SPECIAL ARTICLEEconomic & Political Weekly EPW august 2, 200841Health Inequality in India: Evidence from NFHS 3William Joe, U S Mishra, K NavaneethamRecent research has witnessed considerable engagement with the task of comprehending the crucial determinants of health outcomes. It is observed that the burden of ill health is borne disproportionately by different population subgroups and that people of lower socio-economic status consist-ently experience poor health outcomes [Macinko et al 2003]. Several empirical studies have also acknowledged such income-related inequalities in health, propounded as the absolute income hypothesis [Kakwani et al 1997; van Doorslaer et al 1997: Humphries and van Doorslaer 2000]. In view of such findings, health promotion of the poor has emerged worldwide as a vital area for policy research and action. Policy initiatives and programmes strongly perceive that inequalities in the health outcomes of different population subgroups are characterised by certain systematic deprivations (such as poverty).Apparently, some of the Indian health policies and program-mes also attempt to eliminate deprivation in the provisioning of healthcare and achieve the objective of health equity.1 In order to achieve this objective, it is important to steer policymaking through timely and systematic assessment of prevailing health inequality,2 a task that so far does not seem to have received serious attention. Although a few studies have presented region-specific or population-subgroup-related health profiles for India, they are at best able only to reflect on disparities and not inequa-lities. While disparities are evaluated based on the positioning around aggregate outcome, inequalities have to be adjudged according to specific ethical or economic ideals. Moreover, for ensuring equitable and efficient allocation of public health resources, it is imperative to unravel the depth and the varied dimensions of health outcomes, especially through measures sensitised for equity concerns. Apart from these considerations, it is also of analytical interest to examine whether income inequa-lity itself poses as a public health hazard. This question has gained much academic attention but most of the findings of studies3 on the topic have remained inconclusive. The literature on health economics, which identifies this question as the relative income hypothesis states that the distribution of income in a society has a larger impact on population’s health than absolute income. Since most of the studies on relative income hypothesis are undertaken in the context of developed countries, it would be worthwhile to gather some insights from the Indian experience to further our understanding of the income-health nexus.In this article, we employ widely accepted measurement techniques to assess inequities in child health across different Indian states and draw some interesting conclusions on the relationship between income inequality and health inequality in the country. As we all know, health of children assumes The authors appreciate the editorial effort of P R G Nair towards making the manuscript read better.William Joe (william@cds.ac.in), U S Mishra (mishra@cds.ac.in) and K Navaneetham (nava@cds.ac.in) are at the Centre for Development Studies, Thiruvananthapuram, Kerala.This article utilises the National Family Health Survey-3 data and presents an empirical assessment of income-related health inequality in India. It undertakes a state-level analysis of inequities in child health by employing the widely accepted measures of concentration curves and concentration indices. It finds that the poorer sections of the population are beleaguered with ill health whether in the quest for child survival or due to anxieties pertaining to child nutrition. Further, an attempt is made to comprehend the relationship between income inequality and health status in the Indian context. The analysis reveals that the degree of health inequalities escalates when the rising average income levels of the population are accompanied by rising income inequalities. The income-poor sections have different needs and therefore, planning and intervention necessitates an understanding of the sources of inequality and recognition of the vulnerablegroups to arrive at efficient resource allocation and policy decisions.
All India Maharashtra Gujarat Madhya Pradesh Egalitarian Line 
SPECIAL ARTICLEEconomic & Political Weekly EPW august 2, 200843provided on the basis of the new international reference popula-tion released by World Health Organisation (WHO) in April 2006 [WHO Multicentre Growth Reference Study Group 2006] and accepted by the government of India [IIPS andORC Macro 2007]. All these variables are specifically defined in Table 1 (p 42). To focus attention on issues of association and causation, we have obtained information also on three other economic variables: One, the state-wise net state domestic product (NSDP) 2004-05 at factor cost, which is obtained through the statistics published by Central Statistics Organisation (CSO). The second is the information on public spending on health as a share of total health spending, which is taken from Rao et al (2005). In addition to these varia-bles we also required information on the income inequality levels across different Indian states. For this purpose we have used the unit level records of National Sample Survey’s (NSS) 61st round on consumer expenditure. Here, the consumption expenditure of the households is taken as a proxy for income and we have computed the Gini coefficient of inequality in per capita monthly consumption expenditure for all the states of India.3 Interstate Differences in Health InequalitiesIn this section, we examine the magnitude of income-related inequalities in health, across the different Indian states. For this purpose, we have computed theCI for the selected indicators of child health across all the Indian states (Table 2). The CI values for a range of child health indicators for the country as a whole are negative, confirming the prevalence of income-related health inequalities that are manifest primarily among the poor. On comparison of these inequalities across varied indicators of child health, inequalities are more pronounced in the case of the under-five mortalities, in undernutrition and the receipt of basic vacci-nations for immunisation. The under-five mortalityCC for all-In-dia as well as for three other major states (Maharashtra, Gujarat and Madhya Pradesh) with higher health inequality levels are shown in Figure 1 (p 42). All these CCs lie above the diagonal and thus, indicate a greater concentration of health eventualities among the poorer groups.While the CI value for under-five mortality at the national level is computed to be (-0.1582), it presents a reasonably wide range across various states with the minimum of (-0.0388) in West Bengal and maximum of (-0.4107) in Uttaranchal. Among the other major states, Maharashtra, Madhya Pradesh, Gujarat, Tamil Nadu and Punjab experience greater income-related inequalities in under-five mortality as against the states of Uttar Pradesh, Rajasthan and Bihar, which show much lower levels of inequalities. Apart from the differences in the magnitude of inequalities across the board, the negative values indicate vulnerabilities among the poor.Other than under-five mortality, similar inequality is assessed for a set of child health indicators, which include nutritional make-up, anaemia and child immunisation. As regards nutritio-nal make-up, the two dimensions namely stunting (low height-for-age) and the underweight (low weight-for-age) manifest inequalities at the all-India level ranging between (-0.1249) and (-0.1600). The all-India level inequality in weight-for-age based nutritional assessment being the largest depicts a similar pattern across states as well. Compared with stunting, the inequality in nutrition according to the underweight criterion has a wider range between (-0.0835) in Madhya Pradesh and (-0.3063) in Goa. The level of overall prevalence of the same could also condi-tion a moderate range of inequality across states for the alterna-tive nutritional measures. This is obvious from the fact that prevalence of undernutrition according to the underweight crite-rion is by far the largest when contrasted with the same evalua-ted on an alternative criterion like stunting. Further, weight-for-age in its own construct has a propensity for larger variation during childhood. As regards stunting the all-India concentration index value is (-0.1249) with a variation range of (-0.0325) in Meghalaya and (-0.2867) in Goa. Not only is the range of varia-tion in this inequality measure relatively lower compared to the same according to the underweight criterion but also the high inequality magnitudes are lesser in this case.For the indicator of anaemia the all-India CI value of (-0.0518) is observably lower and could be due to the widespread preva-lence of anaemia across the population but still the poorer sections are found to remain at a higher disadvantage. The inequities in child-anaemia do not vary significantly across the major states. However, the states of Mizoram (-0.1,363), Goa (-0.1,126), West Bengal (-0.0919) and Orissa (-0.0851) are found to be more inequitable. In addition to these health outcome Table 2: CI for Inequalities in Child Health IndicatorsStates CIU5MRCIANECIH/ACIW/ACINFIAndhra Pradesh -0.0704 -0.0367 -0.1311 -0.1650 -0.0963Arunachal Pradesh -0.1401 -0.0587 -0.1167 -0.1816 -0.1296Assam -0.0541-0.0581-0.1302-0.1373-0.1079Bihar -0.0882-0.0389-0.0861-0.0962-0.1340Chhattisgarh -0.0764-0.0389-0.0669-0.1133-0.1443Delhi -0.1835-0.0666-0.1313-0.1410-0.2079Goa -0.1282 -0.1126 -0.2867 -0.3063 -0.2893Gujarat -0.2198-0.0658-0.1127-0.1432-0.1542Haryana -0.1304-0.0524-0.1408-0.1260-0.3341Himachal Pradesh -0.2186 -0.0406 -0.1305 -0.1323 -0.1589Jammu and Kashmir -0.1656 -0.0169 -0.1690 -0.2258 -0.2341Jharkhand -0.0546-0.0624-0.0803-0.0876-0.1131Karnataka -0.1325-0.0339-0.1284-0.1648-0.1823Kerala -0.1274-0.0314-0.1628-0.2026-0.2719Madhya Pradesh -0.2081 -0.0406 -0.0683 -0.0835 -0.1810Maharashtra -0.2481-0.0444-0.1427-0.1796-0.1795Manipur -0.3458-0.0097-0.1409-0.1805-0.1975Meghalaya -0.1152-0.0525-0.0325-0.0811-0.0957Mizoram -0.1942-0.1363-0.1606-0.2400-0.2130Nagaland -0.1646NA-0.1328-0.1645-0.1113Orissa -0.0844-0.0827-0.1865-0.1811-0.1328Punjab -0.1688-0.0331-0.2082-0.2597-0.2505Rajasthan -0.0801-0.0198-0.1043-0.1337-0.0898Sikkim -0.0581-0.0171-0.08480.0200-0.0725Tamil Nadu -0.1749 -0.0346 -0.1463 -0.1936 -0.0523Tripura -0.2251-0.0381-0.1113-0.1421-0.2306Uttar Pradesh -0.0960 -0.0271 -0.0885 -0.1181 -0.0754Uttaranchal -0.4107-0.0710-0.1924-0.1997-0.2302West Bengal -0.0388 -0.0919 -0.1716 -0.1660 -0.1231All India -0.1582 -0.0518 -0.1249 -0.1600 -0.1595The CI ranges between +1 and -1 and takes negative (positive) values when the ill health outcomes are concentrated among the poor (rich).CIU5MR-(CI) for under-five mortality, CIANE-CI for anaemia, CIH/A-CI for stunting, CIW/A-CI for underweight and CINFI- CI for not fully immunised.Source: Computed by authors using NFHS 3 (2005-06) unit level records.
H2 Health H1 H3 H4 Y1 Y2 Y3 Y4 Income
SPECIAL ARTICLEEconomic & Political Weekly EPW august 2, 200845which undoubtedly help detect the ailment accurately but impor-tantlyrequire additional expenditure. It must be noticed that many such expenses are indivisible and unavoidable under conditions of feeble health systems as is the case in many developing countries. Such specific difficulties in accessing quality medical care provide inadequate (or lower) returns to health at low levels of income. This line of reasoning is conceived in terms of the initial convex region of the income-health function depicted in Figure 2.Under such a framework, richer individuals are likely to end up with higher levels of health but it also suggests that individual incomes have to exceed a certain threshold (somewhere close to Y3) to able to meet the initial expenditure requirements for medical care in order to reap greater health benefits. For instance, consider two individuals with incomes Y1 andY2 respectively as shown in Figure 2. In the absolute sense, both these incomes are low and thus, lead to low levels of health. But still, there exists a certain degree of income inequality between these individuals (absolute and relative income inequality, given byY2 – Y1, Y2/Y1) that leads to health inequalities (given byH2 – H1, H2/H1). It is important to note that the inequalities in health, both in absolute and relative senses, are smaller than the inequalities observed in the income distribution and suggest that at lower levels of income, health inequalities are also low. But if individuals are around the threshold income level beyond which they would be able to afford better healthcare, then the relationship between income and health inequalities worsens. To demonstrate this fact, consider two individuals with incomes Y3 andY4 respectively and allow for a considerable degree of income inequality between them (i e,Y4 – Y3, Y4/Y3). Here, unlike in the earlier case, we observe that despite similar degrees of income inequalities, the level of health inequality (H4 – H3, H4/H3) has increased with increase in incomes.In a nutshell, the modification of the income-health function allows one to infer that for a given level of income inequality, if overall income levels are lower (higher) then health inequalities are also lower (higher). It also suggests that the levels of incomeinequality also have significant bearing upon the extent of health inequality but that the impact becomes more observable if the income inequalities are associated with higher levels of income. More importantly, under conditions of lower incomes and high-income inequality, the health inequality levels would get enhanced whereas if income levels are higher and income inequality levels are low, they would have a moderating impact on health inequality levels. Another related discussion that is relevant here is the impact of public health spending upon healthinequality. Although it is desirable that such facilities should be distributed more evenly across the population, the actual result may be undesirable as health facilities provided through public health spending often tend to be concentrated inparticular regions such as urban areas or certain other target-locations thereby, often failing in guaranteeing universal accessand opportunity. Any such bias in the provisioning of public health could thus worsen the distribution of health across individuals. In order to quickly verify the predictions of these two different frameworks in the Indian context, a simple regression exercise is undertaken here. This analysis could also be viewed as a preliminary attempt to comprehend the differences in health inequality across the different states of India in terms of income inequality, per capita income and share of public health spending. We have selected the negative of the under-five mortality CI as an indicator of child health inequality. As explanatory variables, the Gini measure of inequality in per capita monthly consumption expenditure is taken as a proxy for income inequality, per capita NSDP at factor cost is utilised to represent the state per capita income and public spending on health as a share of total health spending is taken to represent the role of subsidies in healthcare. The results from the regression analysis are presented in Table 4.Model 1 shows that in the Indian context, income inequality is positively but insignificantly related with levels of health inequality. The R-squared value suggests that hardly 2 per cent of the variations in health inequality are actually explained by the differences in income inequalities. This finding is similar to what Wagstaff (2002) finds while comprehending the differences in health inequality across developing countries. Given the inability of income inequality alone to capture the variations in health inequality, we add other important variables to comprehend the causation. Specifically, in model 2 we control for income inequality and public health spending levels and thereby attempt to elicit the role of per capita income in determining health inequalities. The results endorse the view that increases in average income also increase the levels of health inequality as indicated by the positive and significant coefficient of NSDP per capita. The theoretical framework discussed earlier has predicted this relationship. However, it is also observed that the coefficient obtained for the variable of public health spending as a proportion of total health spending possesses a negative sign, suggesting its favourable effect for reducing health inequalities. However, the effect turns out to be statistically insignificant.The overall results obtained here (in models 1 and 2), especially in relation to income inequality and average income, are partly in agreement with the framework but do not lend any concrete support to the relative income hypothesis. In other words, it may also be opined that the concave relationship between income and health is somewhat unable to capture the conditions prevalent in developing countries. Hence, now we go on to test the alternate framework namely of the convex-concave relationship between Table 4: Regression Results for CIs of Under-Five MortalityVariable Model 1 Model 2 Model 3 ParameterParameterParameter (t-statistics) (t-statistics) (t-statistics)Constant 0.0760.0800.048 (0.992) (0.920) (1.482)Gini coefficient 0.183 0.160 (0.767) (-.0546) NSDP per capita 9.49E-06** (2.429) Public spending on health as % total -0.00012 (-0.124) Avg of Gini and normalised NSDP per capita 0.211*** (2.894)F statistic for model 0.588 2.438* 8.378***R-squared 0.0240.2490.267Adjusted R-squared -0.017 0.147 0.235N 26 26 25*** Significant at the 1% level, ** significant at the 5% level.SPECIAL ARTICLE
health and income. What is necessary here is to conceive of a variable, which should be sensitive to both, average income levels as well as income inequality levels. For this purpose we develop a simple composite index of income and income inequality in two small steps as follows. Firstly, we normalise the NSDP per capita across the states in such a way that the state with the highest income obtains a value of one and the lowest-income state is assigned a value of zero. In the second step, we obtain the composite index for each state, by taking a simple average of the Gini coefficient for the state and the normalised NSDP per capita values for the respective states obtained through the previous step. This new index represents the relative levels of both income and income inequality and is used as an explanatory variable in the analysis to comprehend health inequality.
Model 3 finds that this new composite index measure given by the average of the Gini and normalised NSDP per capita turns out to be a statistically significant factor explaining the variations in health inequality with an explanatory power of 27 per cent, greater than that of the previous two models. Besides, the coefficient value too is higher (0.211) which is statistically significant at the 1 per cent level of significance. This result supports the theoretical prediction that if higher levels of average incomes are accompanied by higher income inequalities then it leads to increase in health inequality. Further, effects on health inequality get cushioned if either the distribution of income is more equitable or if the average income levels are lower. Although here we have not performed a rigorous analysis of the said framework, this exercise could be considered as a preliminary attempt at gauging the proposed health-income relationship. Undoubtedly, a more comprehensive examination would be able to help draw further insights because income alone is insufficient to describe the larger variations observed in health inequity, as is evident from the explanatory power of the regression.
5 Policy Notes and Conclusion
We shall begin this concluding section by reiterating the motivation behind this present engagement. The concern beneath this empirical exercise is the sheer urgency to unravel the inadequacies of the summary measures in vogue of health outcomes and to evoke policy concerns pertaining to social justice and equity. It is disconcerting to witness, especially from an ethical perspective, that poorer populations in India are bearing the brunt of health disadvantages. Although certain policy interventions are in place to deal with such adversities, greater attention needs to be directed towards the assessment of health deprivation and inequities in India. Also, we would like to stress upon the recognition of differential constraints in accessing medical care across regions. For instance, for some, availability may be an issue while for others it may not actually be the major worry. Similarly, availability alone may not be sufficient; unless it is supported by a policy of greater subsidisation of health facilities through special schemes for maternal and child healthcare. The problem may as well be one of poor levels of awareness for some others. Given such possibilities, the social planner has to acquire
august 2, 2008 EPW Economic & Political Weekly
SPECIAL ARTICLEEconomic & Political Weekly EPW august 2, 200847more complete information with regard to the sources of inequal-ity and identification of the vulnerable groups. Undoubtedly, such an exercise would go a long way to optimise resource allocation and enhance the targeting efficiency of such interventions. The present analysis pursues this thinking and endeavours to enhance the informational base for policymaking, by incorporating into the summary measures, a slightly more elaborate account of health inequality caused by a factor such as income inadequacy.While analysing the income component, it was observed that poorer sections of the population were beleaguered with ill-health whether it be their efforts for child survival or anxieties pertain-ing to child nutrition. Another highlight worth mentioning here relates to the inequality levels being higher with higher levels of the event as against lower levels. Undoubtedly from a policy perspective, focused attention needs to be paid towards improv-ing the mortality situation in backward states and perhaps inequality aversion measures need to be promoted in states like Maharashtra and Gujarat with lower under-five mortality rates (of 47 and 61 per 1,000 live births respectively) in order to obviate the concentration of this misfortune among the poor. Further, a simple theoretical model was resorted to comprehend associa-tional and causational factors. The analysis revealed that the degree of health inequalities escalates when the rising average income levels of the population are accompanied by greater income inequalities. On the one hand, such an association does reflect that the product of economic growth in the form of rising average income and income inequalities presents certain impedi-ments to attaining equitable health by allowing the better-off population to secure greater benefits of the growth process. On the other hand, it is also evident that the summary measures of health also improve with the betterment of the income profile of a region. The interplay of these two impacts may actually help policymakers trade off a little bit of health inequality for gaining higher health levels. However, a social planner needs to sail through such quandaries and should arrive at prudent mecha-nisms to utilise the resources and technology obtained through economic growth, by allocating greater resources towards those sections of the population who have been excluded from the growth process [Wagstaff 2002]. Even the countries with the shallowest health gradients, such as Sweden and England, have viewed their own health inequalities as unacceptable and have initiated policy measures to mitigate those [Daniels et al 2000].We now turn to the larger question, namely, the one relating to the type of social policies that could be pursued by the state to reduce health inequalities. Scholars have advocated for a policy matrix, which not only accommodates immediate or direct health interventions such as medical facilities but also consists of basic interventions indirectly related with the health of individuals. Such investments are largely sought in the form of investment in basic education, better housing, water and sanitary conditions as well as the introduction of programmes to provide income security. By suggesting a holistic policy matrix, our contention here is not as much to argue for allocation of resources but rather to suggest an exercise that would integrate these basic invest-ments, at least, at an analytical level while arriving at resource allocation decisions for the health sector itself. Decisions on resource allocations for public health, taken in isolation from other pertinent factors, may actually affect the efficacy of the policy matrix in toto. Perhaps the state should acknowledge the fact that social sector expenditures, particularly on health and education, are complementary in nature and if put together do produce large individual as well as social benefits.Notes1 Equity as defined by the International Society for Equity in Health is: “The absence of potentially remediable, systematic differences in 0ne or more aspects of health across socially, economically, demographically, or geographically defined population groups or subgroups”.2 Recently, the World Bank, in cooperation with the Dutch and Swedish governments, has sponsored a set of reports providing basic information about health inequalities within countries. As a result of this collective initiative, the basic information (for 1992/93 and 1998/99) about health, nutrition and population inequalities is published in the report on India [Gwatkin et al 2007].3 Wagstaff and van Doorslaer (2000) conducted a literature review of individual level studies on the impact of income inequality on health. In their review of six major studies, they found that the literature reveals strong support for the absolute income hypothesis and little or no support for the relative income hypothesis. Also see Macinko et al (2003).4 Correlation matrices for the CIs obtained through Pearson correlations and Spearman’s rank corre-lation suggest that there are high and significant correlations between the CIs for these indicators. The CIs for the indicators of inequalities in under-five mortalities are significantly correlated with the indicators of malnourishment and with the inequities in incomplete immunisation. In other words, states with a high level of inequality in under-five mortality consistently have high levels of inequality on the other indicators.5As discussed in Section 2, the consumption expenditure of the households is taken as a proxy for income and we have computed the Gini coeffi-cient of inequality in per capita monthly consump-tion expenditure for all the states of India using unit level records of the NSS 61st round on consumer expenditure.ReferencesDaniels, N, B Kennedy and I Kawachi (2000):Is Inequality Bad for Our Health?, Beacon Press, Boston.Gwatkin, et al (2007): ‘Socioeconomic Differences in Health, Nutrition and Population: India’, Country Report on HNP and Poverty, The World Bank, Washington DC.Humphries, K H and E van Doorslaer (2000): Income Related Health Inequalities in Canada,Social Science and Medicene, 50, pp 663-71.International Institute for Population Sciences (IIPS) and ORC Macro (2007): National Family Health Survey (NFHS 3), 2005-06, IIPS, Mumbai, India.Jenkins, S (1988): ‘Calculating Income Distribution Indices from Microdata’, National Tax Journal, Vol 41, No 1, pp 139-42.Kakwani, N C (1980): Income Inequality and Poverty; Methods of Estimation and Policy Applications, Oxford University Press, New York.Kakwani, N C, A Wagstaff and E van Doorslaer (1997): ‘Socioeconomic Inequalities in Health: Measure-ment, Computation and Statistical Inference’, Journal of Econometrics, Vol 77, No 1, pp 87-104.Lerman, R I and S Yitzhaki (1989): ‘Improving the Accuracy of Estimates of Gini Coefficients’, Journal of Econometrics, Vol 42, No 1.Macinko J A, L Shi, B Starfield, J T Wulu, Jr (2003): ‘Income Inequality and Health: A Critical Review of the Literature’,Medical Care Research and Review, Vol 60, No 4, pp 407-52.National Sample Survey Organisation (2007): ‘Consumer Expenditure Survey 61st Round’, Central Statistical Organisation, New Delhi.Rao, K S et al (2005): ‘Financing of Health in India’, Background paper on Financing and Delivery of Healthcare Services in India, National Commission on Macroeconomics and Health, New Delhi.van Doorslaer E et al (1997) ‘Income Related Inequali-ties in Health: Some International Comparisons’, Journal of Health Economics, Vol 16, pp 93-112.Wagstaff, A (1986): ‘The Demand for Healthcare: A Simplified Grossman Model’,Bulletin of Economic Research, Vol 38, No 1, pp 93-95.– (2002): ‘Inequalities in Health in Developing Countries: Swimming against the Tide?’, Policy Research Working Paper No 2795, The World Bank, Washington DC.Wagstaff, A and E van Doorslaer (2000): ‘Equity in Healthcare Finance and Delivery’ in A J Culyer and J P Newhouse (eds),Handbook of Health Economics (1B), Elsevier: Amsterdam, Chapter 34, pp 1803-62.Wagstaff, A, P Paci and E van Doorslaer (1991): ‘On the Measurement of Inequalities in Health’,Social Science and Medicene, Vol 33, No 5, pp 545-57.SPECIAL ARTICLE
august 2, 2008 EPW Economic & Political Weekly
