knowledge by people about insurance

market. In this section, we are going more deeply on the information

structure to capture their familiarity with the insurance system as a whole.

For this, the present study takes into account the 'Insurance Habit' of

102

people in making decision to buy HI. The 'Insurance Habit'23 is a new

concept to capture the consciousness and familiarity of people with the

insurance system. Insurance system in India is an emerging market and

most of the people have no idea about the system and mechanism of

insurance. In such a society. it can be expected that higher the familiarity

of people with different forms of insurance, higher will be the probability to

join the HI, keeping 'other things' constant. The familiarity can be

measured in two ways-(l) their knowledge about different forms of

insurance. and (2) the joining status of the people with different forms of

insurance (I.e .. 'other forms of Insurance'). In this section, we conSider the

second aspect of the issue. It is quite possible that a person with an

insurance policy such as Life. Motor or any other forms of insurance to be

more motivated to buy a HI policy than his counterpart (provided that there

is no agency problem) because he will be more aware of the importance of

insurance and also about the modalities involved in joining, making

premium payments, renewal and finally about making claims. The

insurance habit is measured by classifYing the insurance system into two

components - 1) Life insurance with both saving and risk components

(R+S), and 2) General insurance with risk component (R) alone. As of now.

the Indian HI is having only risk component. therefore it is expected that a

person with general insurance would be more willing to join for HI than a

person with life insurance alone. Here. on the basis of the joining status to

insurance we can claSSifY the households into (1) with life insurance

only(S+R components)' (2) with general insurance only (R component). (3)

with both life and general Insurance, and (4) with no insurance (except HI

In the case of insured sample).

Table 5.5 explains the 'Other forms of insurance' status of both Insured

and Non Insured. The other forms of insurance mentioned here are those

bought only after buying HI (for health insured people). Among the Insured.

around 55% are having Risk Insurance and 68% are having Risk plus

Saving Insurance. When it comes to Non Insured people, only 11% are with

Risk Insurance but around 44% are with Risk plus Saving Insurance.

Further. the Insured and Non Insured households with both types of

23 Th t f' urance habit is developed by the present study borrowing some ideas e concep 0 inS d fin d th f 'Ii .ty f th

c t 11 d bankl'ng habit which has been e e as e ami an a e .rom a concep ca e . . . people with the banks by making various transactions-borrowing. depOSiting etc.

103

insurance are 37% and 9%, respectively, From this, it can be deduced that

the Insured people are featured With the status of more Risk Insurance

than the Non Insured, but there is not much difference about Risk plus

Saving Insurance.

Table 5.5. 'Other Insurance Enrolment Status' of the Insured and Non Insured (%)

jrypes of insurance scheme Insured Non Insured

1) Have Risk Insurance 55 11

2) Risk plus Sa\1ng Insurance 68 44 (excluding the pure risk insurance)

3) Both Risk Insurance and Risk 37 9 Plus Saving Insurance

Source: Primary Survey; N= Total number of observations

5.4. Asymmetric Infonnation Dissemination Channel on Coverage

and Infonnation Health Insurance

We have found from the above discussion that the familiarity of the people

With different concepts and aspects of HI are very low with people. even the

insured people are also not an exception to this. In such a situation. we

examine the main source of information on the HI scheme. Table 5.6

illustrates the main source of information on HI policy for both the Insured

and Non Insured.

Table 5.6 Main source of information on health insurance (Mediclaim Policy) scheme for both Insured and Non Insured (%)

Main source of infonnation Insured (N=200) Non Insured (N=200)

Insurance Agents 76 7

Media 2 1.5

Frtends. workplace etc 12 4

Office of the insurance 10 1.5

companies

Total 100 14

Source: Primary Survey; N= Total number of observatIOns

104

It can be seen that the main source of information for both category of

population is an Insurance Agent, Le., 76% of the insured and 7% of the

uninsured came to know about Mediclairn policy through Insurance

Agents. The role of media as a source of information is very less, only 2% of

the insured and 1.5% of the uninsured came to know about Mediclaim

through media. It is interesting to note that even though HI is a product for

the insurance companies, they are not giving sufficient publicity to It

through media that is reachable to all sections of the society. The other

aspect of the issue is that even if insurance companies are giving publicity

to HI products, It Is not powerful enough to reach the people because the

concept of HI itself is much broader and distinctive as compared to other

market products. Instead, Insurance Agents emerge as the main source of

information on HI for the people. Moreover, even if people come to know

about HI from other sources, as per the rules of the insurance companies',

people buy HI schemes only through licensed Insurance Agents.

The figure below presents the structure of PHI system and the role of

insurance agents.

Figure 5, 1. PHI model (Partner-Agent model)

Product

Insurer

Commission

Insurance Agent

Reimbursement

Health care provider

Product sale& Product servicing

Premium

CHent

Care& Bill ,!:D::::ac!.vm=e~n:..::t ___ -J

105

From the figure. it can be seen that an Insurance Agent is an important

stakeholder between Insurer and Clients. In this context. we examine the

role of Insurance agents in the scaling up of HI schemes and also their

potential role toward selection bias.

5.5. A Model of Insurance Agent's Rational Choice

In general. an insurance agent faces a situation of promoting a high income

or net profit oriented life insurance policies versus one or other high risk

oriented. 'after sales senice' oriented policies like the HI. In the normal

circumstances then. giving his individual choice and rationality. he may

choose only promoting and selling life insurance policies. However. either

because of official compulsions. social obligations and moral commitments

he may choose to promotes some HI poliCies. Therefore. the following type

of questions arises about his insurance promoting behaviors.

a) What will he do if the net income from selling HI is negative: but for the

fact that he is compelled to sell some?

b) What will he do if the net income from selling HI is positive. just as that

from the sale of life insurance poliCies?

An attempt is made here to model the behavior of a representative

insurance agent under two different circumstances: (a) when the net

income from selling HI is negative. and (b) the situation when income-wise.

it is attractive to promote HI also.

ConSider the possible hypothetical and very highly simplified situation. An

agent has fixed amount of time and effort (and energy) to spend on

promoting one or other insurance policies. Generally. it takes lot more

efforts and time to promote a HI policy than a life insurance policy.

Therefore. given his time or effort. there is a trade off between selling health

or life insurance or both in some mix. Stated with a simplified linear

fashion. mathematically.

106

(1)

Where. NH and NL are the numbers of health and life insurance policies that

can be sold using the entire of his time or efforts. Since. as compared to a

life insurance policy. it takes lot more efforts for the sale of one HI product,

it is expected that ~ is positive and greater than unity. and Q is positive.

Let nh and n, be the net profits or incomes by selling a single health or life

insurance products. respectively. The net total profit or income for all

efforts taken together can be stated as:

n = nh Nfl + n, NL (2)

Using (lJ. the same can be restated as:

n = Q n, + (n" - ~ n, ) NH (3)

For briefly. (nil - f3 n, ) is called e.

Case I: n" < 0: The net income per unit health product is negative. In other

words. considering the efforts (both immediate and after sale services taken

together) It does not pay to promote this product. From the expression (nh -

f3 nil = e, It then follows that e is negative, From equation (3) it then follows

that for any unconditional maximization of profits the agent would like to

choose only life insurance policies. and zero HI policies. That makes his

total maximized profits to be Q n,

However. if there are any compulsions (social. administrative. or legal) he

can choose at best a maximum of N H • and not beyond as shown in Figure

below. There are no incentives to sell HI policies otherwise.

Case 2: nh > 0: This is the situation of having some incentive to promote HI

poliCies as there are net positive gains. However. even in this situation.

there are two possibilities. namely.

(a) e > 0 :Thls will imply that nt.. the net profits from sale of HI is much

higher than nl. that from selling life insurance. which is only a hypothetical

case. Then. it pays fully to promote only HI poliCies as can be seen from

Figure below.

107

(b) e < O:This is a strong possibility. as the net profits from sale of life

insurances is likely to be more that from HI. and 13 is greater than unity.

Then. with e < 0 , once again, as in the situation of Case 1, the profit

maximizing strategy would be to use all efforts and time only to sell life

insurance, and not any HI policies. However, if there are any compulsions.

the agents can choose to sell HI up to a maximum of NH '.

In summary, a rational insurance agent in general, tries to avoid HI

schemes, unless compelled.

5.5.1. Insurance Agent and Selection Bias

We have already noted that client has more information on Ws health

status than Insurer. In contrast to this, Client has less information on the

available HI schemes. To recall. the present study considers the issue of

lnformation asymmetry on both the health risk of Clients and the

insurance schemes. In addition to the relative advantage of asymmetry in

information of Clients and Insurer on health risk and HI schemes,

respectively, the Insurance Agents also have some comparative advantage

on information on the two issues.

As far as information on health risk of client is concerned, the clients will

be having most information about their health risk. and insurance agents

will be having the least information about it as compared to clients but

more information as compared to Insurer. Stated mathematically.

IR < IRa < IRe ....................................................................... (I),

·IR,'. 'IR,' and 'IRe' refer to the level of information on the health status

(health risk) of the client by Insurer, Insurance agent and Client himself,

respectively.

There is high probability for the Insurance agents to know the Clients well

than Insurers because Insurance Agents will be mostly from the locations

of Clients.

109

When It comes knowledge about health insurance policy, the insurers Will

have more Information about it than clients but in , surance agents Will have more information on it than clients but less than or equal to that of insurer. The same can be stated as

IHI ~ IH" > IHo ................................................................. (2).

'!HI', ·IHa •• and 'IH,' refer to the level of information on the health insurance

scheme by Insurer. Insurance agent and Client. respectively.

We have already found that Insurance Agents are the main intermediaries

of information for the people about the HI. Therefore, the information

source through insurance agents can be considered as a strong indicator

for information on HI. In India. health insurance polices are sold to clients

on the basis of 'good faith· rather than of any kinds rigorous medical tests.

In this situation. the selection bias is better understood by Insurance

Agents because of his comparative advantage of information on the health

risk and HI policy. Thus, it is up to the discretion of the insurance agents24,

whether to supply infonnation to the clients on particular insurance policy

or not. and motivate them to buy HI policy packages.

In India. many insurance agents hold license to sell both the general (Shop

Insurance. Vehicle Insurance. HI etc) and non-general insurance (Life

Insurance) polices. As we are stUdying the case of only the individual HI

schemes. we do not consider the overseas HI schemes and corporate HI

schemes (and employees HI schemes) offered by the Indian Insurance

companies either through Insurance Agents or directly for the present

analysis.

In a situation where the insurance agents are the main source of

Information and have significant role in the level of HI coverage. the simple

241t has been widely accused that the main rea~on for the purchase of life Insurance policy 'd th "b' h' ", on'enled counseling of the Insurance agents, It has become a IS ue e raIn was In", '

b I, f th eople that some insurance agents misuse therr personal relation common e Ie among e p b th I t JOI

'n 'or an insurance pollcy. Thus. people have a negative y compelling e peop eo" c . gents as a person approaching them .or hIs own material attitude towards Insurance a d C I

I, th d I I to buy insurance (llfe) may be a force one .or many peop e. we .are. In short. e ec s on

110

question here is do they have any role in th ' e resulting of adverse selection

or cream selection and in case it is so why h ld th ' s ou ey have such roles?

As people are more famil' 'th th ' , lar WI e hfe Insurance pOlices where they get the facility of saving in add it' t th ' lOn 0 e nsk coverage (saving plus risk insurance), the populartty for life insurance is high25 A t thi ' s a response 0 's,

the numbers of insurance agents are also increasing, In this context, some

of the insurance agents are selling health insurance to those who bought

life insurance as a complimentary product26 , That may be one reason why

many of the health insured have other forms of insurance coverage also,

which we ha\'e found in the earlier discussion under the title of 'insurance

habit'. Moreover. selling HI to those who have other forms of insurance may

reduce the average cost of HI for the insurance agents, It is possible

because the time and effort needed to sell a HI policy to a person who is

very much familiar to insurance system is very less.

The probability of making a claim is very high in HI as compared to life

insurance or any other insurance policies. Many insurance companies

appreciate their agents for selling insurance poliCies to the low risk people

to reduce the insurance claims: otherwise agents may face a loss of their

reputation with their Insurance companies. At present, in India the

enrolment into HI policy is mainly on the basis of good faith rather than

any medical test: an insurance agent can sell HI policy to those he prefers.

As the people themselves know much more about their health status and

future health needs than insurance company. an insurance agent also has

some more information about the health status of his prospective customer

than insurance company, In other words. an insurance agent has more

information on the health status of customer than an insurance company

and has lesser information than the customer. As insurance agents have

relatively more informational advantage of the health risk of the potential

customers. they will sell HI policies to those known to them. Even though

the final decision to issue and renew the insurance policy (the insurance

25 Although the Insurance awareness and habit are very low in India. it can b~ expected to , h b't with the l'ncrease In income and matenal well bemg of people, mcrease the Insurance a 1 , , th ~ f which may result in the propensity to save and also to have coverage for nsk m e onn 0

Insurance, t II I 26 m etition among the Insurance agents 0 se non-genera

Because of the Increasing co P of the agents try to concentrate on the sale Insurance schemes such as life Insurance. som~d other fonns of Insurance. of general Insurance. and sell health insurance

III

contract is generally for one year) to people I'S at th d' . . e Iscretion of msurance

company, an insurance agent has the crucial role in selecting the clients

for HI. Given the different pay-off situation that has been before an

insurance agent, it is expected that an insurance agent will not only try to

reduce the sale of HI policy but also will sell to low risk people. Thus by

selling HI to low risk people will not only increase the reputation of

Insurance Agent with the Insurance companies but also reduce the many

cost incurred by him for helping the Clients to make insurance Claims. In

such situations, there will be less chance for adverse selection in the PHI

market.

Further, an insurance agent prefers to sell HI to the high income people, It

can be mainly due of the following reasons:

1) A protlt maximising insurance agent prefers to earn maximum of net

profit from the sale of each HI policy by maximising profit and reducing the

average cost. A High Income household perhaps buys high amount of HI

than a low income household. As the income for the insurance agent is

proportional to the insurance amount sold out, it means an insurance

agent could maximise the net profit from the sale of each HI policy.

2) Several studies on the relationship of income on health status state that

income has a positive impact on the health status of the people. It can be

expected that high income people will be healthier than the low income

people. Therefore. selling HI to the high income household means that the

HI is sold out to low health risk people, therefore, an insurance agent can

reduce adverse selection by selling HI to high income households.

In the above discussion, we have found that Insurance Agents have a key

role in the scale up process of PHI in India. Increasing the profit rate for the

Insurance Agents for the sale of HI by Insurance Company with

government subsidy can be considered as one solution for the scale up of

PHI in India. However, the scale up process of HI is not completely

determined by the behaviour of Insurance agents; it also depends on the

'Insurance Habit' of the people. The 'Insurance Habit', an indicator of the

familiarity of the people with different forms of insurance and insurance

b . d by promoting other forms of insurance too. system, can e Increase

112

In the next section. we empiIically test the theoretical model that we have

been discussing so far. As a first step. we will empirically test whether

adverse selection is there or not in PHI market.

5.6. Empirical Estimation of the Presence of Adverse Selection: Significance of Health Risk

By now it is clear that the main vartable that determines the selection

bias In HI market In the fonn of adverse selection or cream selection Is the

status of health Iisk of the clients. But it Is not an easily quanttfiable

entity. Demand for HI is a deIived one from the demand for health care

services. Health Iisk Is a mUltidimensional phenomenon and is

detennined by many observable and non-observable factors. Many studies

that we have cited earlier have used the self-reported health status by the

people as an indicator for the health Iisk. The reported health status is

measured on a scale ranging from I to 5: 1) very good. 2) good, 3) average,

4) bad. and 5) very bad. Thus. people reporting 'very good' and 'good'

health is categoIized as low Iisks and those reporting 'bad' and 'very bad'

are categortzed as high risks. The self reported health status is highly

subjective that may vary according to perception and understanding of

the respondents. Other indicators used to measure health risk are age,

gender. working conditions etc. For example. presence of elderly

population and women members in the family, and people working at

high-Iisk conditions are considered as indicators of high health risk. As

we are considering household as the unit of analysis, we find the presence

of many elderly and women population In the households Irrespective of

whether they are Insured or Non Insured. therefore. it is not appropIiate

to consider age or gender to measure health risk of people for the present

analysis. Yet another measure of health risk is the revealed infonnation

on health care expenditures. Therefore. the actual health care expenditure

of the household can be conSidered as another indicator to measure the

health risk. A household with higher health expenditure is considered as

high risk as compared to a household with low health expenditure. This

t h any serious limitations. The literature reveals that measure men as m health expenditure is positively income elastic and also vartes according

113

to the health care seeking behavior27 . If we use health care expenditure as

an indicator to measure health risk in the present analysis we have to

make a comparison of the health care expenditure between the Insured

and the Non Insured after adjusting it with the income ability and the

health care seeking behavior of the households. However. the health care

expenditure of the Insured households is also influenced by their behavior

of moral hazard in terms of over utilization of health care services. Hence

we cannot take health care expenditure as an indicator to measure the

health risk of the households either.

Another indicator to measure the health risk is that whether members of

the household have any 'bad health or bad medical situation' in the form of

permanent health problems. the same illness again and again, and chronic

health situation. We found this way of measuring health is more reliable

and more objectively focused than the self reported health status, which is

highly subjective and biased. Hence, we use the 'bad health or bad medical

situation' as an Indicator of the health risk of the family. Therefore. to test

the issue of adverse selection for the present analysis, at least one family

member reporting 'bad health or bad medical situation' is coded as a 'bad

risk'~8.

d f d'cal care is more income elastic in the poorer, 27 The empirical evidence shows Ihal the demlan or meElngel curve estimates for medical care in Birdsall

, , 'h h r tndustna countnes. . I' ' . developtng countnes than 10 t e nc e . . I t' 't'es close to unity whereas loCO me e astlcllies

( 1981) Irl loCO me e as ICI I ' and Chuhan (! 986) and Musgrave . rep', .. (V n de Ven and Van der Gaag 1982. Holtmand

, f d f 'duslnal countnes a . between 0 2 and 0 , are tYPically oun or 10 d G man 1978 Phelps 1975, and Manning et aI

,.. '1978 Goldman an ross ' and Oben 1978. Colic and Grossman . f I income and high income households. 1987) The same findings can be applied to the caseD .0

1 W f PHI insured from health insurance coverage. and

2 ' . • '1 ded in the taml Y 0 II Moreover some members an.:: exc U I h ' k . f h asurement of hea t ns .

therefore, we have omitted them rom t e me

114

Figure 5.2: Households reporting bad health b d at least one among the members in the or a medical situation

I 100 ~

~ 90 88

~ 80 <II

:5! 70 0 .J:.

'" 60 ., :::0 0

50 :z: -0 40 '" 01 !l 30 c: 22 '" u 20 1 ~

II ... 10 1

L 0

Insured Households

Sources:Prtmary Survey

household in PHI scheme (%)

67

33

Non Insured Households

o Bad Health

• Good Heallh

In testing for selection biases. we have to consider the society as one entity

and have to see out of the share of high risk Insured to high risk Non

Insured. Thus. It can be seen that 22% of the PHI clients are with high risk

whereas among PH Non Insured It Is 33%.

5.7. Econometric Estimation

It is clear that the joining status of the health insurance for a household is

determined by factors such as the familiarity of people with different

aspects of Insurance. the role of insurance agents and health risk of the

household members. So far. we have analyzed the factors determining the

decision to Join HI scheme In a bivariate framework. In this section. we

discuss the significance of each factor on a decision to buy HI with the help

of a multivariate analysis. Therefore. we undertake a maximum likelihood

estimation of binary Probit model on the probability to buy HI. Moreover.

an empirical Justification of the theoretical model presented earlier is

attempted in this section.

115

If a household has HI. the revealed probability of having HI is one; hence

the dependent vartable is equal to one. and the variable is zero if the

household does not have any health insurance. Thus the specification for

the Probit analysis Is based on a binary dependent Variable.

The response vartable Y*i ~ I3'Xi + Pi (1)

Where Y*i is unobservable. What we observed is a dummy Variable y

defined by

(have no HI).

y = 1 if Y*i >0 (have health insurance) y = 0 otheIWise

As (3'X is E (Y*1 / xil. we get

Prob (Yl = I) = Prob (lli > - Wxd

= 1- F (- f) ·xd. Where F is the cumulative distIibution function for p.

Table 5.7 Definitions of variables

Variables Definition Healthinsurance 1 if the household has health insurance. (Dependent o otheIWise VaIiableJ

Income Percapita annual household income of the household

Education Years of schooling

Householdsize Total family size of the household

Rlsklnsurance 1 = If any of the household member is having the pure Iisk insurance. o otherwise

Rlskpl usinsurance 1= If any of the household member is having the risk plus saving insurance. o otherwise

Healthrisk 1= If any of the household member has bad health situation; o otherwise

member had been of the household Information 1= If any Insurance Agent to talk about approached by an

health insurance ;0 otherwise

116

Table No 5.8 Probability t h o ave health i model results of PHI h nsurance coverage- Probit sc emes

Dependent variable: 1 if the household h h o Otherwise as ealth insurance;

Variables Coefficient Marginal Effect Income .00005 .00002

(5.23)*

Education .06855 .02685

(1.92)**

Householdsize -.05721 -.02241

(-0.46)

Healthrisk -.46372 -.18300

(-1. 76)***

Riskinsurance l.2674 .44072

(5.18)*

Riskp1USinsurance .45797 .17872

(2.19)**

Information 2.0419 .66124

(9.18)*

Constant -3.3879

(-3.97)*

Y - Pr(insured) (predict) .57578

Log likelihood -9l.3371

LR chi2 (10) 37l.84 ,

Pseudo R2 0.6706

Number of observations 400

Values in the parentheses refer to the 'Z' statistics Level of statistical significance: • = 1 %: •• =5%: ••• = 10%

From the above model. it can be seen that as against the theoreUcal

expectation there is an indicaUon of low health risk preference on the decision

to opt for HI. If there were an adverse selection problem. the insured people

would be having bad health risk. In other words. there must be significant

difference between the Insured and uninsured about the health risk. But tht'

present model indicates that there is a statistically negative and significant

117

difference between the health risk statuses of both g Th roups. is means that PHI is characterized by more healthy people than what th th . e eory predIcted. Hence, there is no evidence of adverse selection in PHI market. In support of

this, from the primary survey we have evidence that a total of 12% of the sick

members of the insured family had been excluded from the PHI coverage

mainly because they are Sick (unhealthy). Thus, the present study accepts the

hypothesis on the presence of adverse selection problem in the Indian

voluntary HI market.

Further, the income of the households has a statistically significant role in

deciding to opt for HI. It means that higher income of the households has a

positive impact on the decision to go for HI. Thus, it can be inferred that the

market is over represented by high income people resulting in horizontal cross

subsidization instead of vertical cross subsidization, where rich pays the

health care cost of all the rich, which means that Indian voluntary HI is not a

sound risk pooling mechanism. At par with the theoretical expectation, the

model reveals that the educational qualification of the households has a

positive impact on the probability to have HI. But it can be seen that 'Risk

Insurance' and Risk plus Insurance' enrolment status of the people have a

positive and practical significance on their decision to opt for HI which is a

clear indication that "the higher the familiarity of the people with the different

forms of insurance. higher will be the probability to join HI scheme, keeping

other things constant". We have already noted from the profit maximizing

behavior of Insurance Agent that he will be more interested to sell HI to those

who have other forms of insurance as a compliment mainly because it will

reduce his Average Cost of selling HI policy. The main implication of this

finding is that the insurance education in the form of familiarity with different

forms of insurance has much role in the probability of getting health insured.

Another related policy implication is that the information supplied by the

Insurance Agents are more powerful than whatever the insurance companies

have been giving through media and their offices.

118

5.S. Chapter Summary

From the above discussion, it is obvious that absence of proper information on

HI schemes Is one of the main causes for the low level of HI coverage, The PHI

coverage is highly influenced by Insurance agents and other insurance

enrolment status of the insured. Even though there is no sophisticated

monitoring mechanism to control adverse selection in PHI, as against the

theoretical expectation, there is no empirical evidence for adverse selection in

PHI coverage. We have found that the Insurance agent can significantly

influence the scaling up of and also the 'selection biases in PHI schemes.

In the next chapter, we revert to our discussion on the equity in MHI coverage

and extend our discussion on the issue of selection bias in MHI schemes on

the scaling up process.

CHAPTER SIX

SELECTION BIAS IN MICRO HEALTH

INSURANCE SCHEMES

6.1. Introduction

In the previous chapter we had analyzed the role of an insurance agent in

the scale up process of HI and also examined the ways in which his

behavior influences the selection bias in the PHI scheme. We found that a

situation of no adverse selection would increase the income of an insurance

agent and motivates him to avoid adverse selection. The empirical finding

also supports the fact that there is no adverse selection in a PHI scheme.

Now. let us extend the discussion on selection bias in the MHI schemes. We

have already noted that while insurance agents act as information

dissemination channels in PHI schemes. the SHGs demonstrate such a role

in MHI schemes. In this chapter. we examine the case of selection bias in

MHI schemes. First. it is examined whether across MHI schemes selection

bias is there or not. Further, the sources of adverse selection across

various income classes are analyzed and the role of SHGs in such outcome

is also examined.

6.2. Adverse Selection in MHI schemes

In the ECCP data for the MHI schemes, the health risk is measured by

asking a question to the respondents "Do some people in your household

have the same illness again and again or even permanently? These can be

illnesses like diabetes, high blood pressure or the like." Answers as 'YES'

and 'NO' are coded as high health risk and low health risk, respectively,

h ld ·th hl'gh health risk are presented in the The percentages of house 0 s WI

table below.

120

Table 6.1 Households reporting bad health Or bad medical situation (high risk) at least one among the members in the household in MHI schemes (%)

HI schemes Insured Non Insured Karuna Trust 11.36 9.77 Yeshasvini Trust 17.05 10.45 Dhan 14.72 8.83 Uplift 21.03 13.03 VHS 33.71 18.23

Sources. ECCP data

The econometric models that we have estimated in Chapter 4 on the

probability to have HI have also included a variable on 'Health Risk', i.e.,

the Variable distinguishing the household with high health risk and the low

health risk. a variable that we used to measure the significance of adverse

selection in PHI scheme. We summarize the findings of the same model

across each MHI schemes in the following table.

Table 6.2 Probit model results specific to the probability of the high risk to have health insurance

MHI schemes Mar--,inal Effect I) Rural MHI Schemes (Karuna Trust, .0998 Yeshas\\.'ini Trust and DHAN) (2.94)" I) Karuna Trust .0444

(0.71) 2) Yeshasv,'ini Trust .1153

( 1.94)"" 3) DHAN .1483

(2.50)" ill Urban MHI Schemes (UPLIFT and VHS) .1827

(5.53)" I) UPLIFT .0109

(2.88)" 2)VHS .2190

(4.94)" , , tics' Values in the parentheses refer to the Z statis ,

" 1°1. .•• -5%' ••• = 10% Level of statistical significance: = ,0, - ,

It is obvious from the above table that in the 'Rural MHI' schemes, the

HI Ie the high-risk households is 10% more as probability to have ,or

ld S' '1 ly in the 'Urban MHI' schemes the compared to low risk househo s. Inn ar

h ld are 18% more likely to have health insurance, high health risk house 0 s 'R al MHI' schemes and the 'Urban MHI' schemes have In short, both the ur

N xt we test the adverse selection problem the adverse selection problems. e ,

Among the 'Rural MHI' schemes, the Karuna within each MHI schemes.

121

Trust does not have any adverse selection bl B pro em. ut, the Yeshaswini Trust and the DHAN Tru t h s ave adverse selection problems. Among the 'Urban MHr schemes. both the UPLIFT d th

an e VHS have adverse selection problems: however. the VHS have a very low level of it.

In the last chapter we have been diSCUSSing ab t th . ou e eqUIty aspects of HI coverage across both the MHI schemes and PHI scheme. We have found

that the MHI schemes are successful in giving access to HI coverage for the

low-income classes in their operational areas. Although they have been

giving access to low-income classes in the rtsk pool. it would be more

equitable if the schemes cover a large number of lOW-income people

through the scaling up of HI. One of the constraints for such scaling up is

the issue of adverse selection. The econometrtc results show that the MHI

schemes sulfer from adverse selection problem except for the MHI provided

by the Karuna Trust. One of the outcomes of adverse selection will be an

increase in the HI premium. which may motivate the low risk people to

drop out from the rtsk pool that may again lead to an increase in the

premium. Nonetheless. such problems would affect the sustainability of the

risk pools. Moreover. it will be difficult for the low-income households to

afford a high premium on health insurance coverage and may force them to

leave the rtsk pool. which is perhaps against the equity objectives of health

Insurance coverage. However. this issue should be viewed in another

perspective too. The fact of the matter is that if the adverse selection has

emerged mainly among the low income households. we can not argue that

It is a situation of welfare reducing outcome. According to the conventional

theories of demand for HI. referring to the additional health care when

persons become insured. moral hazard is a welfare loss to the society (M. V.

Pauly. 1968. 1983: M. S. Feldstein. 1973: W.G.Manning and M.S.Marquis.

1996). whereas according to the new theory proposed by J. A Nyman

(2004l, moral hazard Is welfare promoting as it will give access to health

care to those who can not afford otherwise. In a similar way. if we look at

the issue of adverse selection. another form of market failure. according to

the conventional theory it is a welfare loss; but if the adverse selection is

caused by over representation of unhealthy from the low income

households. it is a situation of welfare promoting and more equitable on

th . f th theory of demand for HI. Thus. adverse selection

e perspectlve 0 e new

122

due to the enrollment of low income households in MHI scheme can be

considered as a situation of welfare promoting as it gives access to health

care to unhealthy people of the bottom sections of the income pyramid.

Furthermore. adverse selection from low income households can be

considered as equity enhancing and welfare promoting outcomes as long as

it will not adversely affect the sustainability of the risk pool. In this context,

we will investigate the SOurces of adverse selection, i.e., whether it is

coming from the lOW-income people or from the high-income people.

We now investigate the sources of adverse selection in MHI schemes with

the help of a binary Problt model on the probability to have HI. As we had

done earlier. households with HI are coded as T and those without health

insurance coded as '0'. We have claSSified the three income classes

consisting of 'Low Income', 'Middle Income' and 'High Income' with and

without high risk, coded in to six dummy variables, Among this six dummy

Variables, in the first estimation the "High Income without high health risk'

Is taken as the reference category. Furthermore, the model is re-estimated

by changing the reference category too. to give us insight of the sources of

adverse selection within each income class.

Table 6.3 Definition and measurement of variables

Variables Description Hhsize Household size SHGmember Household is a member of SHG 1; 0 otherwlse Lower_edu Highest Educational qualification of the household is till 4th

years of schoolln~ I; Otherwlse-O Medlum_edu Highest Educational qualification of the household is 5 th to

10th years of schooling -1; Otherwlse=O 11th Higher_edu Highest Educational qualification of the ~ousehold is

and aboveyears of schooling -I; Otherwlse-O Headeducatlon Highest education attained by the head of the family,

measured in terms o(y~ars of schoolin~ Lowinc_highrisk Low Income household with at least one m~mber of the

households is with hi~h health risk = 1; Otherwlse=O . h hold with at least none of the member of LoWinc_ lowrisk Low mcome ouse . 0

I h Ids is with high health risk= 1; Otherwlse= the lOuse 0 h - h . hold with at least one member of t e Midinc highrisk

Middle i7~~~ewit~u~fgh health risk -1; Otherwlse=O househo h ehold with at least none of the member Midinc lowrisk Middle inco~e Id ous with high health risk = 1; Otherwlse=O of the house 0 s IS h Id with at least one member of the

Higinc highrisk i h - 'orne house 0 H g me . with high health risk - 1; Otherwlse-O households iSh h Id with at least none of the member of

Higinc lowrisk H h' come ouse 0 - . ig m d' 'th hi~h health risk -1; Otherwlse-O the househol s IS Wl

123

In the model, we have taken 'High Income With low lisk' as the reference

category: in short, all our compansons are in terms of that group, Equally

we would like to know the probability that how the 'low income class with

high lisk' is likely to have HI as compared to 'Low Income With low lisk' and

the 'Middle Income class With high lisk' as compared to the 'Middle Income

with low lisk', and the 'High Income with high lisk' to 'High Income with

low risk', For this purpose we have estimated the Probit model by changing

the reference category, first by taking the 'Low Income With low lisk' and

secondly by taking 'Middle Income With low lisk', and also 'High Income

with low risk' as the reference categolies separately in each estimation. As

we are changing only the reference category, it does not alter any other

parameters and test values of the model. In the table below we present the

results by taking the 'High Income with low risk' as reference category and

compare with probability of other income groups with high and low risk. In

the subsequent table, we present Probit results on the likelihood of each

income classes with high risk of having the HI as compared to low health

risk households of their respective income classes.

Table 6, 4 Probability to have health insurance coverage- Marginal effect of Probit model estimate of Rural MHI schemes

Dependent variable: 1 if the household has health insurance; o Otherwise

Variables Rural MHI Karuna Yeshaswini DHAN schemes

Hhsize .00172 -.01862 .01568 .03773 (0.26) (-1.52) (l.40) (2.17)'*

SHGmember -.00008 .OIl34 ,15364 -.00015 (-0.28) (0.29) (3.44)' (-0.51)

Lowecedu -.16554 -.15209 -.23646 -.14011

(-4.90)* (-2.81)* (-3.85)' (-2.13)**

Medium_edu -.09813 -.10835 -.1273 -.03309 (-3.64)* (-2.13)" (-2.91)* (-0.66)

Headeducation -.01002 -.012372 .00718 -.02830

(-3.83)* (-2.60)* 0.59) (-5.47)'

-.11187 .15835 -.22405 -.12222 Lowinc_highrisk ( 1.61)*" ( 1.24) (-1.87)**' (-1.11)

.16603 -.31141 -.13516 LoWinc lowrtsk -.04930 (-6.65)* (-2.50)* (-1.62)*" (3.34)'

.09293 -.08806 .15470 Midinc highrisk .03331 (0.97) (-0.91) ( 1.52) (0,57)

124

Variables Rural MHI Karuna Yeshaswini DHAN schemes Mldinc_lowrisk -.07366 .038705 -.10560 -.13464 (-2.71)- (0.78) (-2.12) (-2.83)-Higinc _highrisk .18328 .08873 -.00280 .13383

(3.56)- (0.92) (-0.03) ( l.49) v-Pr(lnsured)(pred1ctj .49854 .50233 .48957 .49952 Log likelihood -1432.90 -471.79 -437.35 -472.89 LR chi2 (10) 66.16 26.79 90.01 52.35 Pseudo R2 0.0226 0.0276 0.0933 0.0524 Number of 2115 700 696 720 observations Reference c.lto.:ories. HIglIlC_ lownsk. HIgher level of educatIOn

Values in the parentheses refer to the 'Z' statistics: Level of statistic-al signific-ance: • = 1%:" =5%: ••• = 10%

Table 6. 5 Probability to have health insurance coverage- Marginal effect of Probit model estimate of Urban MHI schemes

Variables Urban MHI UPLIFT VHS schemes

Hhsize -.03041 -.04189 -.01269 (-344)- (-3.09)- (-1.03)

SHGmember .00010 .00019 .00010 (052) (0.48) (0.41)

Lower_edu -.18711 .19313 -.08724 , (-3.20)- (-2.46)- (-0.90)

Medlum_edu -.00149 .08026 -.05130 ( 0.05) 0.76)-** (-1.16)

Headeducation -.00282 .01027 .01392 (-0.71) (1.82)--- (-2.40)-

Lowinc_hlghrisk .23339 .14187 .29803 (4.48)' ( 1.50) (94.86)-

Lowinc_lowrisk .01596 -.08938 .08134

(0.42) (-1.52) ( l.34) .27320 .20966 Mldlnc_hlghrisk .23179

(3.85)- (3.74)- (2.49)' .04517 .03331 Mldlnc_lowrtsk .05065

(1.28) (0.89) (0.49)

.10407 .19445 Hlglnc _hlghrisk .12512 ( 1.20) (2.33)-' (2.05)'-.493329 .50347 '.i Pr(lnsured)(predlct) .49817 -453.34 -462.45

Log likelihood -940.54

LR chi2 (10) 54.16 52.58 35.77

0.0280 0.0548 0.0372 Pseudo R2

692 693 Number of 1396 observations

er level of educaUon Reference calegories: HI!(lnc_ lowrlsk. High

£ t the 'Z' statistics: Values in the parentheses re e~ _ ~ 1%' •• =5%: .,. = 10% Level of statistical significance. - .

125

In the tables above, the probability to hav ' e HI by each mcome class as

compared to low risk of High income cia (H" ss Igmc_Iowrtsk) is presented, In

the Rural MHI schemes, there is a statistically signlficant difference

between the health risk of both the Insured and th N I e on nsured among the 'Low Income' Households. In other words th' d ' ere IS no a verse selection problem from the 'Low Income' households of th MHI h e sc emes, Among the 'Middle Income' households, the households Wl·th h'gh ri k I S are 10% more likely to join the HI as compared to the households with low risk of the

same income class. Interestingly, the 'High Income with high risk' is 18%

more likely to have health insurance as compared to low risk households of

the same income class. In short, we can see that if we consider each

income as separate entity, Insured households among the 'Middle Income'

and the 'High Income' classes are high risk as compared to the Non Insured

households of the respective income classes,

So far we ha\'e discussed the sources of adverse selection within each

income classes in the 'Rural MHI' schemes, Next we examine the

probability of high risk and low risk households of each income class as

compared to the 'High Income with low risk', Thus, as compared to the

'High Income with low risk', the 'Low income households with high health

risk' is 11 % less likely; and the 'Low income Class with low risk' is 5% less

likely to join for health insurance. In other words, the source of adverse

selection is not from the Low Income households as compared to the High­

Income households, In short, the main source of adverse selection in 'Rural

MHI' schemes is the presence of high-income class, We have found in the

earlier chapter that the 'rural MHI' schemes have covered less proportion of

the 'Low Income' households as compared to the 'High income' households,

Moreover, our finding from the above model reveals that the 'High Income'

households who are members of these schemes are high risk also.

In the case of the 'Urban MHI' schemes, being in the category of 'Low

Income' households with high risk tncreases the likelihood to have HI by

22% as compared to the low risk of the same income class. Similarly, being

in the cat{!gories of 'Middle Income' and 'High Income' households with high

risk increases the probability to have HI by 19% and 13%, respectively, as

h I ri k households of their respective income classes. As compared to t e ow s

126

compared to the 'High Income with low risk', the 'Low Income households

with high risk' and 'Low Income households with low risk' is 23,3% more

and 2% more likely to join for HI. respectively. Likewise, being in the

category of 'Middle Income' households with high risk increases the

likelihood to have HI by 23% as compared to the low risk of the same

income class. To recall a finding from the previous chapter on the

proportion of various income classes in the 'Urban MHI' schemes, we found

that these schemes have covered more proportion of low-income people as

compared to the high-income class in the risk pool.

In the above discussion we extended the empirical estimation of the

adverse selection in MHI schemes which reveals that both the 'Rural MHI'

and 'Urban MHI' schemes are having adverse selection problems. It was

also found that the main source of adverse selection in the 'Rural MHI'

schemes is the high-income class as compared to the low income, but the

schemes hm'e covered higher proportion of high-Income households. The

empirical findtngs prompt that the 'Rural MHI' schemes have not only

covered proportionately more of the high-income people in the insurance

schemes but also there are more high-risk households from the high­

income households in the risk pool as compared to the low-income

households. From this point of view, a first theoretical prediction is an

increase in the premium to be paid which may result in the drop out of the

low risk households from the risk pool. In such a situation the low-income

class will be more probable to go out of the risk pool, which is against the

equity objectives of health insurance coverage. But the 'Urban MHI'

schemes have covered more proportion of low-income households; at the

same time, the source of adverse selection is alSQ from the same income

group.

However, a full interpretation of the model is beyond the scope of our

current discussion. We therefore would limit the discussion to the relevant

parameters. We are interested to know the probability of having HI between

high risk and low risk within each income class. Therefore, we re-estlmate

the Probit model thrice by changing the reference categories as Lowinc_

10WIisk, Mldinc_ lowrisk, and Hlginc_ lowrisk, respectively. Only the

relevant parameters are presented below because all other values and

127

properties of the model are similar to the value th t s a we presented in the above tables.

In the table below, first let us have a look at the case of adverse selection

within each income class:

a) Low Income Class Households: In the case of low income class, it is

obvious that there is no statistical significant difference between high rtsk

and low rtsk to have HI in the Rural MHI schemes such as Karuna,

Yeshaswin1, and Dhan. Therefore we can conclude that in Rural MHI

schemes both the high rtsk and low rtsk have equal probability to have HI.

However. in Urban MHI schemes, in UPLIFT has more adverse selection,

that is. the high rtsk of the low income class has 22% more probability to

have HI as compared to the low rtsk of the same income class. In contrast

to this, in another urban MHI scheme called VHS, the high rtsk has 23%

less probability to have HI: it can be a case of cream selection of excluding

the high rtsk from enrollment.

b) Middle Income Class Households: Among the middle income class

the probability to have HI for the high health rtsk as compared to low rtsk

is statistically insignificant in Karuna, Yeshaswini and Dhan. however, in

aggregate there is adverse selection in Rural MHI schemes (10% more).

Among the Urban MHI schemes, the high risk UPLIFT and VHS are 24%

and 18% more probable to have HI coverage, respectively.

id 'Rural MHI' c) High Income Class Households: When we cons er

schemes and 'Urban MHI' schemes, the high risk are 18% and 13%

probability to have HI, respectively. However. when we conSider the MHI

h ·t wi only l'n the VHS the high rtsk case is statistically sc emes as urn se. different from the low rtsk case to have HI coverage. by about 19% more.

128

Table 6.6 Probability to have health insurance coverage in MID scheme- Marginal effect of Problt model estimate of some selected parameters

Rural KlUUDa Yesbas DHAN Urban

MHI Wlnl MHI

Lowinc_highrisk -.063 -.005 .087 .012 .219

(Reference (-.88) (-004) 10.62) 10.11) (4.39)'

Category:

Lowinc _loWTisk)

Mldlnc_hlghrlsk .106 054 .017 .271 .186

(Reference 11.84)" 1057) (0.17) (3.14)' (3.08)'

Category:

Midlnc_lowrisk

Hlglnc _hlghrlsk .183 .088 -.0021- .133 .125

IReference 1356)' 1092) 0.03) 11.49) 12.05)"

Category:

H Iglnc Jo\lIrlsk)

Reference categories: Lov.;nc_lowrisk: MldlncJownsk: HIgmc_lowTlsk;

Values In the parentheses refer to the 'Z' statistics; Level of sta tistical significance: • = 1%: .. =5%: ••• = J 0%

UPLIFT VHS

.224 -.234

(2.55)' (-3.95)'

.235 .179

(3.09)' (2.13)"

.104 .194

11.20) 12.33)"

We have already noted that except for KanIna Trust. the other four MHI

units suffer from adverse selection. While splitting the source of adverse

selection among the three income classes. none of the income classes is

found to be significantly causing any adverse selection. In the case of

DHAN. the adverse selection is significantly contrtbuted by the middle

income classes.

However. in UPLIFT the adverse selection has spread across all the three

income classes. Interestingly. in VHS instead of adverse seleCtion. there are

evidences of creamy selection among the low income class which means

that the relatively lower risk are enrolled in the HI among the low income

households. Moreover. there is adverse selection in VHS from the middle

income and high income households.

129

6. 3. Role of SHGs in Adverse Selection

We have presented the structure of MHI schemes in the previous chapter

and found that SHGs act as both infonnation dissemination channel on HI

and intermediary between MHI units and the Clients. One of the

distinguishing features of the SHGs is that they are not-for-profit agency as

compared to the profit driven Insurance agents who maximize profit in the

PHI market. Moreover. SHGs do not have the issue of trade off between the

issue of selling health insurance and other fonns of insurance. As MHI

units and the SHGs target the low-income people even though the risk

pools need high-income people also as members. due to the issue of

sustainabillty of the risk pool, one could expect that the SHGs will not only

target the low-income people but also those who are in immediate need of

health care (who have high health risk) among the low income class.

:l: OJ > co

.J:!

0 ->-.t::

.D ..

.D 0 ~

Il.

~

~

2 _

Probability to have HI coverage for the households of each income class with SHG membership as compared to the

households without SHG memership of the same income class, by various income class across MHllocations

-- • Low Income

• Middle Income

o f:liiJh income

0%

--~ --r-

Yeshas~ni DHAN -20% -----

-40%

MHI Locations

th t on the role of SHGs to have HI It can be seen from the figure above a h t SHGs play significant role on having HI coverage. we have found t a

130

coverage in the three MHI units (YeshasWini, DHAN, and UPLIFT) except

for Karuna Trust and VHS. In VHS we have seen that there are evidences

of creamy selection among the low income. In fact. the presence of SHGs

is very insignificant in VHS. It means that in VHS the HI is sold directly to

the clients and it is leading to creamy selections (exclusion of the high

risks) among the low income class. In UPLIFT, the role of SHGs is very

high to have HI coverage and there is adverse selection from the low

income class. As the SHGs are not-for-profit agenCies. it targets not only

the low-income people (a case of equity promoting behavior of SHGs) but

also those who are in immediate need of health care (who have high

health risk) among the low income class. So there is high probability for

adverse selection in risk pool by selling HI through the SHGs. However, as

against the theoretical expectation. there is no adverse selection in many

of the MHI schemes where SHGs have significant role for having HI

coverage. As the SHGs sell HI to majOIity of its members where selling HI

to heterogeneous group rather than to selected individual is a method to

prevent adverse selection. the presence of SHGs in the scale up process of

MHI schemes would result in no or low level of 'selection bias'. it can be

cited as one of the reasons for the low severity in adverse selection in MHI

6. 4. Chapter Summary

Both the 'Rural MHI' and the 'Urban MHI' schemes are subjected to

bl i re severe in the sense that source adverse selections. Their pro em s mo , f th 'ncome classes which they have of adverse selection IS also rom ose 1

d to other income groups. There given relatively more coverage as compare

is a mixed result when we investigated the source of adverse selection

across various income classes. Given the positive role of SHGs to have an

population. SHGs did not promote adverse equitable HI coverage for the

selection as such.

, , of the members may perhaps buy HI In each SHG. 29 It may be due to the fact that ma)onty d th i dividual health risk will

g the members an e n then the health risk will be spread arnon , , HI to majority of the SHG members results m no

become InsIgnificant. In short. selhng , '

th arne time ensure adverse selection and ate s HI coverage for the hIgh nsk too.

131

So far in the previous and in the present chapters, we have examined the

nature HI coverage across both PHI and MHI schemes. Moreover, we have

examined the role of determinant factors on the scale up process of HI on

an equitable basis. And also investigation was made on the selection bias

that can be a constraint to scale up of HI. Moreover, we have found that

people are not aware of insurance in general and HI in particular, which

can be cited as one of the reasons for the low level of HI in India. Suppose

we consider a situation that people are willing to and able to pay for HI

and are completely aware of the importance of HI and also of the

prevailing schemes in the market, and there is no adverse selection

problem in PHI and MHI schemes, insurance agents are adequately

compensated to motivate them to sell more PHI schemes. As SHGs are

widespread in the rural and urban slums to reach the clients, can we

expect that HI will be scaled up? Equal to all the above necessary factors

for the scale up process of HI is that the schemes available in the market

must reflect the preference for various HI benefits by the people. The next

chapter addresses this issue.

CHAPTER SEVEN

PREFERENCES FOR THE HEALTH INSURANCE

BENEFITS AND THE HEALTH INSURANCE SCHEMES

7. I. Introduction

In the previous chapter, the discussion was confined to the issue of

universality in HI. and we dealt with the main factors detennining the

scaling up process of HI in PHI and MHI schemes, One of the main

assumptions dUring the discussion was that Individuals with insurance

co\'erage do not have to pay for medical expenses in case of falling sick. In

other words, the out of pocket health expenditure of the Insured

individuals are assumed to be at zero. In fact. it Is not always true that HI

schemes reimburse all Out of Pocket Spending (OOPS) of the insured. In

fact. a HI may perhaps cover either pari or full share of all medical

expenses of Insured people; It depends on the composition of the benefits

in HI package. An important factor that may attract people to opt for HI

resulting in the scale up process of HI is the comprehensiveness of HI

schemes. In fact. a v1able HI scheme can not cover all health care benefits

at a premlum that is affordable to the major sections of the Indian society.

DeSirable composition of benefit packages that individuals and

communities in India have expressed within a limited budget (Dror et al ..

2007). If the prevailing HI schemes do not refiect the preferences and

reqUirements of the clients. it becomes unattractive. and would lead to

low level of health insurance coverage. In this chapter, we examine the

meeting point between the preference of people for a desirable HI benefit

packages and the prevailing HI schemes.

7.2. Analytical Aspects

First of all. let us classifY the total health care expenditure into Outpatient

(OP) health expenditure and Inpatient (IP) health expenditure.

133

Both the OP and IP d . expen itures can be further classified as direct and indIrect health expenditures.

M = MD + MI

Where MD is 'Direct Health Expendit 'and MI Expenditure'. ure is 'Indirect Health

Further, MD (Direct Health expenditure) = All types

expenditure + Drugs Expenditure + Tests of Consultation

MI (Indirect Health expenditure) = Travel expenditure + Wage I oss.

Health Expenditure

I Inpatient Care

Direct Expenditure

(=Hospitalization + Drugs + Lab and image tests)

t

Indirect Expenditure

(=Transport costs + Compensation for the loss of income)

1 Outpatient Care

Direct Expenditure

(=Consultations/G P +Drugs + Lab and image tests)

Indirect Expenditure (=Transport costs + Compensation for the loss of income)

An insurance scheme can not cover all health care services mainly

because of (i) the ability to pay of the people and (iI) the problem of moral

hazard and adverse selection. As we have noted before, insurance is a

financial protection against the uncertain illness episodes. It has two

aspects: (1) buying health insurance, but not making any claim; (2)

buying HI and getting claim, either partial or full.

134

Not Sick No Claim

Insurance Enrolment

Sick Claim

Thus. utility of having a HI with full reimbursement will be higher than

that of the utility of having a HI with partial reimbursement. Nonetheless.

insurance coverage gives a kind of psychological security to Insured

people irrespective of whether he/she falls sick or not. Therefore, a person

who intends to buy HI may be expecting a scheme covering all health care

costs in case of falling sick or at least preferred HI benefit packages at a

reasonable premium. The absence of HI schemes reflecting the preference

of the individuals and communities can be cited as one reason for the low

level of HI in India. In this context. we test the [ollowing hypothesis: "Low

or no uptake of HI in India is explained by the fact that benefits on offers

differ significantly from what the client want."

We measure the preferences of the people for various health care benefits

that are classified into different categories under two situations: (1)

Without a Budget constraint, and (2) With a Budget constraint. The

preferences without budget constraint are measured from the responses

of those who are willing to pay for health insurance, for which we use the

ECCP data. And. the preferences with budget constraint are measured

with the help of a decision tool called Choosing Health plans All Together

(CHAT exercise). Thereafter. we analyze whether the preferences of the

people match with the various benefits on offer with the available HI

schemes in India.

7.3. Preferences of the People for Various Health Care Benefits without Budget Constraint

Although the proportion of population covered by health insurance

h . I any studies indicate that Indians are willing to join

sc emes IS very ow. m

(M tho hagan 1998' Gumber et al .. 2000; Gupta

and pay for HI a lyaz ' '

06) A Per the ECCP data. 70% of sample

Indrani. 2000, Dror 20 . s

135

households are willing to pay Rs. 100 per individual per year for a viable

rural HI schemes. People not only expressed their willingness to pay and

join, but also revealed their preferences for different types of benefits in

the HI package.

Let us consider a situation of people expressing their preferences for

different types of health care benefits in a HI scheme without budget

constraints. The total health care benefits are claSSified into 7 categories:

Hospitalizations. Primary Care/Consultation with a General Practitioner

(GPl, Drugs. Lab Tests. Maternity. Transportation Costs. and

Compensation for loss of income due to illness.

The following figure presents the percentage of households who would like

to have various HI benefits with the HI scheme for which they are willing

to pay (as reflected in the ECCP data).

Figure 7.1

100 ~ 90 ~ 80 C 70 ., '0 60 0 50 .c .. 40 ., :::I 30 0

20 ~

10 o·

Preferences for various health care benefits among those who are willing to pay for health insurance (N=2390)

91 88

77 79 75 69 59

-~ --,-

~ c: c:

'" ~Cl. '" 2 .9 0_ c Ol W '" row ~oQ) 0 ",(9 :0

'" <.> EQ:; ~ - t::_ "'WE

ro 0 .D '" o W cwo .- ~ . .", Q).2 (,) N ~ '" '" E 0..0 Cl. <.> -' w <.> o..~c

!'l '" c: E 0 .-iii ~ 0-

0.. ::;;: r- 0 '" 0 :r:

Health Care Benefits

Source: ECCP data

136