Hrev_master [page 26] [Healthcare in Low-resource Settings 2014; 2:1866] Efficiency of social sector expenditure in India: a case of health and education in selected Indian states Brijesh C. Purohit Madras School of Economics, Kottur, India Abstract Social sector expenditure in India captures a number of important aspects including health, nutrition, education, water supply, san- itation, housing and welfare, among others. Over a period of time, besides budgetary outlay on this sector, private sector has also played a considerable role. Thus, efficiency of expendi- ture in this sector by state government has to be reckoned both in terms of relative levels of various aspects across the states and in terms of comparable benchmarks for different aspects of the sector. This paper attempts an analysis of social sector efficiency focusing on two major aspects: health and education. Unlike other studies on the Indian context, this analysis focusing on major states in India uses both non-parametric and parametric approaches. Although both approaches provide benchmarks to judge relative efficiency across states, the former provides a yardstick more at an aggregative level without parametric restrictions, whereas the latter is used for major focus on health care aspects. Results of free disposal hull analysis are suggestive of a considerably more scope for improvement in efficiency of public expenditure in health rela- tive to education. Our results of stochastic frontier analysis indicate considerable state level disparities which could be reduced through a mix of strategies involving realloca- tion of factors (namely, manpower and supply of consumables) within the sector, mobilizing additional resources possibly through enhanced budgetary emphasis, or encouraging more private sector participation. Based on our results, this may enhance efficiency by nearly 20% in health care sector and increase avail- ability and equity across low performing and poorer states like Madhya Pradesh and Uttar Pradesh. Introduction Social sector comprises an important item in the state budgetary expenditure. It has remained around 5.8% of gross domestic prod- uct and its share in total state expenditure has varied between 36.8 (in 1990-1995) to 39.2% (2010-2011).1 Within social sector, major chunk (nearly 57%) is being spent on educa- tion, sports, art and culture (46.1%) and med- ical and public health (10.5%). The other items which include: family welfare and water supply and sanitation, housing, urban development, welfare of scheduled castes, scheduled tribes and other backward castes, labour and labour welfare, social security and welfare, nutrition, natural calamities and the rest, comprise a low percentage which varies from 1.3% (natural calamities) to 9.6% (social security and wel- fare) of total social sector. It becomes pertinent therefore to analyse whether the major expen- diture sectors like health and education are performing satisfying the criteria of efficiency. Several approaches for measuring the efficien- cy of government expenditure have been pro- posed in the literature.2 In general, these approaches are broadly of four types. First, studies which have concentrated on gauging and enhancing efficiency by focusing on cer- tain types of government spending in a specif- ic country. Secondly, those only which use data on inputs of government spending in quantita- tive terms, but not on outputs. Third, those using only outputs, but not inputs. Finally, those which have looked at both inputs and outputs; these studies, however, have not made a consistent comparison of the efficien- cy of government spending among countries.3- 6 These studies do not explicitly analyze the relationship between government spending and social indicators. Within each of the approaches, however, one may distinguish the studies which have focussed only on developed country (or countries) or only on developing country (or countries) and further in terms of their interest in education and health sector also. Thus, the issue of gauging and enhanc- ing government efficiency continues to inter- est policymakers and researchers alike.2,7-9 This interest received a boost with the initia- tion of wide-ranging institutional reforms by some of the developed nations10-12 which aimed at improving the efficiency of the public sector. These reforms basically were to separate poli- cy formulation from policy implementation, create competition between government agen- cies and between government agencies and private firms, and develop output-oriented budgets using a wide array of output indica- tors. This practice of result-oriented public expenditure management has generated a wealth of information on how to control pro- duction processes within the government and how to enhance their efficiency. Pertaining to education sector, for instance, there are certain studies which analyse both inputs and outputs. For instance, Harbison and Hanushek13 provide an overview of 187 studies of education production functions in the United States and 96 studies of education pro- duction functions in developing countries and investigate the relation between education inputs and outputs. Another type of analysis, for instance by Tanzi and Schuknecht14 assess- es the incremental impact of public spending on social and economic indicators in industri- al countries and conclude that higher public spending does not significantly improve social welfare. In most studies of developing coun- tries, it is found that teacher education, teacher experience, and the availability of facilities have a positive and significant impact on education output, and that the effect of expenditure per pupil is significant in half the studies; the pupil-teacher ratio and teacher salary have no discernible impact on education output. Likewise, Jimenez and Lockheed15 also assess the relative efficiency of public and pri- vate educations in several developing coun- tries by taking into account both inputs and outputs. In regard to health care sector, for instance, among developed nations, using regression analysis and focusing on inputs, a study of OECD member countries covering 20 years analyzed the efficiency of health care systems. They show that public-reimburse- ment health systems, which combine private provision with public financing, are associated with lower public health expenditures and higher efficiency than publicly managed and financed health care systems.16 This is traced by looking at factors associated with a high rel- atively expensive in-patient care and the lack of a mechanism to restrain demand for special- Healthcare in Low-resource Settings 2014; volume 2:1866 Correspondence: Brijesh C. Purohit, Madras School of Economics, Gandhi Mandapam Road, Kottur, Chennai-600025, India. Tel. +91.044.2230.0304 - Fax: +91.044.2235.4847. E-mail: brijeshpurohit@gmail.com Key words: social sector expenditure, India, health, education. Acknowledgments: an earlier version of this paper was presented at National Conference on Social Sector in India: Issues and Challenges, March 29-30, 2013, Golden Jubilee Celebrations 2012-13, Centre of Advanced Studies, Department of Analytical and Applied Economics, Utkal University, Odisha, India. Thanks are due to par- ticipants of this conference for their valuable comments. Received for publication: 7 August 2013. Revision received: 2 October 2013. Accepted for publication: 3 November 2013. This work is licensed under a Creative Commons Attribution 3.0 License (by-nc 3.0). ©Copyright Brijesh C. Purohit 2014 Licensee PAGEPress, Italy Healthcare in Low-resource Settings 2014; 2:1866 doi:10.4081/hls.2014.1866 Non -co mmerc ial us e o nly [Healthcare in Low-resource Settings 2014; 2:1866] [page 27] ized health care. Countries without ceilings on in-patient care were also found to have higher public health expenditure. A number of studies have laid emphasis on the overall health sys- tem performance and its impact on health out- comes.17,18 More often an idealized yardstick is developed which is used to evaluate economic performance of health system. There are a number of studies in health care sector which employ either non-parametric approaches like free disposable hull (FDH) or data envelop- ment analysis (DEA) or parametric approaches like stochastic frontier analysis (SFA). In the former category with a focus on devel- oped world one may include, for instance, Aubyn19 who used FDH covering both the health and education sectors in Portugal, Hofmarcher and colleagues20 for an Austrian province, Puig-Junoy and Gannon21,22 for Ireland, Magnussen23 for Norway, Jeffrey and Coppola24 relating to USA, Bates and col- leagues25 for the USA, Kontodimopoulos and colleagues26 for Greek hospitals, and Spinks and Hollingsworth27 for OECD countries. Likewise, with a focus on developing nations some notable studies include a report on dis- trict hospitals in Namibia,28 Masiye29 for Zambian hospitals, Mathiyazhgan30 for hospi- tals in Karnataka State in India, Mirmirani31 for transition economies of former socialist block including Albania, Armenia, Russia and others, Kittelsen and Magnussen32 for Norway, Li and Wang33 relating to Chinese public acute hospitals, Hajialiafzali and colleagues34 relat- ing to Iran, and Suraratdechaac and Okunadeb35 for Thialand. In the latter type of studies using SFA, one may include with a focus on developed nations, studies for instance, by World Health Organization36 covering different nations, Murray and Frenk,37 Worthington,38 Jamison and colleagues,39 and Salomon and others40 relating to inter country comparison, Schmacker and colleagues41 relating to USA, Evans and others42 for a cross country compar- ison and Greene,43 Farsi and others44 relating to Switzerland, Wang and others45 for New South Wales, Kris and others46,47 relating to Texas, Rosko48 relating to USA, Yong and Harris49 relating to Australia, Hollingsworth and Wildman50 for a cross country comparison, Mortimer and Peacock51 relating to Australia, and Jayasuriya and Wodon52 for a comparison among nations. Among studies focused on India one may include Sankar and Kathuria53 and Purohit.9,54-56 These latter types of studies have deployed frontier efficiency measure- ment techniques which involve a production possibility frontier depicting a locus of poten- tially technical efficient output combination that an organization or health system is capa- ble of producing at a point of time. An output combination below this frontier is termed as technically inefficient.57-59 Despite its nascent nature of application in healthcare sector, an exhaustive review of studies applying these methods has been attempted which provides us in detail the steps and empirical problems that have been highlighted by researchers.38,60 Notably there are very few studies in the devel- oping countries’ context and except a few par- ticularly in the Indian context, which have focused on this aspect; the literature is nearly marked by absence for recent period. Our study thus covers this gap for India for the latest period. Hypothesis and objective We hypothesize that States differ in their technical efficiency pertaining to health and education systems due to factors which require emphasis in facility planning in these sec- tors.9,53 It is also hypothesized that these factors differ from State to State according to their level of development.9 It is presumed that estimated efficiency parameters (from both types of analy- sis, i.e. non-parametric and parametric approaches) should help the health and educa- tion policy makers to improve State level system performance pertaining to these sectors. Materials and Methods Non-parametric approach: free dis- posable hull In this paper we use two types of tech- niques, namely non-parametric and paramet- ric, that allow for a direct measurement of the relative efficiency of government spending among countries or states within a nation. In the former type we apply FDH analysis which assesses the relative efficiency of production units in a market environment. This analysis consists of, first, establishing the production possibility frontier representing a combination of best-observed production results within the sample of observations (the best practices), and, second, measuring the relative inefficien- cy of producers inside the production possibil- ity frontier by the distance from the frontier. The major advantages of FDH analysis are that it imposes only weak restrictions on the pro- duction technology, while allowing for a com- parison of efficiency levels among producers. The only assumption made is that inputs and/or outputs can be freely disposed of, so that it is possible with the same production technology to lower outputs while maintaining the level of inputs and to increase the inputs while maintaining outputs at the same level. This assumption guarantees the existence of a continuous FDH, or production possibility frontier, for any sample of production results. Thus, FDH analysis provides an intuitive tool that can be used to identify best practices in government spending and to assess how gov- ernments are faring in comparison with these best practices.61-63 In our analysis using FDH, the term producer is meant to include govern- ments. A producer is relatively inefficient if another producer uses less input to generate as much or more output. A producer is relative- ly efficient if there is no other producer that uses less input to generate as much or more output. In the Appendix and Appendix Figures A and B, this is illustrated for the case of one input and one output. If a producer is engaged in the production of multiple outputs using more than one input, it becomes more difficult to establish relative efficiency. In such a situa- tion (of multiple inputs), it is postulated that a producer is relatively inefficient if he uses as much or more of all inputs to generate as much or less of all outputs than all other producer, with at least one input being strictly higher, or one output strictly lower. Depending upon the availability of latest and comparable informa- tion, we have applied this technique for data on major and smaller Indian States for educa- tion covering different cross sections from 2003-2011 and for health covering the period 2001-2010. This analysis covers 15 major Indian States [which include Andhra Pradesh (AP), Assam, Bihar, Gujarat Harayana, Karnataka, Kerala, Madhya Pradesh (MP), Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu (TN), Uttar Pradesh (UP), and West Bengal (WB)] and 10 smaller States [which include Arunachal Pradesh, Chhatisgarh, Goa, Himachal Pradesh (HP), Jammu and Kashmir (JK), Jharkhand, Manipur, Meghalaya, Mizoram and Nagaland]. Parametric technique: stochastic frontier method In the application of parametric techniques, stochastic methods can be used to correct for measurement and other random errors in the estimation of the production possibility fron- tier. In any parametric techniques a functional form is postulated for the production possibili- ty frontier, and then a set of parameters is selected that best fit the sample data. Model specification In the estimation of health system efficien- cy, our specification is based on a general sto- chastic frontier model that is presented as: lnqj = f(ln x) + vj- uj (1) where: ln qj is the health output [life expectan- cy (LEXP) or inverse of infant mortality rates (IMR)] produced by a health system j; x is a vector of factor inputs represented by per capi- ta health facilities (including per capita avail- ability of hospital beds, per capita primary health centers (or sub centers), per capita doc- tors, per capita paramedical staff, per capita skilled attention for birth; vj is the stochastic Article Non -co mmerc ial us e o nly [page 28] [Healthcare in Low-resource Settings 2014; 2:1866] (white noise) error term; uj is a one-sided error term representing the technical ineffi- ciency of the health system j. Both vj and uj are assumed to be independently and identically distributed with variance sv2 and su2, respective- ly. From the estimated relationship ln q^j=f (ln x) - uj, the efficient level of health outcome (with zero technical inefficiency) is defined as: ln q*=f (ln x). This implies ln TEj=ln q^ j - ln q*=- uj. Hence TEj=e-uj, 0<= e-u j<= 1. If uj=0 it implies e-uj=1. Health system is technically efficient. This implies that technical efficiency of jth health system is a relative measure of its output as a proportion of the corresponding frontier output. A health system is technically efficient if its output level is on the frontier which in turn means that q/q* equals one in value. Study design: sample and sampling technique This study uses secondary data published in official documents of government of India and State governments. Applying this data in any empirical study does not require any ethical approval. The study makes use of a purposive sampling and therefore focus is on 15 major Indian States. The purpose is to carry out an analysis which reveals broadly the country’s scenario at state level disaggregation. Data used thus are presumed to be authentic and therefore reliable. Validity of the results is thus subject to the reliability of official publications and underlying statistical techniques deployed in the study. For parametric approach, we cover 15 major Indian States [which include Andhra Pradesh (AP), Assam, Bihar, Gujarat Harayana, Karnataka, Kerala, Madhya Pradesh (MP), Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu (TN), Uttar Pradesh (UP), and West Bengal (WB)] and use panel data for 2005-2011. Use of panel data is preferred since it does not require strong assumptions about the error term and unlike the cross section data, the assumption of independence of tech- nical efficiency from factor inputs is not imposed.64,65 We extend our estimation to the second stage which presumes that differences in technical efficiency pertaining to health sys- tem can be discerned at the health facility planning level from non-health related param- eters. Thus, we explain the dispersion in tech- nical efficiency by a set of variables which includes per capita income, literacy, urbaniza- tion, per capita budgetary expenditure on health and rural water supply. Thus, our model in the second stage is: dispersion in technical efficiency=f (per capita income, literacy, urbanization, per capita budgetary expenditure on health and rural water supply) + error term (2) Thus main dependent variables used in the study are LEXP and dispersion; independent variables include per capita income and others namely, number of primary health centers Article Table 1. Input efficiency score: education (2008-2011). States Public expenditure Net enrolment primary IES Literacy IES (2008-09) (2008-09) (2008-2009) (2011) (2011) Major Andhra Pradesh 1195.59 79.12 0.67 67.66 0.85 Assam 1374.02 83.58 0.95 73.18 0.74 Bihar 725.89 53.38 1.00 63.82 1.00 Gujarat 1015.67 59.75 0.79 79.31 1.00 Harayana 1615.77 74.14 0.81 76.64 0.92 Karnataka 1429.04 69.14 0.92 75.60 0.71 Kerala 1661.71 84.71 0.79 93.91 1.00 Madhya Pradesh 799.49 97.28 1.00 70.63 1.00 Maharashtra 1487.72 88.93 0.88 82.91 1.00 Orissa 1193.44 69.16 0.67 73.45 0.85 Punjab 1395.89 74.15 0.94 76.68 0.73 Rajasthan 1096.43 76.54 0.73 67.06 0.93 Tamil Nadu 1310.20 119.56 1.00 80.33 0.78 Uttar Pradesh 763.40 56.35 1.00 69.72 1.00 West Bengal 943.52 87.17 0.85 77.08 1.00 Minor Arunachal Pradesh 3684.77 115.15 1.03 66.95 0.90 Chhatisgarh 1211.87 88.30 1.00 71.04 1.00 Goa 4648.96 62.04 0.81 87.40 0.81 Himachal Pradesh 3299.52 115.11 1.00 83.78 1.00 Jammu and Kashmir 1497.35 100.69 1.00 68.74 Jharkhand 1162.75 73.18 1.00 67.63 1.00 Manipur 2054.26 83.20 0.73 79.85 1.00 Meghalaya 2110.56 83.46 0.71 75.48 0.97 Mizoram 3780.70 104.75 1.00 91.58 1.00 Nagland 2339.54 88.34 0.64 80.11 1.00 IES, input efficiency score. Figure 1. Independently efficient states based on infant survival in 2003 and per capita public expenditure on health in 2001-2002. Non -co mmerc ial us e o nly [Healthcare in Low-resource Settings 2014; 2:1866] [page 29] (PHCs), sub-centers (SCs), community health centers (CHCs), hospitals and dispensaries, health manpower-medical and paramedical, and socio-economic parameters like income, education, and basic amenities, etc. Database This study is based on secondary data. Information is collected for the years 2005-11 from various sources including RBI Bulletin,1 Health Information of India66-72 and other pub- lished sources. At the all-India level, main vari- ables used in the study are LEXP, IMR, per capi- ta income and other parameters related to health infrastructure including number of PHCs, SCs, CHCs, hospitals and dispensaries, health manpower-medical and paramedical, and other variables relevant for depicting healthcare facilities, their utilization, health outcomes, socio-economic parameters like income, education, and basic amenities, etc. Statistical analysis tools used by our study include frontier regression technique applying STATA software. Results The results of our FDH analysis for educa- Article Table 2. Input efficiency score: health (2001-2005). States Public expenditure Infant survival IES Public expenditure Infant survival IES (2001-2002) (2003) (2004-05) (2006) Major Andhra Pradesh 182 941 0.81 191 944 0.91 Assam 176 933 0.83 162 933 1.07 Bihar 92 940 1 93 940 1.00 Gujarat 147 943 1 198 947 0.87 Haryana 163 941 0.90 203 943 1.00 Karnataka 206 948 0.95 233 952 0.88 Kerala 240 989 1 287 985 1.00 Madhya Pradesh 132 918 0.69 145 926 0.64 Maharashtra 196 958 1 204 965 1.00 Orissa 134 917 1.09 183 927 0.95 Punjab 258 951 0.93 247 956 0.83 Rajasthan 182 925 0.81 186 933 0.93 Tamil Nadu 202 957 1.18 223 963 0.91 Uttar Pradesh 84 924 1 128 929 0.73 West Bengal 181 954 1 173 962 1.00 Smaller Arunachal Pradesh 627 966 0.55 841 960 0.35 Chattisgarh 121 930 1 146 939 1.00 Delhi 426 972 0.81 560 963 0.53 Goa 685 984 1 861 985 0.34 Himachal Pradesh 493 951 0.49 630 950 0.46 Jammu and Kashmir 271 956 0.66 512 948 0.57 Jharkhand 146 949 1 155 951 1.00 Manipur 345 984 1 294 989 1.00 Meghalaya 407 943 0.85 430 947 0.68 Mizoram 836 984 1 867 975 0.34 Pondicherry 841 976 0.99 1014 972 0.29 Sikkim 825 967 1.01 1082 967 0.27 Tripura 301 968 1 328 964 0.90 Uttarakhand 178 959 1 280 957 1.00 Nagaland na na na 639 980 0.46 IES, input efficiency score; na, not available. Figure 2. Independently efficient states based on infant survival in 2006 and per capita public expenditure on health in 2004-2005. Figure 3. Independently efficient states based on infant survival in 2010 and per capita public expenditure on health in 2008-2009. Non -co mmerc ial us e o nly [page 30] [Healthcare in Low-resource Settings 2014; 2:1866] tion and health sector using data for Indian states, both major and smaller ones, are pre- sented below in Figures 1-5 and Tables 1-3. Free disposable hull analysis It can be observed that for per capita public expenditure on health (in 2001-02), independ- ently efficient states that emerged from FDH for major states are UP, Bihar, Gujarat West Bengal, Maharashtra and Kerala (Figure 1). Among the smaller states the independently efficient states are Chhatisgarh, Jharkhand, Uttarakhand, Tripura and Manipur (Figure 1). Likewise, in Figure 2 (for 2004-2005 per capita public expenditure), the situation is somewhat changed for UP whereas other independently efficient states remain the same. Among smaller states a changed situation with lower efficiency is depicted for Tripura only (Figure 2). Free disposable hull for public expenditure in 2008-09 for health sector (Figure 3) depict additional states namely WB and Tamil Nadu among independently efficient states (Figure 3) and inclusion and exclusion of Goa and Chhatisgarh respectively in the category of such (independently efficient) states (Figure 3). In education sector, using literacy (2011) and public expenditure (2008-09), the states like Bihar, UP, WB, Gujarat, Tamil Nadu. Maharashtra and Kerala (among major states) and Jharkhand, Chhatisgarh, Manipur and Article Table 3. Input efficiency score: health (2010). States Public expenditure (2008-2009) Infant survival rate (2010) IES Major Andhra Pradesh 410.00 954.00 1.00 Assam 471.00 942.00 0.96 Bihar 173.00 952.00 1.00 Gujarat 270.00 956.00 1.00 Harayana 280.00 952.00 0.99 Karnataka 419.00 962.00 0.98 Kerala 454.00 987.00 1.00 Madhya Pradesh 235.00 938.00 0.74 Maharashtra 278.00 972.00 1.00 Orissa 263.00 939.00 1.06 Punjab 360.00 966.00 0.77 Rajasthan 287.00 945.00 0.97 Tamil Nadu 410.00 976.00 1.00 Uttar Pradesh 293.00 939.00 0.95 West Bengal 262.00 969.00 1.00 Smaller Arunachal Pradesh 771.00 969.00 0.90 Chhattisgarh 378.00 949.00 0.87 Delhi 840.00 970.00 0.83 Goa 1149.00 990.00 1.00 Himachal Pradesh 884.00 960.00 0.96 Jammu and Kashmir 845.00 957.00 0.82 Jharkhand 328.00 958.00 1.00 Manipur 695.00 986.00 1.00 Meghalaya 690.00 945.00 0.91 Mizoram 1611.00 963.00 0.71 Puducherry 1333.00 978.00 0.86 Sikkim 1446.00 970.00 0.79 Tripura 740.00 973.00 0.94 Uttarakhand 630.00 962.00 1.00 IES, input efficiency score. Figure 4. Independently efficient states based on literacy in 2011 and per capita public expenditure on education in 2008-2009. Figure 5. Independently efficient states based on net enrolment primary in 2008-2009 and per capita public expenditure education in 2008-2009. Non -co mmerc ial us e o nly [Healthcare in Low-resource Settings 2014; 2:1866] [page 31] Himachal Pradesh (among smaller states) emerge as independently efficient states (Figure 4). By and large a similar observation could be made using net enrolment primary in 2008-09 (Figure 5). Using this FDH analysis, input efficiency scores (IES) are presented in Tables 1-3. It could be observed that there is a range of 7-25% for major states and a scope of nearly 10% for smaller states to improve their input efficiency relative to nearest independ- ently efficient states in 2011 for education sec- tor (Table 1). In case of health sector, this range is much higher for some years like 2004- 2005 (Table 2) and it has been 1-13% for major states and 6-30% for smaller states for the year 2010 (Table 3). Stochastic frontier method In the application of parametric techniques, stochastic methods can be used to correct for measurement and other random errors in the estimation of the production possibility fron- tier. In any parametric techniques a functional form is postulated for the production possibili- ty frontier, and then a set of parameters is selected that best fit the sample data. Results of our panel data estimation using frontier model for India (Males and females) are pre- sented in Table 4. It is observed that all the independent variables to explain LEXP have emerged with appropriate positive signs. Three of these variables, i.e. rural specialists (total specialists), auxiliary nurse midwife (ANM)/female health worker, and total number of blood banks are statistically significant. Discussion Results of our FDH analysis are suggestive of a considerably more scope for improvement in efficiency of public expenditure in health rela- tive to education. Further parametric approach of SFA indicates factors that could be isolated to suggest ways to improve efficiency in the pub- lic expenditure in the sector. As mentioned ear- Article Table 4. Stochastic frontier panel data model for India: life expectancy male and female (2005-2011). Variables Coefficient z M F M F Total specialists 0.004 0.004 1.83** 1.8** Auxiliary nurse midwife 0.014 0.017 2.12* 2.57*** Total no. blood bank 0.043 0.048 3.25*** 3.21*** Constant 3.929 3.942 52.360*** 46.21*** Mu 0.081 0.112 3.520*** 4.59*** Lnsigma2 -5.802 -5.546 -10.910*** -11.810*** Ilgtgamma 2.879 3.144 4.890*** 6.09*** Sigma2 0.003 0.004 - - Gamma 0.947 0.959 - - Sigma_U2 0.003 0.004 - - Sigma_V2 0.000 0.000 - - Time-invariant inefficiency model number of observation=105 per group (min=7). Wald chi2(3)=29.19 Log likelihood=275.66912; Prob>chi2=0.0000. We also tried the alternative model using random effects. However, the results of Hausman test indicated fixed effect model. *5% level of significance; **10% level of significance; ***1% level of significance. Table 5. Actual and estimated life expectancy for males and females in selected Indian States (2010). State Actual Potential Actual as % Ranks of States according LEXP LEXP of potential to realization of potential LEXP LEXP M F M F M F M F Andhra Pradesh 65.40 69.40 76.17 82.01 85.86 84.62 14 12 Assam 61.60 62.80 70.12 74.67 87.85 84.10 11 13 Bihar 67.10 66.70 70.24 74.84 95.52 89.13 4 9 Gujarat 67.20 71.00 72.33 77.33 92.90 91.82 6 5 Haryana 67.90 69.80 69.61 74.03 97.54 94.29 2 3 Karnataka 66.50 71.10 73.57 78.74 90.39 90.29 8 7 Kerala 72.00 76.80 73.10 78.13 98.49 98.29 1 1 Madhya Pradesh 62.50 63.30 72.91 78.04 85.72 81.11 15 14 Maharashtra 67.90 81.78 75.99 87.19 89.35 93.79 10 4 Odisha 62.30 64.80 70.99 75.65 87.76 85.65 12 11 Punjab 68.70 71.60 70.83 75.45 96.99 94.90 3 2 Rajashthan 66.10 69.20 72.06 77.00 91.73 89.87 7 8 Tamilnadu 67.60 70.60 75.17 80.70 89.92 87.49 9 10 Uttar Pradesh 64.00 64.40 74.48 79.95 85.93 80.55 13 15 West Bengal 68.20 70.90 72.34 77.37 94.28 91.64 5 6 LEXP, life expectancy. Non -co mmerc ial us e o nly [page 32] [Healthcare in Low-resource Settings 2014; 2:1866] lier, we hypothesize that States differ in their technical efficiency pertaining to health system due to factors which require emphasis in health facility planning. It is also hypothesized that these factors differ from State to State according to their level of development. It is presumed that estimated efficiency parameters should help the health policy makers to improve State level health system performance. As presented in the results above our findings indicate positive impact of governmental inter- vention in expansion of PHC facilities and the desirable impact of having rural specialists like surgeons, obstetrician and gynaecologists, physicians and paediatricians for enhancing life expectancy. The fact that the ANM has emerged with positive signs is indicative of the desirable role of the various inputs provided through paramedical manpower. Statistical sig- nificance of these inputs at the conventional level of significance and the variable of blood bank suggest that the system has indeed worked towards providing some of the desirable inputs. However, whether these have been utilised as efficiently as to be considered as optimum is revealed through our comparison of actual and estimated LEXP for males for the year 2010 in Table 5. These depict Kerala as the most efficient State with its actual LEXP being the highest in the estimated LEXP. This is fol- lowed by Punjab and Haryana. Further, the low- est efficiency for males is depicted by Madhya Pradesh followed by Andhra Pradesh and Uttar Pradesh. In case of Female life expectancy these rankings for the latter type (i.e., moving from lowest ranking state) are depicted by Uttar Pradesh followed by Madhya Pradesh and Assam (Table 5). Reasons for these inter-State disparities can be deciphered from major inputs for health sector in the States. Notably, the distributions of: per capita hospitals, PHCs, SCs, CHCs and beds in the States are highly inequitable. In fact, there is a considerable dif- ference between maximum and minimum val- ues for each of the parameters.72 Pertinently population served per government hospital bed is the highest (5606) in Bihar, followed by Assam (3912) and Uttar Pradesh (3499). Similar order holds true with regard to Population Served per govt hospital with high- est figure for Bihar (451325) followed by Uttar Pradesh (229118) and Assam (194863). The magnitude of the highest and the lowest Population Served Per Government hospital bed and hospitals in the States is ranked slight- ly different from order that of life expectancy and its achievements (i.e., actual vs potential life expectancy) in our results. However, obser- vations pertaining to other facilities like PHCs, SCs and CHCs depict higher numbers per thou- sand populations in Uttar Pradesh, which is in contrast to its lowest ranking of life expectancy outcomes thus depicting inadequate utilisation of these facilities. It is pertinent to note that Kerala does not have the highest number for any of the categories of these.72 In fact, in terms of manpower again Uttar Pradesh seems to have highest per thousand specialists at CHC (1.89), health assistants (4.52) and ANMs (22.46) and it has the second highest number for doctors at PHCs (2.86) and lady health visi- tor (2.04) in the country. This pattern also rein- forces the lower utilisation of manpower in the state. It points to the inadequate or ineffective utilization of staff inputs in poorly performing states. However, in most of the States, neither the inadequate availability of healthcare sector inputs nor merely inefficient utilization of these inputs explains the differentials in achievements in life expectancy. Besides the factors within the health system, as noted by us earlier, there are influences external to the sys- tem that may lead to differentials in efficiency at the State level. Some of these factors could be per capita income, per capita budgetary health expenditure, literacy, access to safe drinking water and urbanization. In general, the differential impact on life expectancy of health system inputs may be due to significant influence of some of these variables. It could be observed from the official publications that the majority of poorly performing States like Uttar Pradesh, Madhya Pradesh and Bihar are among the low income category States.73 Even the budgetary expenditure (as percent to total state budget) is lower in some of these States like Madhya Pradesh but this also holds for some of the relatively better off States like Punjab, Haryana and Maharashtara.73 Although Kerala does not have the highest figures in terms of either per capita income or budgetary expendi- ture on health, yet it has an outstanding posi- tion in terms of overall literacy which is 90.91percent as per the 2011 census.73 In con- trast, many of the poor and poorly performing States, in terms of life expectancy, have much lower levels of literacy. A similar situation pre- vails in terms of level of urbanisation in poorer states relative to their counterparts in better off states.73 Thus, in order to explore such external factors, we used dispersion in efficiency as a dependent variable in the second stage of our regression exercise using panel data for the state level. These are presented in the Appendix and Appendix Table A. The positive sign of per capita income indicates the impact of inequality in income across states influenc- ing the inequality in health outcomes towards greater disparities. The negative sign of gross enrolment indicates that an increased level of awareness about health related facilities and issues have helped to reduce regional disparity in efficiency of health system across states. However, this has not been able to compensate for other deficiencies of low investments and poor utilization of existing heath care facilities. Conclusions Results of our FDH analysis are suggestive of a considerably better scope for improvement in efficiency of public expenditure in health relative to education. Further parametric approach of SFA applied for health care sector indicates factors that could be isolated to sug- gest ways to improve efficiency in the public expenditure in the sector. The results of the frontier model, using panel data for 15 major Indian States in the years 2005-2011, indicate that the efficiency of public health delivery sys- tem remains low. Considerable disparities across States in terms of per capita availability and utilization of hospitals, beds and manpow- er inputs has had an adverse impact on improving the life expectancy in the poorer States. Overcoming these factoral disparities within the health system may lead to an improvement in the State level efficiency of the public health system. This may also help to improve life expectancy speedily and more equitably in the poorly performing States of Madhya Pradesh and Uttar Pradesh possibly as much as by 20%. However, this has to be sup- ported with other adequate infrastructure facilities like more budgetary expenditure to improve availability of medicines and materi- als at rural facilities and better management of health personnel in the rural areas to ensure their adequate utilisation. Learning from the remarkable achievements of Kerala, an emphasis on literacy by reducing dropout rates along with better utilization of health infra- structure and manpower resources could go a long way in improving life expectancy. This may require a considerable re-orientation of current healthcare set-up, particularly in the rural areas in the poorly performing States. These could reallocate surplus manpower from within and also make the rural infrastructure more useful to the needy through adequate inputs of building, equipment and medicines. In fact, there is a considerable differential in budgetary expenditure per capita between bet- ter off and poorer States. This in turn reduces the availability of basic medicines and materi- als in the public health system and reduces its reliability for the poor making them more dependent on the costlier private sector. Part of this problem could be tackled through funds from National Rural Health Mission and also by improving rural sanitation in poorer States. The results also suggest lack of appropriate links and coordination between economic and social sector policies leading to sub-optimal health outcomes for the poorer States in the country. Our results of SFA for 2005-2011 cor- roborate the analysis for earlier periods from other studies like Sankar and Vinish53 and Purohit.9 Article Non -co mmerc ial us e o nly [Healthcare in Low-resource Settings 2014; 2:1866] [page 33] References 1. Reserve Bank of India. State finances 2011. Mumbai: Bank of India Publ.; 2012. 2. Gupta S, Honjo K, Verhoeven M. The effi- ciency of government expenditure: experi- ences from Africa. Washington, DC: International Monetary Fund; 1997. 3. Isenman PJ. Basic needs: the case of Sri Lanka. World Dev 1980;8:237-58. 4. Sen AK. Public action and the quality of life in developing countries. Oxford B Econ Stat 1981;43:287-319. 5. Aturupane H, Glewwe P, Isenman PI. Poverty, human development and growth: an emerging consensus? Washington: The World Bank, Human Resources and Operations Policy Department Publ.; 1994. 6. Kakwani N. Performance in living stan- dards: an international comparison. J Dev Econ 1993;41:307-36. 7. Ke-Young C, Hemming R, eds. Public expenditure handbook: a guide to public expenditure policy issues in developing countries. Washington, DC: International Monetary Fund Publ.; 1991. 8. Ke-Young C, Gupta S, Clements B, et al. Unproductive public expenditures: a prag- matic approach to policy analysis. Washington: International Monetary Fund Publ.; 1995. 9. Purohit BC. Health care system in India. New Delhi: Gayatri Publ.; 2010. 10. Scott GC. Government reform in New Zealand. Washington: International Monetary Fund Publ.; 1996. 11. Oxley H, Maher M, Martins P, Nicoletti G. The public sector: issues for the 1990s. Paris: Organization for Economic Cooperation and Development Publ.; 1990. 12. OECD. Performance management in gov- ernment: performance measurement and results-oriented management. Paris: Organization for Economic Cooperation and Development Publ.; 1994. 13. Harbison RW, Hanushek EA. Educational performance of the poor: lessons from rural Northeast Brazil. Oxford: Oxford University Press; 1992. 14. Tanzi V, Schuknecht L. Reconsidering the fiscal role of government: the internation- al perspective. Am Econ Rev 1997;87:164-8. 15. Jimenez E, Lockheed ME. Public and pri- vate secondary education in developing countries: a comparative study. Washington, DC: World Bank; 1995. 16. Gerdtham VG, Jonsson B, MacFarlan M, Oxley H. New directions in health care pol- icy. Paris: Organization for Economic Cooperation and Development Publ.; 1995. 17. Evans DB, Tandon A, Murray CJL, Lauer JA. The comparative efficiency of national health systems in producing health: an analysis of 191 countries. Geneva: World Health Organization Publ.; 2001. 18. Wang J, Jamison DT, Bos E, et al. Measuring country performance on health: selected indicators for 115 coun- tries. Washington, DC: IBRD/World Bank; 1999. 19. St. Aubyn M. Evaluating efficiency in the Portuguese health and education sectors. Economia 2002;26:1-53. 20. Hofmarcher MM, Paterson I, Riedel M. Measuring hospital efficiency in Austria: a DEA approach. Health Care Manage Sci 2002;5:7-14. 21. Puig-Junoy J. Technical efficiency in the clinical management of critically ill patients. Health Econ 1998;7:263-77. 22. Gannon B. Technical efficiency of hospi- tals In Ireland. Dublin: Economic and Social Research Institute Publ.; 2004. 23. Magnussen J. Efficiency measurement and the operationalization of hospital pro- duction. Health Serv Res 1996;31:21-37. 24. Jeffrey P, Coppola MN. Efficiency of feder- al hospitals in the United States. J Med Syst 2004;28:411-22. 25. Bates LJ, Kankana M, Santerre RE. Market structure and technical efficiency in the hospital services industry: a DEA approach. Med Care Res Rev 2006;63:499- 524. 26. Kontodimopoulos N, Nanos P, Niakas D. Balancing efficiency of health services and equity of access in remote areas. Health Policy 2003;76:49-57. 27. Spinks J, Hollingsworth B. Health produc- tion and the socioeconomic determinants of health in OECD countries: the use of efficiency models. Clayton: Centre for Health Economics, Monash University; 2005. 28. Government of the Republic of Namibia. The technical efficiency of district hospi- tals in Namibia. Windhoek: Ministry of Health And Social Service; 2004. 29. Masiye F. Investigating health system per- formance: an application of data envelop- ment analysis to Zambian hospitals. BMC Health Serv Res 2007;7:58. 30. Mathiyazhgan MK. Cost efficiency of pub- lic and private hospitals: evidence from Karnataka State in India. Hon Sui Sen: Institute of South Asian Studies; 2006. 31. Mirmirani S. Health care efficiency in transition economies: an application of data envelopment analysis. Int Business Econ Res J 2008;7:47-56. 32. Kittelsen Sverre AC, Magnussen J. Testing DEA models of efficiency in Norwegian psychiatric outpatient clinics. Oslo: Health Economics Research programme at the University of Oslo 1998-2003; 1999. Available from: http://www.frisch.uio.no/english/publica- tions/?pubid=93 33. Li L, Wang J. Relative efficiency of the Chinese public acute hospitals: an empiri- cal data envelopment analysis application. Hong Kong: Department of Management, The Hong Kong Polytechnic University; 2008. 34. Hajialiafzali H, Moss JR, Mahmood MA. Efficiency measurement for hospitals owned by the Iranian social security organisation. J Med Syst 2007;31:166-72. 35. Suraratdechaac C, Albert A. Okunadeb AA. Measuring operational efficiency in a health care system: a case study from Thailand. Health Policy 2006;77:2-23. 36. WHO. The world health report, 2000. Health systems: improving performance. Geneva: World Health Organisation; 2000. 37. Murray CJL, Frenk J. A WHO framework for health system performance assess- ment, Global programme on evidence and information for policy. Washington, DC: World Bank; 1999. 38. Worthington AC. Frontier efficiency meas- urement in health care: a review of empir- ical techniques and selected applications. Med Care Res Rev 2004;61:135-70. 39. Jamison DT, Sandbu M, Wang J. Cross country variation in mortality decline, 1962-87: the role of country specific tech- nical progress. Geneva: Commission on Macroeconomics and Health, World Health Organization; 2001. 40. Salomon JA, Mathers CD, Murray CJL, Ferguson B. Method for life expectancy and healthy life expectancy uncertainty analysis. Geneva: Global Programme on Evidence for Health Policy, World Health Organization; 2001. 41. Schmacker ER, McKay NL. Factors affect- ing productive efficiency in primary care clinics. Health Serv Res 2008;21:60-70. 42. Evans DB, Tandon A, Murray CJL, Lauer JA. The comparative efficiency of national health systems in producing health: an analysis of 191 countries. Geneva: World Health Organization; 2000. 43. Greene W. Fixed and random effects in sto- chastic frontier models. New York, NY: Department of Economics, Stern School of Business, New York University; 2002. 44. Mehdi F, Filippini M, Lunati D. Economies of scale and efficiency measurement in Switzerland’s nursing homes. Paris: OECD; 2007. 45. Jian W, Zhao Z, Mahmood A. Relative effi- ciency, scale effect, and scope effect of public hospitals: evidence from Australia. Newcastle: University of Newcastle; 2006. 46. Knox KJ, Blankmeyer EC, Stutzman JR. Efficiency of nursing home chains and the implications of nonprofit status: a com- ment. J Real Estate Port Manage 2001;7:177-82. Article Non -co mmerc ial us e o nly [page 34] [Healthcare in Low-resource Settings 2014; 2:1866] 47. Knox KJ, Blankmeyer EC, Stutzman JR. Technical efficiency in Texas nursing facilities: a stochastic production frontier approach. J Econ Fin 2007;31:75-86. 48. Rosko MD. Cost efficiency of US hospitals: a stochastic frontier approach. Health Econ 2001;10:539-51. 49. Yong K, Harris A. Efficiency of hospitals in Victoria under casemix funding: a stochas- tic frontier approach. Clayton: Centre for Health Program Evaluation, Monash University; 1999. 50. Hollingsworth B, Wildman J. The efficien- cy of health production: re-estimating the WHO panel data using parametric and nonparametric approaches to provide additional information. Clayton: Centre for Health Programme Evaluation, Monash University; 2002. 51. Mortimer D, Peacock S. Hospital efficiency measurement: simple ratios vs. frontier methods. Clayton: Centre for Health Program Evaluation, Monash University; 2000. 52. Jayasuriya R, Wodon Q. Measuring and explaining country efficiency in improving health and education indicators. Washington, DC: World Bank; 2002. 53. Sankar D, Vinish K. Health system per- formance in rural India efficiency esti- mates across states. Econ Polit Weekly 2004;27:1427-33. 54. Purohit BC. Efficiency variation at sub- state level: the health care system in Karnataka. Econ Polit Weekly 2010;47:70- 6. 55. Purohit BC. Efficiency of health care sys- tem at sub-state level in Madhya Pradesh (India). Soc Work Public Health 2010;25:42-58. 56. Purohit BC. Efficiency of health care sec- tor at sub-state level in India: a case of Punjab. Online J Health All Sci 2010;8:2. 57. Fried HO, Lovell CA, Schmidt SS. the measurement of productive efficiency: techniques and applications. New York, NY: Oxford University Press; 1993. 58. Charnes A, Cooper WW, Lewin AY, Seiford LM. Data envelopment analysis: theory, methodology and applications. Boston: Kluwer; 1995. 59. Coelli T, Rao DSP, Battese G. An introduc- tion to efficiency and productivity analysis. Boston: Kluwer; 1998. 60. Hollingsworth B, Dawson PJ, Maniadakis N. Efficiency measurement of health care: a review of non-parametric methods and applications. Health Care Manage Sci 1999;2:161-72. 61. Deprins D, Simar L, Tulkens H. Measuring labor-efficiency in post offices. Amsterdam: North-Holland; 1984. 62. Tulkens H, Vanden Eeckaut P. Non-para- metric efficiency, progressmand regress measures for panel data: methodological aspects. Eur J Oper Res 1995;80:474-9. 63. Fakin B, de Crombrugghe A. Fiscal adjust- ment in transition economies: social transfers and the efficiency of public spending. Comparative analysis with OECD countries. Washington, DC: World Bank; 1999. Available from: elibrary.world- bank.org/doi/pdf/10.1596/1813-9450-1803 64. Pitt M, Lee M. The measurement and sources of technical inefficiency in the Indonesian weaving industry. J Dev Econ 1981;9:43-64. 65. Schmidt P, Sickles RC. Production fron- tiers and panel data. J Bus Econ Stat 1984;2:309-16. 66. Government of India. Health Information of India, 2005. New Delhi: CBHI, DGHS, MOHFW; 2006. 67. Government of India. Health Information of India, 2006. New Delhi: CBHI, DGHS, MOHFW; 2007. 68. Government of India. Health Information of India, 2007. New Delhi: CBHI, DGHS, MOHFW; 2008. 69. Government of India. Health Information of India, 2008. New Delhi: CBHI, DGHS, MOHFW; 2009. 70. Government of India. Health Information of India, 2009. New Delhi: CBHI, DGHS, MOHFW; 2010. 71. Government of India. Health Information of India, 2010. New Delhi: CBHI, DGHS, MOHFW; 2011. 72. Government of India. Health Information of India, 2011. New Delhi: CBHI, DGHS, MOHFW; 2012. 73. Government of India. Census of India, 2011. New Delhi: The Registrar General Office; 2012. Article Non -co mmerc ial us e o nly