Successfully reported this slideshow.

Financing and delivery of health services ncmch


Published on

  • Be the first to comment

Financing and delivery of health services ncmch

  1. 1. Background Papers of the Background Papers of the National Commission on Macroeconomics and Health Financing and Delivery of Health Care Services in IndiaNational Commission on Background PapersMacroeconomics and Health Financing and Delivery of Health Care Services in India National Commission on Macroeconomics and Health MINISTRY OF HEALTH AND FAMILY WELFARE GOVERNMENT OF INDIA, 2005 EQUITABLE DEVELOPMENT • HEALTHY FUTURE
  2. 2. NCMH Background Papers Financing and Delivery of Health Care Services in IndiaNational Commission on Macroeconomics and Health Ministry of Health & Family Welfare Government of India, New Delhi August 2005
  3. 3. © Ministry of Health & Family Welfare, Government of IndiaSeptember 2005ISBN 81-7525-632-8This Report does not address tertiary care and related areas such as super speciality hospital development in the publicor private sector, telemedicine, medical tourism, environmental pollution or food safety etc. though they are all equallyimportant. The Commission Report is based on background papers which can be accessed from the NCMH They have also been published in two companion volumes. This report was written during theperiod April 1, 2004 - March 31, 2005.Printed at: Cirrus Graphics Private Limited B 261, Phase I, Naraina Industrial Area, New Delhi 110 028 Tel: + 91 11 51411507/1508 Fax: +91 11 51417575 email: cirrusgraphics@touchtelindia.netEditors: Pranay G. Lal and Byword Editorial ConsultantsCover design: Quote Design Studioii Financing and Delivery of Health Care Services in India
  4. 4. PrefaceIN PURSUANCE OF THE RECOMMENDATIONS MADE BY THE COMMISSION ON MACROECONOMICS ANDHealth, WHO, India established the National Commission on Macroeconomics and Health (NCMH) in March,2004. The main objective of the NCMH was to establish the centrality of health to development and make anevidence-based argument to increase investment in health. The Terms of Reference of the NMCH were mainlycentered on identifying a package of essential health interventions that ought to be made available to all citizensand also list systemic constraints that need to be addressed for ensuring universal access to this package ofservices. The NCMH was also to indicate the resources required and targets that ought to be achieved by 2015. The Terms of Reference of the NCMH were very widespread and spanned across a wide range of issues.Foraddressing each of the major concerns a broad outline of the approach to be adopted was prepared and sharedwith a large number of researchers, policy makers, experts from donor agencies and health activists. Based onthe suggestions received, topics to be addressed were identified and studies / papers commissioned. Every paperwas also peer reviewed by experts in that field. In all over 35 papers were commissioned. Due to limitations ontime and resources, original field surveys were limited to a hundred percent facility survey in eight districts ofKhammam(AP), Ujjain(MP), Varanasi(UP),Udaipur(Rajasthan), Kozhikode(Kerala), Jalna(Maharashtra), Nadia (West Bengal) and Vaishali (Bihar). For arriving at the estimates of public spending, we obtained informationfrom other government departments, PSUs, FIIs etc. and analyzed the data under the National Health AccountsFramework. Analysis of consumer surveys, the 57th. Round Survey National Sample Survey Organization onestablishments, and other data bases related to drug manufacture and sales, import and export of medical devicesetc. were also analyzed. Principal focus was on critically evaluating the current status of the health system - its organizational structure,financing mechanisms, regulatory frameworks etc. The three key drivers of health costs - namely human resources,drugs and technology were specially studied in detail as the main concern for the future is going to be the rapidescalation of costs. Such analysis highlighted and reiterated several shortcomings in the countrys health systemwhich are well known and have been recognized for long. Clearly, a well conceived and sequenced system ofreform emerged to be the priority area for policy attention so as to develop the capacity to absorb the promisedfunding of 2-3% of GDP in the next five years committed in the Common Minimum Program. What also emergedwere that solutions for many of the issues have been known for long, but routinely ignored and not acted upon.It was impossible not to conclude that if only timely attention to the large number of recommendations alreadyavailable had been accorded, the health system need not have been so inefficient, insensitive, dysfunctional andin such a crises as we find it today. Financing and Delivery of Health Care Services in India iii
  5. 5. The background papers formed the basis for the main report of the Commission and its recommendations.We have attempted to bring into the public domain all the data and analysis that has been carried out by theNCMH, both in printed form ( 2 volumes) as well as in the website of the NCMH - The mainpurpose has been to stimulate greater debate and research that would be useful for policy formulation. If thishas been achieved even in a small measure, we would be content that our efforts have been worthwhile. I wish to thank my colleagues at the Sub-Commission - Dr. Ajay Mahal, Dr. Avtar Dua, Dr. Sakthivel, Dr. SomilNagpal, Ms. Madhurima Nundy and Shri Sunil Nandraj and Dr. Rama Baru for their help and assistance. I alsothank all the contributors and reviewers for taking time off to write the paper or review it and helping us in everypossible way, very often at short notice. And finally a special thanks to Dr. Ranjit RoyChaudhury , member ofthe NCMH and chair of the sub-commission for his constant support, encouragement and advise. I am gratefulto each and every one of them. Sujatha Rao Secretary, NCMHiv Financing and Delivery of Health Care Services in India
  6. 6. List of Contributors and Reviewers AuthorsAJAY MAHAL M. GOVINDA RAOAsstt. Prof. DirectorHarvard School of Public Health National Institute of Public Finance & Policy,Boston, U.S.A. 18/2, Satsang Vihar Marg, Special Institutional Area, Near JNU,ANIL VARSHNEY New Delhi 110067Consultant90/2, Malaviya Nagar, MADHURIMA NUNDYOpp. Govt. Senior Secondary School, Research Scholar,New Delhi Centre for Social Medicine in Community Health, School of Social Sciences, JNUANUP K. KARAN New DelhiFellowInstitute for Human Development MARCELO TOKMANIAMR Building, 3rd Floor, Director, Economic PolicyI.P. Estate, New Delhi Ministry of Finance ChileASHOK D.B. VAIDYAMedical and Research Director MITA CHOUDHARYBhartiya Vidya Bhavans SPARC, Economist13th N.S. Road, J.V.P.D. Scheme, National Institute of Public Finance & Policy,Juhu, Mumbai 400049 18/2, Satsang Vihar Marg, Special Institutional Area, Near JNU,AVTAR SINGH DUA New Delhi 110067Asstt. Prof., Deptt. of PSMSMS Medical College, MUKESH ANANDJaipur Senior Economist National Institute of Public Finance & Policy,MS. CONSUELO ESPINOSA MARTY 18/2, Satsang Vihar Marg,Senior Health Economist and Advisor, Special Institutional Area, Near JNU,Health Care Reforms New Delhi 110067Ministry of Finance,Chile N. RAVICHANDRAN Assistant ProfessorDHIRENDRA KUMAR Indian Institute of Health Management & ResearchAssociate Professor 1, Prabhu Dayal Marg,Indian Institute of Health Management & Research Sanganer Airport,1, Prabhu Dayal Marg, Jaipur 302011Sanganer Airport,Jaipur 302011 N. VEERABHRAIAH Andhra Pradesh Vaidya Vidhan ParishadK. SUJATHA RAO Department of HealthPrincipal Secretary, Govt. of Andhra Pradesh,Government of Andhra Pradesh, HyderabadHyderabadAndhra Pradesh P. DURAISAMY ProfessorLALIT MOHAN NATH Department of EconometricsFormer Dean (AIIMS) University of MadrasE-21, Defence Colony Chepauk, Chennai - 600005New Delhi 110003 Financing and Delivery of Health Care Services in India v
  7. 7. SHIV CHANDRA MATHUR SOMIL NAGPALDirector WHO Consultant,State Institute of Health & Family Welfare TB DivisionJhalana Institutional Area, New DelhiSouth of DD KendraJaipur 302004 S. SELVARAJU Consultant,S.D. GUPTA BD-3 G, DDA Flats,Director Munirka, New DelhiIndian Institute of Health Management & Research1, Prabhu Dayal Marg, T. DILEEP KUMARSanganer Airport, Advisor (Nursing), Dte.GHS andJaipur 302011 President, Indian Nursing CouncilS. SAKTHIVEL Nirman Bhavan, New DelhiResearch Associate,Institute of Economic GrowthDelhi University Enclave,Delhi 110007vi Financing and Delivery of Health Care Services in India
  8. 8. ReviewersALAKA SINGH GIRISH CHATURVEDIWorld Health Organisation, Joint Secretary (Insurance)Geneva, Switzerland Ministry of Finance Jeevandeep Building, Parliament Street,ANURAG BHARGAVA New DelhiConsultantJan Swasthya Sahyog, GIRISH N. RAOVillage & Post: Ganiyar, Managing DirectorDistrict Bilaspur 495112 TTK Health Care Services Pvt. Ltd.,Madhya Pradesh #7, Jeevan Bhima Nagar, Main Road HAL III Stage,BARUN KANJILAL Bangalore-560075DeanIndian Institute of Health Management & Research, G.P. DUBEY1 Prabhu Dayal Marg, Sanganer Airport, ProfessorJaipur Department of Biofeedback, Institute of Medical Sciences,C.H.S. SASTRY Banaras Hindu University,Director(Retd.), National Institute of Ayurveda, Jaipur Varanasi, Uttar Pradesh3-599/4, Congress Office Road,Near Ayappa Temple, INDRANI GUPTAUndavalli, Tidapalli (Mandal) Institute of Economic GrowthDistt. Guntur, Andhra Pradesh University Enclave, Delhi - 110 007CHARU GARGWorld Health Organisation J.V. MEENAKSHIGeneva, Switzerland IFPRI, WashingtonD. NARAYANAProfessor, Department of Economics, JAYAPRAKASH MULIYILCentre for Development Studies, PrincipalThiruvananthapuram Christian Medical College, VelloreDARSHAN SHANKARDirector K.S. RAGHAVANFoundation for Revitalisation of local Health Traditions Consultant74/2, Jarakabanda Kaval 102, Jyothi Manor, Plot No.41, Srinagar Colony,P.O. Attur, Via Velahanka Hyderabad 500073Bangalore- 560064 M.S. VALIATHANDINESH AGARWAL Honorary Advisor,Technical Advisor, Manipal Academy of Higher Education,UNFPA, 53, Jor Bagh Manipal 576104New Delhi MIRA SHIVAD.K. SRINIVAS Senior ConsultantRajiv Gandhi University of Health Sciences, Voluntary Health Association of India4-T Block, Jayanagar, B-40, Qutub Institutional AreaBangalore (Karnataka) New Delhi- 110 016GANGA MURTHY N.K. SETHIAdditional Economic Advisor DirectorMinistry of Health & Family Welfare National Institute of Health & Family Welfare,Nirman Bhavan, New Delhi Munirka, New Delhi Financing and Delivery of Health Care Services in India VII
  9. 9. NARENDRA BHATT RAMA BARUVice President Centre for Social Medicine & Community HealthIndian Association for the Study of Traditional Asian School of Social Sciences,Medicine Jawaharlal Nehru University,15 - Bachubhai Bldg. New Delhi 1100067J. Bhatnagar Marg, ParelMumbai- 400 012 SEETA PRABHU United Nations Development ProgrammePRAKIM SUCHAXAYA Lodhi EstateFaculty of Nursing New DelhiChiang Mai University,Chiang Mai, SRINIVASAN RThailand Former Secretary (Health) D-402, Kaveri Apartments,RAVI NARAYAN Alaknanda, New DelhiGlobal SecretariatC/o Community Health Cell S. SRINIVASANNo.359 (Old No.367) LOCOST, 1st Floor,Srinivas Nilaya, Jakkssadlu, First Main, Premanand Sahitya Sabha Hall,1st Block, Kormangala, Opp. Lakadi Pool, Dandiya Bazar,Bangalore 560002 Baroda 390001RAMESHWAR SHARMA SUNIL NANDRAJConstultant National Professional Officer,B-32, Vijay Path Health Systems Developments,Tilak Nagar, WHO, Nirman Bhavan,Jaipur-302004 New DelhiR.D. BANSAL V.N. PANDITConsultant Sri Sathyasai Institute for Higher Learning,Kothi No.3059 Prasantinilayam,Sector 19 D Distt. Ananthapur,Chandigarh-19 Andhra Pradesh 515134RAVI DUGGAL VAIDYANATHAN A.Coordinator, CEHAT Madras Institute of Development StudiesAram Society Road, 79, Second Main Road,Vakola, Santacruz(E) Gandhinagar, Adyar,Mumbai -400055 Chennai 600020R.L. MISHRA WILAWAM SEMARATAMAFormer Secretary Health Assistant ProfessorNo.4403, Qutub Enclave, Chiang Mai UniversityDLF Phase IV, Chiang Mai,Gurgaon 122002 ThanilandRAMESH BHATProf. of FinanceIndian Institute of ManagementVastrapur,Ahmedabad- 380 015VIII Financing and Delivery of Health Care Services in India
  10. 10. ContentsPreface iiiList of Contributors and Reviewers vSECTION I: Health, Poverty and Economic Growth in India 1 Health, Poverty and Economic Growth in India 3 Health, nutrition and poverty: Linking nutrition to consumer expenditures 19SECTION II: Delivery of Health Care Services in India 37 Primary Health Care in India: Review of Policy, Plan and Committee Reports 39 Delivery of health services in the public sector 43 Training for effective delivery of health services 65 Effective Integration of Indian Systems of Medicine in Health Care Delivery: 77 Peoples Participation, Access and Choice in a Pluralistic Democracy Delivery of health services in the private sector 89 The not-for-profit sector in medical care 125 People’s Partnership for Health Towards a Healthy Public in India 135SECTION III: Drivers of Health Care Costs 151 Human Resources for Health 153 Nursing for the delivery of essential health interventions 175 Access to Essential Drugs and Medicine 185 Appropriate Policies for Medical Device Technology: The Case of India 213 Annexure 1: Medical equipment use pattern in the public and private 226 sectors in India: Policy implicationsSECTION IV: Financing of Health Care in India 237 Financing of Health in India 239 Annexure 1: National Health Accounts for India 256 User charges in India’s health sector: An assessment 265 Health insurance in India 275 Resource Devolution from the Centre to States: Enhancing the Revenue 297 Capacity of States for Implementation of Essential Health Interventions Financing and Delivery of Health Care Services in India vii
  11. 11. SECTION IHealth, Poverty and Economic Growth in India
  12. 12. SECTION I Health, Poverty and Economic Growth in India T HE IMPORTANCE OF ECONOMIC GROWTH, MEASURED BY INCREASES IN GROSS domestic product (GDP) and GDP per capita, for policy purposes can hardly be overem- phasized. Economic growth is commonly used as an indicator of a nation’s economic performance, and the level of GDP per capita is a key component of the Human Development Index of the United Nations Development Programme, a popular indi- cator of national well-being. The benefits of economic growth are so pervasive that it has been a central agenda everywhere and countries have accorded top priority to achieving high rates of growth. Some experts and policy-makers have also argued that it is difficult to achieve declines in poverty rates by relying on redistribution strate- gies alone, without a concomitant improvement in size of the national economic cake, as reflected in the magnitude of real GDP and real GDP per capita. It is difficult to imagine a sustained decline in poverty unaccompanied by a simultaneous improve- ment in aggregate economic performance. There is now a large body of theoretical and empirical research on the determinants of economic growth. Much of the early work highlighted growth in labour and the stock of physical capital as the key determinants of economic growth. However, early empirical work was unable to ‘explain’ a significant portion of the growth in GDP and GDP per capita, by the growth in labour force and capital alone, and so atten- tion turned to other factors-most notably technological change embodied in capi- tal goods, and on the quality and quantity of labour, referred to as human capital, in promoting economic growth. Two key elements of human capital are the extent to which the labour force is educated, and the level of its health. Recent empirical work has sought to assess the association between human capital and aggregate eco- nomic performance and found that, given labour and capital, improvement in health status and education of the population lead to a higher output (Barro and Sala-i- Martin 2004). The role of health in influencing economic outcomes has been well understood at the micro level. Healthier workers are likely to be able to work longer, be generally more productive than their relatively less healthy counterparts, and consequently able P. DURAISAMY to secure higher earnings than the latter, all else being the same; illness and disease DEPARTMENT OF shorten the working lives of people, thereby reducing their lifetime earnings. Better ECONOMETRICS health also has a positive effect on the learning abilities of children, and leads to bet- UNIVERSITY OF MADRAS ter educational outcomes (school completion rates, higher mean years of schooling, CHENNAI 600005, INDIA achievements) and increases the efficiency of human capital formation by individu- E-MAIL: als and households (Strauss and Thomas 1998; Schultz 1999). However, more recent research has also established a strong causal association run- ning from health to aggregate economic performance. Thus Bloom, Canning and Sevilla (2004) report evidence from more than a dozen cross-country studies and all these studies, with a single exception, show that health has a positive and statisti- AJAY MAHAL cally significant effect on the rate of growth of GDP per capita. The causal relation- HARVARD SCHOOL ship does not run in only one direction-from health to aggregate economic per- OF PUBLIC HEALTH formance-and there is strong case for considering a reverse link, running fromDEPARTMENT OF POPULATION ‘wealth to health’. Higher incomes potentially permit individuals (and societies) toAND INTERNATIONAL HEALTH afford better nutrition, better health care and, presumably, achieve better health. There BOSTON MA 02115, USA is some cross-country evidence that such a relationship holds at the national level E-MAIL: (Pritchett and Summers 1996; Bhargava et al. 2001). Several experts believe, how- ever, that the causal direction from health to economic performance is stronger. The previous empirical findings have implications for the role of health improve- Financing and Delivery of Health Care Services in India 3
  13. 13. SECTION I Health, Poverty and Economic Growth in Indiaments among workers in influencing another key policy objec- these differentials. Unfortunately, there have been only a fewtive-poverty reduction. First, to the extent that improvements recent attempts at examining the relative growth perform-in health result in improvements in national income, poverty ances of the States in India (Ahluwalia 2001; Sachs et al.could decline on account of both the standard ‘trickle-down’ 2002), and most of the major studies do not emphasize theeffects and an increased financial capacity of nations to set role of health in influencing economic performance. The onlyup safety nets. There is a good deal of evidence suggesting study that sought to do this in the Indian context was onethat countries that experience a steep rise in growth rates of by Gupta and Mitra (2003), which examined the link betweenreal GDP per capita also experience impressive declines in growth, health and poverty in India. While useful, the chiefpoverty (Barro and Sala-i-Martin 2004). Second, improve- drawback of this paper is that its empirical specification wasments in health, when directed at the poor, can contribute essentially ad hoc, and not influenced by developments inmore directly to poverty reduction and serve as an element the economic growth literature. As a consequence, there areof a ‘pro-poor’ growth strategy. The poor bear a dispropor- legitimate concerns with their model specifications, includ-tionately higher burden of illness, injury and disease than ing the criteria used for the inclusion (or exclusion) of explana-the rich. The poor suffer ill health due to a variety of causes, tory variables. There are now newer and more powerful meth-poor nutrition for instance, which reduces the ability to work ods to assess the links between health, poverty and economicand weakens their resistance to disease. With their body growth. For these reasons, we believe that the estimatesoften being their main income-earning asset, sickness and reported in their paper are unlikely to be robust.disability have significant adverse implications in terms of lossof work and incomes, compounded by their inability to obtain Economic growth and health: A review ofadequate health care. Frequently, treatment expenditure cross-country and regional studiesand loss of earnings force poor families to exhaust their sav-ings and assets, and take recourse to borrowing, leading to Modern growth literature includes, in addition to the stan-more poverty and poor health status. dard labour and capital variables, indicators of human capi- This paper contributes to existing analyses of the health- tal-the stock of education and health-among the determi-poverty-income nexus by examining these relationships at the nants. Particularly, the influential works in this area are theState level in India, using the most recent empirical methods cross-country studies by Barro (1991, 1997) and Barro andavailable in the literature (Bloom et al. 2004). Our analysis is Sala-i-Martin (2004) and the theoretical framework devel-carried out using a cross-State panel dataset for 14 major oped by Mankiw, Romer and Weil (1992). A comprehensiveIndian States for the years 1970/71, 1980/81, 1990/91 and review of empirical evidence on the new macroeconomics of2000/01, spanning a thirty-year period. growth is contained in Temple (1999). Barro (1991) used a The analysis of this paper is important for several reasons. cross-sectional framework and the human capital variableFirst, there is no denying the policy significance of under- was restricted to school enrolment rates at the primary andstanding the determinants of economic growth and its rela- secondary levels. He showed, using cross-section data for 98tionship with poverty and improvements in health. If health countries, that the growth rate of real GDP per capita overturns out to have significantly influenced India’s economic the period 1960-85 was positively related to the initial (1960)performance, this may call for investing more public funds enrolment rate, and inversely related to the starting (1960)in health, given that health budgets have been severely level of real per capita GDP. In subsequent analyses, Barroresource-constrained in recent years. One way this could hap- (1997) and Barro and Sala-i-Martin (2004) used a panel datasetpen is by greater emphasis on the commitments India has of countries, and included health as a determinant (lifemade to meet the targets set by the Millennium Declaration. expectancy at birth [LEB]) besides years of educational attain-These targets include significant improvement in health ment and other factors that could potentially influence thethrough reduction in infant and child mortality by two-thirds growth of real income per capita. Their results indicate thatby 2015 (World Bank 2004). Conversely, this also calls for the log of LEB has a positive and statistically significant effectunderstanding better the impact of economic growth on on growth rate with a coefficient of 0.042, which implies anhealth, so that one can assess the improvements in economic annual rate of increase of per capita real GDP of 4.2%. Fogelperformance necessary to achieve the desired goals. Sec- (1994) showed that about one-third of the increase in incomeond, unlike the existing literature which relies on cross- in Britain during the nineteenth and twentieth centuries couldnational data, our paper examines the interlinkages between be attributed to improvements in health and nutrition. Mayerhealth and economic performance within a single country. (2001) concluded that improvements in adult survival wereIntracountry analysis has the advantage of being much bet- causally linked to improvements in growth performance inter equipped to handle data-comparability issues relating Brazil and Mexico; and Weil (2001) found that health (indi-to health, education and economic performance. At the same cated by average height and LEB) explained about 17% of thetime, the significant variation in inter-State performance in variation in income per capita across countries. Gyimah-Brem-health and economic achievement means that our estima- pong and Wilson (2004) find that 22% and 30% of the growthtion procedures yield estimates that are reasonably robust. rate of per capita income in sub-Saharan Africa and OECDThird, significant inter-State differences in India’s eco- countries, respectively, can be attributed to health.nomic performance call for enhanced efforts in understanding Bloom, Canning and Sevilla (2004) review several studies4 Financing and Delivery of Health Care Services in India
  14. 14. Health, Poverty and Economic Growth in India SECTION Ithat include health as an explanatory variable in growth equa- aggregate GDP growth to at least 8%-10% and sustain it attions, in addition to presenting new results, based on a cross- that level for a sufficiently long period. Bourguignon (2004)national panel dataset for countries. They use a production examines theoretically the interrelationship between growth,function model of economic growth with a measure for human inequality and poverty, and shows that both growth and changescapital which takes account of the indicators of health, edu- in inequality contribute to changes in poverty. However, thecation and labour market experience. There are two note- relative effects of these phenomena may be country-specificworthy findings from their analysis. First, their analysis rec- and depend on initial income level and inequality.onciles microeconomic analyses of the rate of return to school- It was noted earlier that health improvements contributeing with macroeconomic analyses of returns to education. to income improvements or growth. With much evidenceSecond, they report a positive and statistically significant also pointing to the growth-poverty reduction nexus, bettereffect of health on economic growth. Their empirical find- health can be seen as a factor that contributes to povertyings reveal that an increase of one year in LEB raises the growth reduction via some form of trickle-down mechanism. Whenrate of GDP by 4%. Bhargava et al. (2001) found that the adult health improvements are concentrated among people livingsurvival rate (ASR) has a positive effect on growth rate of per close to, or below the poverty line, both a trickle-down mech-capita GDP and that a 1% increase in ASR increases the growth anism and a redistributive one work to reduce poverty. Roughrate by 0.05% for the poorest countries. computations by the World Bank, using National Sample Sur- While there is compelling evidence that health contributes vey (NSS) data, suggest that ill-health and associated eco-significantly to economic growth, there is also voluminous lit- nomic losses cause as much as 22 lakh Indians, most livingerature that focuses on causality in the reverse direction-from marginally above a poverty line standard of living, to tem-income to health. Much of this work is based on micro-level porarily fall below the poverty line each year, owing to adata that focus on the impact of income on the health status combination of income losses on account of being unable toof households and their members (Behrman and Deolalikar work and declines in non-medical care consumption. The NSS1988; Strauss and Thomas 1998). There has also been some for India for 1995-96 also reveal that when the poor fall sick,recent work at the macro-level, using cross-national panel they are often unable to afford treatment, and even whendatasets; and much of the current work using cross-country they do decide to get treated, tend to sell off productive assetstime series data has tended to account for reverse causality and rely on borrowing, all of which have the potential ofand inter-dependence between health, income and economic decreasing their long-run earning capacity-and the capacitygrowth. Thus, Pritchett and Summers (1996) estimate the effect to take advantage of any trickle-down labour market advan-of income on health, measured by infant and child mortality tages offered by a growing well as life expectancy. Some authors have also inquired There are several studies in India on health status and health-into the distributional aspects of the income-health relation- seeking behaviour. In an early attempt Kannan et al. (1991)ship. For instance, Preston (1975) used cross-country evi- analysed the linkages between health, development and socio-dence to suggest that the effect of income improvements on economic factors in Kerala. Vaidyanathan (1995) examinedhealth was greater for the poorest countries than for the rich- the measurement issues related to nutritional and healthest countries. Deaton (2001) argued that income inequality is status and the adequacy of currently available data for assess-not a major determinant of health of the population. ing nutrition-health status. A number of studies examined How about the relationship between health, income and levels and changes in morbidity and health expenditurepoverty? In a purely accounting sense, increases in real GDP using the National Sample Survey 1986-87 health survey dataper capita will be accompanied by simultaneous declines in (Visaria and Gumber 1994, Krishnan 1995, Duraisamy 1995,the number of people living in poverty, provided the distribu- and Gumber 1997). Sundar (1995) studied the levels andtion of income remains more or less constant. Growth may be changes in health status and health expenditure based onessential to reducing poverty and one might presume that poli- NCAER survey. These studies are mainly descriptive and refercies promoting distributional improvements will prove diffi- to earlier periods.cult to sustain in the absence of long-term increases in real The relationship between income, health and productivityGDP per capita-that is, economic growth. Empirically, Barro has been analysed at the household level based on micro-and Sala-i-Martin (2004) demonstrated that regions of the econometric framework. Duraisamy (1998, 2001) foundworld that experienced higher growth rates also witnessed evidence of a strong negative effect of income or total con-steeper declines in poverty. Bourgoignon (2004) cites studies sumption expenditure on morbidity and household assetsthat provide evidence on the poverty-reducing impact of growth emerged as an important determinant of child survival andgiven that income distribution remains the same, and of increases preventive health care (Duraisamy and Duraisamy 1995). Deo-in poverty with a worsening of income distribution. In India, lalikar (1988) demonstrated that health was a significantpoverty levels have declined the fastest over periods that determinant of labour productivity using farm level data. Aexperienced the highest growth rates, during the1990s study on health, wages and labour supply by Duraisamy and(Ahluwalia 2001). According to Srinivasan (2003), there was Sathiyavan (1998) revealed that a 10% increase in the bodyno perceptible decline in poverty in India until growth accel- mass index of males and females increased their wage rateerated in 1980s and hence a necessary condition for eradi- by 7% and 2% respectively and labour supply by 20% andcating mass poverty is to accelerate average annual rate of 11% respectively. Financing and Delivery of Health Care Services in India 5
  15. 15. SECTION I Health, Poverty and Economic Growth in India At the macro level, very little is known on the association variables for the study period. The States included for thebetween income/economic growth and health (Gupta and study account for 90% of India’s population and 83% of theMitra 2003, World Bank 2004). Gupta and Mitra (2003) exam- country’s total land area at present.ined the relationship between health, poverty and economic State-level income and per capita income are representedgrowth in India for the years 1973/74, 1977/78, 1983, 1987/88, by the respective State’s NSDP and the per capita NSDP1993/94, 1999/2000 based on data for 15 Indian States. Their (PCNSDP). Data on the NSDP and PCNSDP are produced oneconometric analysis showed that per capita public health a regular basis by the Central Statistical Organisation (CSO)expenditure positively influences health status, that poverty of the Government of India. We obtained these data fromdeclines with better health, and that growth and health have publications of the EPW Research Foundation (2002a, 2003)a positive two-way relationship. Despite reporting what appear and CSO (2004). The value of NSDP and PCNSDP in these isto be significant findings, this study suffers from certain reported in current prices and this has been converted intomethodological drawbacks as indicated earlier. Identifica- constant price series using a GDP deflator.tion restrictions in the model specification appear to be arbi- The poverty variable is the head count measure, i.e. thetrary rather than based on economic theory, or empirical lit- proportion of the population living below the poverty line.erature. Their empirical specification with growth of net State In India, the poverty line is defined as the minimum expen-domestic product (NSDP) as the dependent variable uses NSDP diture required for achieving a basic calorie requirement,(not per capita NSDP) in the base year as an explanatory plus comparable non-food consumption expenditures. Thevariable, a procedure not used previously in the literature, and source of poverty data for this paper is the Planning Com-for which no justification is provided. The same specification mission, which computed poverty levels from the Nationalomitted population as an explanatory variable, an assump- Sample Survey Organization (NSSO) ‘consumer expenditure’tion which appears not to be standard (Bloom and Freeman surveys using the ‘expert group methodology’. Poverty data1986; Bloom and Williamson 1998). Many of the estimated are available for the years 1972/73, 1983, 1993/94 andcoefficients in their analysis turned out not to be statistically 1999/2000, respectively, and for the purposes of our statis-significant. For example, in the growth equation (growth of tical analysis, are taken to correspond to the years 1971, 1981,NSDP), poverty, infant mortality rate, initial NSDP and liter- 1991 and 2001.acy are statistically not significant even at the 10% level of The health status of the population is captured through twosignificance. In the same equation, the infrastructure (INF) indicators-LEB and the infant mortality rate (IMR). Data onvariable has a significant negative effect on growth rate. these two health indicators were obtained from the Registrar In a World Bank (2004) study, the effects of per capita General of India (1999) and updated for recent years usingGDP, per capita health expenditure and female literacy on the Sample Registration System (SRS) Bulletin published byinfant mortality were examined using State-level data for the Registrar General of India. Data for 1971 for Bihar andthe period 1980-99 based on econometric framework. The West Bengal were extrapolated using the time series data ofresults show that both per capita public spending on health the concerned States. LEB estimates for 1961 were taken fromand per capita GDP are inversely related to IMR, but they are the estimates published by the Registrar General of India,not very robust to alternative specifications of the model. which is based on the population census of that year.However this study does not examine the effect of per capita Apart from health, human capital is measured along twoincome on LEB, an alternative and perhaps better measure additional dimensions-average years of schooling and workof health status of the population. experience. First, we computed years of schooling using cen- The lack of consistent findings in the literature, and possi- sus data on completed levels of education by age and sex ofbly specification problems in the early works, lend further jus- the population. The completed years of education for vari-tification to the empirical analysis that we pursue in this paper. ous levels are assumed to be as follows: literate below pri- mary-4 years; primary-5 years; middle-8 years; secondary-Database 10 years; higher secondary/pre-university-12 years; techni- cal and non-technical diploma-13 years; graduate and above-To empirically examine the linkages between health, poverty 16 years. The variable for average years of schooling is con-and economic growth at the sub-national (State) level in India, structed from the census tables on completed levels of edu-we constructed a panel dataset of 14 States, including obser- cation by age and sex of the population given in the Socialvations every ten years-1970/71, 1980/81, 1990/91 and and Cultural Tables, Census of India, published by the Reg-2000/01. This study is confined to the major Indian States istrar General of India for various census years, weighted byfor which consistent time series data are available. The States its appropriate population share. Second, following Bloom,excluded from the study are: Jammu and Kashmir, Goa and Canning and Sevilla (2004), the years of labour marketHimachal Pradesh, eight north-eastern States, and seven Union experience is constructed using the age and gender distri-Territories. In the year 2000, three of the States included in bution of workers provided in the General Economic Tables,our sample, Bihar, Madhya Pradesh and Uttar Pradesh, were Census of India, Registrar General of India for various years.bifurcated. We have merged the data on the new States (Chat- The ‘years of experience’ is defined as age minus years oftisgarh, Jharkhand and Uttaranchal) with their respective par- schooling minus six, the age of entry into schools as used inent States and constructed a comparable series of all the the micro-studies in labour economics. The average work6 Financing and Delivery of Health Care Services in India
  16. 16. Health, Poverty and Economic Growth in India SECTION Iexperience is the weighted average of the age- and sex-group government expenditures on health, water supply and san-specific potential experience with the respective group’s share itation, and family welfare were compiled from the RBI Bul-of the total. The data on the number of workers for various letin for various years. We also constructed two variables tocensus years were obtained from the General Economic Tables represent political power: (i) the percentage of votes gainedof the population census (Registrar General of India [various by the ruling party at the Centre in the Assembly elections;years]). Total workers include both main and marginal work- and (ii) the percentage of votes secured by socialist anders. The total population in the working age groups of 15- communist parties in the respective State Assembly elections.59 years were also collected from the decennial population The data for these variables were gathered from the Elec-census for the respective years. As the age distribution of tion Commission.the population for 2001 was not available when this work Using the above data we first present a descriptive analysiswas completed, projected instead of actual population by to understand the association between some of the variablesage groups was used. used in the study. This is followed by the specification of the Physical capital is another key explanatory variable in analy- econometric model and discussion of the of economic growth. Unfortunately, data on gross cap-ital formation or the level of investment at the State level Health, poverty and economic growth:comparable with the national-level data on physical capital Inter-State descriptive analysisfrom national accounts statistics are not available. Data ongross fixed capital formation (GFCF) is available only for a The basic socioeconomic characteristics of the 14 States andfew States from 1993-94 onwards. However, data on the for all of India are given in Table 1. Clearly, there is largevalue of fixed capital for the industrial sector are available inter-State variation in the level of PCNSDP for the most recentfrom the Annual Survey of Industries (ASI) published by the year (2000/01). The richest State is Punjab, with a per capitaCSO and compiled and published by the EPW Research Foun- income of Rs 15,390; with Bihar being the State with the low-dation (2002b). These values were expressed in current prices est income per capita of Rs 4123.and have been converted into a constant price series using The estimated growth rates of real per capita income overthe GDP deflator. the thirty-year period 1970-2000, also shown in Table 1, reveal Public expenditure on health is an important determinant similar trends. The range of variation in growth rates is fromof the health status of the population. State-level a low of about 0.9% and 1% respectively in Madhya Pradesh Table 1 Basic characteristics of the States included in the study State PCNSDP Annual average Life expectancy IMR 2000 Poverty Population Annual average 2000/01 rate of real at birth 1995-99 (per 1000 1999-2000 2001 rate of population (Rs)Life PCNSDP growth (years) live-births) (% below (in thousands) growth 1971-2001 1970-2000 (%) poverty line) Andhra Pradesh 9,982 2.6 63.1 55 15.77 75,728 1.8 Bihar 4,123 1.0 60.2 62 42.6 82,879 1.3 Gujarat 12,975 3.6 62.8 62 14.07 50,597 2.1 Haryana 14,331 2.8 61.5 67 8.74 21,083 2.5 Karnataka 11,910 3.5 64.0 57 20.04 52,734 2.0 Kerala 10,627 1.9 73.5 14 12.72 31,839 1.3 Madhya Pradesh 7,620 0.9 56.4 87 37.43 60,385 1.2 Maharashtra 15,172 3.8 65.8 48 25.02 96,752 2.2 Orissa 5,187 1.7 57.7 95 47.15 36,707 1.7 Punjab 15,390 3.1 68.1 52 6.16 24,289 1.9 Rajasthan 7,937 2.5 60.5 79 15.28 56,473 2.6 Tamil Nadu 12,779 3.5 64.6 51 21.12 62,111 1.4 Uttar Pradesh 5,770 1.2 58.4 83 31.13 166,053 2.1 West Bengal 9,778 2.8 63.4 51 27.02 80,221 2.0 India 10,376 2.4 61.7 68 26.1 1,027,015 2.1 Note: 1. Data for Bihar, Madhya Pradesh and Uttar Pradesh include the three newly formed States of Jharkand, Chhattisgarh and Uttaranchal, respectively. 2. The data for India includes all States and Union Territories. 3. Per capita income (PCY) refers to real per capita NSDP . Sources: PCNSDP from EPW Research Foundation (2003), growth rate in PCNSDP is based on the authors computation, LEB and IMR are from the Sample Registration System Bulletin (2004) published by the Registrar General of India, poverty estimates are from Planning Commission (from, population for 2001 is from the Registrar General of India, GOI and the growth rate in population is computed by the authors. Financing and Delivery of Health Care Services in India 7
  17. 17. SECTION I Health, Poverty and Economic Growth in Indiaand Bihar to a high of 3.8% in Maharashtra. the graph, and the declining slope of the curve indicates that The relationship between initial real per capita income the effect of LEB increases faster at lower than at higher income(1970/71) and annual average rates of growth of real income levels. The relationship is similar to the cross-country evidenceper capita is indicated in Fig. 1. In general, States with lowinitial incomes also witnessed low growth rates except Andhra Fig 2Pradesh, West Bengal and Karnataka. Conversely, States withhigher starting incomes experienced higher growth rates, with Relationship between initial incomethe notable exceptions of Kerala and Madhya Pradesh. and growth rate Next, we examine the relationship between economic growth NSDP: net State domestic productand initial per capita income pooling the data for the threeperiods, 1970-80, 1980-90 and 1990-2000. The computedgrowth rate is the decadal rate for the periods and the initialincome corresponds to the beginning year of the respectivedecade. The scatter plot with a trend line is exhibited in Fig.2. It is amply evident that there is a positive association between Fig 1 Per capita income and growth rate by States, 1970-2000 NSDP: net State domestic product Fig 3 Trends in life expectancy at birth (LEB), 1970-99, Indiainitial income and growth rate. At first glance, this is at vari-ance with the cross-country results and the regional evi-dence reported in Barro and Sala-i-Martin (2004). However,the simple association of Fig. 2 does not control for con-founding factors such as human capital stock, and addi-tional analyses are called for to reach firmer conclusions.This issue will be explored further later in the paper. Figure 3 presents all-India trends in life expectancy atbirth (LEB) during the period 1970-2000. It is immediatelyapparent that India experienced a remarkable improvement on the association between LEB and per capita income (inin LEB over this period, from 49.7 years during 1970/75 to 1985 purchasing power parity (PPP) in dollars) shown by61.7 years during 1995/99. The inter-State disparity in LEB Pritchett and Summers (1996) as well as in Preston (1975).in 1995/99 is laid out in Table 1. LEB is highest in Kerala The positive association between LEB and per capita income(73.5 years) and lowest in Madhya Pradesh (56.4 years), imply- could be due to (i) increased income causing better health;ing a difference of 18.1 years. Bihar, which is one of the or (ii) healthier workers being more productive and hence hav-States with the lowest per capita income, seems to have ing higher incomes; or (iii) a common factor that leads to bothfared better than Madhya Pradesh, Orissa and Uttar Pradesh better health and higher incomes. Thus, the simple associa-in this health status indicator. tion between LEB and per capita income cannot tell us exactly It is instructive to compare the simple association between what the nature of the relationship is. This issue is furtherLEB and per capita income, pooling the three-period data, examined below using multivariate shown in the scatter plot (Fig. 4). The positive association The inter-State variation in the second health status indi-between income and life expectancy is vividly brought out in cator-IMR-is seen in Table 1. Kerala again stands out with8 Financing and Delivery of Health Care Services in India
  18. 18. Health, Poverty and Economic Growth in India SECTION Ithe lowest IMR of 14 per 1000 live-births, compared to 95 in Our discussion on the correlations that exist between indica-Orissa, which has the highest IMR. Interestingly, the second tors of economic growth-income and health-cannot be inter-lowest IMR is 48 in Maharashtra, nearly three times higher preted as a cause-effect relationship, including also the pos-than Kerala’s IMR. This clearly shows that even States with sibility of two-way causality among the above-mentionedbetter health status than all of India have a long way to go variables. We now develop an econometric framework to examine the causal or simultaneity relationships among these three Fig 4 variables. Life expectancy at birth (LEB) and per capita NSDP, 1970-2000 Model specification and estimation issues NSDP: net State domestic product Following the cross-country empirical studies on the deter- minants of economic growth (Barro 1991; Barro and Sala-i- Martin 2004; Bloom and Canning 2004, the growth rate of real per capita income function can be specified as (1) Git = α0 + α1lnYit + α2lnHit + α3 Sit + α4 lnWit + a5GWit + uit, i=1,2,..,N States, t=1,2,..T periods Here Git =1/m[lnYit+1 - lnYit] is the growth in per capita real income over the period t and t+1, lnY is the natural log- arithm of initial real per capita income, lnH is the logarithm of health indicator, namely LEB, S is another dimension of human capital, namely average years of schooling of the adult population, W is the ratio of working age to total popula- tion, GW is the rate of growth in W, m is the length of t -to ‘catch up’ with Kerala. (t+1), αi are parameters to be estimated and uit is the ran- The association between initial per capita income and dom disturbance term distributed with zero mean and con-IMR, pooling the three-period data is shown in Fig. 5. Per stant variance (see Bloom and Canning 2004 for the theo-capita income and IMR are negatively related. The decline in retical derivation of the model). Several other variables to cap-IMR as income increases is not uniform across all income ture the economic geography and quality of governancelevels. The decline is higher at the low-income levels and lower such as openness, institutional quality, ethnolinguistic frac-at high-income levels. tionalization, landlocked, tropical area, average government savings rates, access to ports, government consumption ratio,Econometric model and empirical analysis rule of law, etc. were included in the cross-country analysis (Barro 1997); Bloom and Williamson 1998). However, some of these variables are not relevant for a study such as this (e.g. openness) and data on many of the variables such as gov- Fig 5 ernance, investment or savings ratio, rule of law, etc. were not available at the State level. Religious and caste compo- Infant mortality rate (IMR) and sition (percentage of the population belonging to various reli- NSDP real per capita gions, and schedule caste and schedule tribes), urbanization NSDP: net State domestic product and population density were considered. Due to high corre- lation between these and other variables, particularly LEB and schooling, these were not included in the final analysis. In addition to average years of schooling, we also tried includ- ing years of labour market experience but due to high collinear- ity between schooling and labour market experience, the expe- rience variable turned out to be statistically insignificant and hence was dropped in the final analysis. The coefficient of the initial income variable Yit is an indica- tor of whether there is a conditional convergence in income per capita or not among countries or regions (States) within a coun- try. The conditional convergence hypothesizes a negative sign of the initial income coefficient. A positive sign would imply increased income dispersion among rich and poor countries (States). A problem with the initial income per capita is that this Financing and Delivery of Health Care Services in India 9
  19. 19. SECTION I Health, Poverty and Economic Growth in Indiavariable is potentially endogenous and also measured with error. in output is regressed on changes in inputs. Let the aggre-The procedure adopted in the growth literature is to predict the gate production function be of Cobb-Douglas form:per capita income using lagged values and the predicted values α β γ δ λare used to compute the growth rate as well as for initial income (3) Yit = AitK it L itH itS itE it(Barro 1997). We have also adopted this procedure. where Y is aggregate output, A is a technology parameter, Pritchett and Summers (1996), Bhargava et al. (2001) and K is physical capital stock, L is labour force, H is health (lifeothers argue that health cannot be treated as an exogenous expectancy), S is mean years of schooling, E is an experiencedeterminant of growth. That is, increased income leads to more vector (experience and experience squared) and , , , , andinvestment in health and thus there is strong case for reverse are the parameters.causality. The current level of health status depends upon the Taking logs of the Cobb-Douglas aggregate productioninitial income per capita and mean years of schooling of the function (3), we can obtain the following modelpopulation as specified in Pritchett and Summers (1996). Thedeterminants of health function can be specified as (4) lnYit = µit + α lnKit + βlnLit + γlnHit + δSit + λEit where µit is lnAit. The inputs K, L and H are endogenous(2) lnHit = β1 + β2lnYit + β3lnHExpit + β4Sit + β5Pit + eit , and also measured with errors. To overcome these problems, the practice adopted in literature is to instrument the inputs i=1,2,..,N States, t=1,2,…T periods where H, Y and S are as using their lagged values and this approach has been used indefined above, HExp is the per capita State government expen- this study for capital stock, total workers and LEB. Some lim-diture on health, water supply and sanitation, and family wel- itations in the data should be noted. For instance, the defi-fare, hereafter referred to as health expenditure in this study, nition of ‘workers’ has changed between 1961 and 1981.P is a measure of political power, βi are the parameters to be The SRS provides data on LEB only from 1970-71 and henceestimated and e is the random error term assumed to be dis- the LEB for the year 1960-61 is based on the estimates of thetributed with zero mean and constant variance. Increases in population census of 1961.The estimation methodologyper capita income of the people and public expenditure are depends upon the assumption we make about the technol-expected to improve the health status of the population. The ogy parameter µit. If µit is assumed to be the same for allpolitical power factor should influence public spending in a States over a period of time, then the production functionwelfare state. Two variables are considered. One is the per- (4) can be estimated by the OLS or by instrumental variablescentage of votes gained by socialist and communist parties (IV) the elections in the decade. The larger the share of votes On the other hand, if all the States are at an identical tech-gained by the socialist and communist parties, the greater nology level but that technology itself changes over time astheir influence on public policy decisions such as govern- shown belowment spending on welfare measures like health. Hence it isexpected that political power will exert a positive effect on (5) µit = µt + withealth status. Another political variable considered is the where wit is the random disturbance, µt is the time-specificpercentage of Assembly seats gained by the ruling party at constant. Under this assumption, the equation can be estimatedthe Centre. The higher the number of Assembly seats won by by the ‘fixed effects’ (time) method or by including a set ofthe ruling party at the Centre in the State, the more likely the dummy variables for time. However, if the technology remainsState to get a higher share in central fund allocation. This vari- the same over a period of time but varies across States, then theable is thus expected to have a positive effect on health. Given assumption about the technology parameter can be stated asthat elections are held once every five years under normal con-ditions, there were at least two elections in a decade. Hence (6) µit = µt + witwe assigned a weight equal to the number of years a partic- Specification (6) of the model can be estimated by intro-ular government stayed in power in a decade. ducing a set of dummy variables for States or by the ‘fixed The initial income per capita is likely to be endogenous effects’ (states) method. The fixed effects method enables usand researchers have instrumented initial income using lagged to control for unobserved time-specific or State-specific fixedvalues of the per capita income variable (Barro 1997). How- factors such as genetic factors, climatic conditions, region-ever, Bhargava et al. (2001) argue that lagged variables should specific health problems, etc. An alternative approach is thebe treated as endogenous. Pritchett and Summers (1996) random effects model that can be estimated by the feasibleexperimented with alternative instruments-terms of trade GLS method (Bloom et al. 2004). We have tested for theshocks, investment/GDP ratio, black market premium and fixed versus random effects specification of the model usingprice level distortions in their cross-country study. These vari- Hausman’s (chi-sqaure) specification test.ables are probably less relevant within a country, and infor-mation on these is not available at the State level in India. Empirical results: Estimates of economic In the above formulation, initial income is used to control growth and health equationsfor the transitional dynamics induced by factor accumula-tion. If, on the other hand, data on factor inputs are avail- Table 2 contains the OLS and two-stage least squares (2SLS)able, it is possible to formulate a model in which the change estimates of the Barro-type growth equation (1). The depend-10 Financing and Delivery of Health Care Services in India
  20. 20. Health, Poverty and Economic Growth in India SECTION Ient variables are the growth rates of real per capita income The 2SLS estimates are reported in column 3. The effect ofover the three periods 1970-80, 1980-90 and 1990-2000, log LEB is positive but the coefficient is significant only atand the explanatory variables are initial levels of income, LEB, the 10% level. The last specification includes two demographicyears of schooling, ratio of working age population to total variables-ratio of working age over total population and itspopulation and the growth rate in the ratio of working age growth rate over the decade. The effect of working age overover total population. The first column provides the OLS total population is positive and statistically significant at theestimates of the initial levels of log per capita income and 5% level. An increase in the share of working age populationlog LEB. The effect of initial income on growth is positive increases the potential labour force which in turn increasesbut not statistically significant even at the 10% level. The pos- the growth rate. However, the growth in the share of work-itive sign of the coefficient of initial income implies that Indian ing age over total population is negative and not statisticallyStates do not converge to a steady-state growth of per capita significant.income. This is largely at variance with the cross-country The IV estimates of the effect of income on health are reportedevidence that supports the conditional convergence hypoth- in Table 3. The estimates, given in column 1, show that bothesis. However, similar findings emerge from a study by Sachs log per capita income and per capita health expenditureet al. (2002) using panel data on Indian States for the period have a positive and statistically significant effect on LEB, as1980-98. In the case of China, there are marked differences expected. A 10% increase in per capita income would increasein findings, with convergence and divergence in various sub- the LEB by about 2% while a thousand rupee increase in perperiods, due to major shifts in economic policy (Sachs et al. capita health expenditure would lead to 1.3% increase in LEB.2002). The effect of log LEB on economic growth is positive Next, the average number of years of schooling is added inand the coefficient is statistically significant at the 5% level. the specification and the results reported in column 2 revealIn the second specification, we include years of schooling and that a substantial effect of per capita income and health expen-we observe that the effect of both log LEB and schooling turns diture is taken away by the schooling variable. Its effect is pos-out to be statistically insignificant. The high correlation between itive and highly significant (at the 1% level or better). The nextthese two variables (r=0.87), indicates that there is a problem specification (column 3) includes the percentage of votesof multicollinearity. Education is expected to influence health gained by the socialist and communist parties in the Assem-and the relationship is apparently quite strong at the macro bly elections. The effect of the political factor variable is pos-level. Hence, the schooling variable was excluded in the remain- itive but not significant. The other measure of the politicaling specifications of the model. variable, namely the per cent of Assembly seats won by the Table 2 Table 3 The effect of health on economic growth in The instrumental variable (IV) estimates of the India, 1970-2000 effect of per capita income and health Dependent variable: Growth rate in per capita NSDP expenditure on LEB, India, 1970-2000 over the decade Dependent variable: Log of life expectancy at birth (LEB) Explanatory variable OLS OLS 2SLS 2SLS Explanatory variable 1 2 3 1 2 3 4 Log per capita NSDP* 0.174 0.078 0.820 (3.01) (1.85) (1.77) Log initial per capita NSDP 1.534 1.519 1.419 2.875 Per capita health expenditure 1.265 0.156 0.179 (1.58) (1.53) (1.27) (2.63) (in thousands) (2.35) (0.39) (0.42) Log initial LEB 5.567 5.172 5.990 -1.801 Average years of schooling 0.069 0.0671 (2.23) (1.29) (1.86) (0.43) (6.74) (5.17) Initial average years of schooling 0.463 Average percentage of votes 0.000170 (0.13) secured by socialist and communist (0.21) Log initial working age over 14.162 parties in the Assembly election total population (2.47) Constant 2.480 3.123 3.092 Growth of working age over -9.837 Adjusted R2 0.521 0.776 0.770 total population (1.37) Number of observations 42 42 42 Constant -33.223 -31.659 -33.690 -5.880 Adjusted R2 0.346 0.329 0.377 0.304 NSDP: net State domestic product Note: ‘t’ values are given in parentheses Number of States 14 14 14 14 * Instrumented using lagged values of per capital NSDP Number of observations 42 42 42 42 Source: Authors calculation NSDP: net State domestic product; LEB: life expectancy at birth; OLS: Ordinary least squares; 2SLS: two-stage least squares. Note: ‘t’ values are given in parentheses Source: Authors calculation Financing and Delivery of Health Care Services in India 11
  21. 21. SECTION I Health, Poverty and Economic Growth in Indiaruling party at the Centre, also had a positive effect on LEBbut the effect was not statistically significant. Table 4 Estimates of the aggregate production function,Panel data estimates of the aggregate India, 1970-2000Cobb-Douglas production function Dependent variable: Log (NSDP)The effect of health on output at the State level is examined OLS (levels) IV (levels) Inputs 1 2 1 2by estimating a production function as specified in equations(4-6). Output is measured by the real NSDP. To overcome theproblem of measurement errors and year to year fluctuations Log Labour 0.461 0.475 0.678 0.686in NSDP, predicted rather than actual values of NSDP are (7.15) (7.74) (10.86) (11.11)used. The NSDP for a particular year is predicted using its lagged Log Capital 0.414 0.384 0.355 0.341values. Two conventional inputs-capital and labour-are used (7.58) (7.23) (5.66) (5.46)in the production functions. At the State level, there is no infor- Log LEB 1.134 0.163 1.960 1.250mation on capital stock or investment, even though data on (4.51) (0.37) (5.74) (2.27)these two variables are available at the national level over a Years of schooling 0.105 0.0718period of time. In this study, we use the value of fixed capital (2.60) (1.64)net of depreciation for the manufacturing sector to capture Constant -7.639 -3.905 -13.793 -11.064the capital input. The capital stock should include public sec- R2 0.934 0.941 0.909 0.915tor investment as well as private investment in other sectors F statistics 244.00 204.89 180.48 140.40also. In the absence of such comprehensive data, the capital Number of States 14 14 14 14measure used in this study captures only the partial and not Number of observations 56 56 56 56the full effect of capital on aggregate output. The labour input NSDP: net State domestic product; OLS: Ordinary least square;refers to the total number of workers including main and IV: instrumental variable; LEB: Life expectancy at birthmarginal workers. The measure of health-LEB-and average Note: ‘t’ values are given in parentheses. The input variables (log labour, log capital and log LEB) in IV (levels) columnsyears of schooling are as defined in the previous section. Poten- are instrumented using lagged values of their values.tial experience and its squared term were also computed and Source: Authors calculationincluded but due to the small variation in these variables andhigh collinearity between the two variables, the parameter esti- stands rejected.mates turned out to be imprecise and hence were dropped from We begin the discussion with specification 1. The effect ofthe final analysis. The input variables-log capital, log labour the changes in the two conventional inputs-labour and cap-and log LEB-are instrumented using their lagged variables as ital-and LEB on the change in output is positive and statisti-in Bloom, Canning and Sevilla (2004). As there is high corre- cally significant (1% level) on output. The results suggestlation between the two human capital variables of LEB and that a 1% improvement in LEB would result in a 1%-2% increaseschooling, all the models are estimated with and without the in output. The effect of health on output is much higher thanschooling variable. the effect of the two conventional inputs. Specification 2 The OLS and IV estimates of the aggregate production func- includes the average years of schooling along with health andtion (4), based on the assumption that the technology is con- other conventional inputs. Schooling exerts a positive and sta-stant over time and across States, are reported in Table 4. The tistically significant effect at the 10% level. Both the magni-OLS and IV results reported in columns 1 and 3 indicate that tude and significance of the health effect on output are reducedthe conventional inputs-labour and capital-and health (LEB) due to inclusion of the education variable.exert a positive and statistically significant effect (1% level)on output. The magnitude of the coefficient of LEB is high, Economic growth, poverty and health:which is puzzling. The average number of years of schooling Theory and empirical evidenceis included in specification 2. The effect of the schooling vari-able is positive and statistically significant at the 1% level. Poverty is a measure of income that indicates inadequate com-However, once the schooling variable is included, the effect mand over material resources. The level of poverty in a coun-of LEB on output became statistically insignificant, which is try or region depends upon the level of income as well as itsdue to the high correlation between the two variables as dis- distribution. Any policies or programmes which alter the dis-cussed above. The coefficient estimates from the OLS and IV tribution of income would affect poverty. In a country or Statemethods are similar in sign but the standard errors of the with a large income inequality there would be a relatively largecoefficients are somewhat higher in the case of IV estimates. number of poor people or people with a low income (below The estimates of the fixed effects model under the assump- a fixed poverty line), even if the country/State has a high pertions made in equations (5) and (6) are reported in Table 5. capita income. A higher rate of economic growth would reduceThe Hausman specification test statistic suggests that the poverty if growth affects the distribution of income in wayserror terms are correlated with the inputs and thus the null that pulls up the bottom tail of the distribution. Countrieshypothesis that the random effects model is appropriate that pursue a growth-oriented strategy firmly believe that12 Financing and Delivery of Health Care Services in India