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 '\
I / '\ ",
: ',I BAI\CALORE: )':.
, . , .... '/
. . I J : ';/
\. , J /
\ -. \.. ' /
'-- " '-'.....,.,.". I
{-"-",.,. ,
" - [:flC -'
. , _ ".
_
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
6792
71
71
74
77
79
85
92
93119
91
94
98
98
102
104
106
109
113
liS
119
120132
120
120
130
131
133153
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 provider
2
. 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 India
4
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
earnings
6
As already noted. studies show that the cost of treatment in
India is very high in both pubJic
7
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 rating
8
. 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
restrictions
9
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 nsured
nsured
nsured nsured nsured nsured
pesUtute
38 62 63 38 52 48 52 48 44 56

Poor
42 58 72 28 54 46 71 29 52 48
Moderate
Poor
57 43 63 38 55 45 49 51 55 45
flulnerabl,

46 54 52 48 53 47 40 60 53 47

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
g r o u p ~ 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 to0
20

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: of the society it can be expected that the of
theshe sChhemes tar
l
g
) 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 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 -
l5
1h
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 4th 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
~ ' 4 6
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 schemes
21
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
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
________________________ --,
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 Non-
Member
.SHG
Member
Figure 4.7-b: The SHGs membership status of the Non Insured
households of the 'Rural MHI' schemes
90
80
70
60
"C
-0 50
! -5; 40
fIl
'" 30
20
10
o
Destitute Extreme
Poor
69
64
Moderate Vulnerable Non-Poor Wealthy
Poor
NonPoor
Income Class
Pearson Chi-Square: 23.89 (p<O.OOO)
Source: ECCP data
D SHG Non-
Member
.SHG
Member
81
Figure 4.7 -c: Proportion of Insured and Non
members in Rural MHI Schemes
Insured among the SHG
9 0 r - - ' ~ - - - - - - - - - - - - - - ______________ ~
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
9 0 ~ - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
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 NonPoor
Income Class
Pearson Chi-Square: 2.453 (p<O.783)
Source: ECCP data
oSHG Non-
Member
.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 10
th
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
f a m i l ~ 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

Low income household
with SHG
membership
= 1: Otherwise-O

Low
income
household
with SHG non-
membership -1; Otherwlse-O

Middle
income
household
with SHG
membership -1; Otherwise-O

Middle
income
household with SHG non-
membership = 1; Otherwise-O
Hlglnc _
High Income household with SHG membership
= I; Otherwlse=O
,
Higinc_
High income
household with SHG non-
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)-

-.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


.442 .141 .371 .658 .254 .460 .100
memb
116.69)' 12.11)" 15.69)' 117.59)' 15.15)' 111.26)' (1.37)
(Reference
category:

nonmembl

.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:

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)

(Reference
category:
Higinc_

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


to have HI coverage for the households of each
Income class with SHG membership as compared to the
households SHG memership of the same income class, by
various Income class across MHllocations
1000
..
. '
800
60,
400
20'0

200
-400
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 plan
22
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 Insu
rance 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 H
I products are featured by a pure risk component
s componen.
98
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 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 (%).
~ e a r d about Insured 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 ___ -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', IH
a
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 agents
24
,
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 r e a ~ o n 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 high
25
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 product2
6
, 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. s o m ~ 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
seriou
s 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 behavior
27
. 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

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
( 198
1) 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 ,
" 11. . -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 ,
Am
ong 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 I; Otherwlse-O
Medlum_edu Highest Educational qualification of the household is 5
th
to
10
th
years of schooling -1; Otherwlse=O
11th
Higher_edu Highest Educational qualification of the is
and aboveyears of schooling -I; Otherwlse-O
Headeducatlon
Highest education attained by the head of the family,
measured in terms of
Lowinc_highrisk
Low Income household with at least one of the
households is with 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 health risk -1; Otherwlse=O
househo h ehold with at least none of the member
Midinc lowrisk
Middle 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 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 Wlth 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 h
ouseholds 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-

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 seve
re 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
pop
ulation. 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 arn
e time ensure
adverse selection and ate s
HI co
verage 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 th
o hagan 1998' Gumber et al .. 2000; Gupta
and pay for HI a lyaz ' '
06) A
P
er 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:
'"

'"
2
.9
0_
c
Ol
W
'"
row

0
",(9
:0
'"
<.>
EQ:;

-
t::_ "'WE
ro
0
.D
'"
o W
cwo
.-
. .",
Q).2 (,)
N
'"
'"
E
0..0
Cl. <.>
-'
w <.>

!'l
'"
c: E 0 .-
iii
0-
0..
::;;: r- 0
'"
0
:r:
Health Care Benefits
Source: ECCP data
136