68542 Policy Notes Unfinished Fiscal Reform Agenda in Indian States June 17,2008 Prepared by Poverty Reduction & Economic Management South Asia, The World Bank POLICY NOTES OX THE UNFINISHED FISCAL REFORM AGENDA IN INDIAN STATES TABLE OF CONTENTS ACKi~OWLEDGEMENTS 1 FOREWORD 2 EXECUTIVE SUMMARY 3 INTRODUCTION :3 FISCAL CORRECTION AND THE UNFINISHED AGENDA 7 L INTRODUCTION 7 II. SOURCES OF FISCAL CORRECTION 7 IlL MEASURING THE STRENGTH OF FISCAL CORRECTION ACROSS STATES 12 IV. LESSONS FOR STATES 15 V. LESSONS FOR THE CENTER 15 LESSONS FROM FISCAL REFOR.,'1 IN THREE STATES 16 ORISSA 16 UTTAR PRADESH 20 MAHARASHTRA 24 MEASURING SPENDING EFFICIENCY IN SELECTED SERVICES 27 ELEMENTARY EDUCATION 28 HEALTH 34 CONCLUSION 38 BIBLIOGRAPHY 41 ACKNOWLEDGEMENTS This report was prepared by a World Bank team led by V. J. Ravishankar and Farah Zahir under the overall guidance of Ijaz Nabi. The peer reviewers are Shahrokh Fardoust (IEGDG), Vikram Nehru (EASPR) and Mr. Senthil (Joint Secretary, Department of Expenditure, Ministry of Finance, and Gol). The task team includes Mohan Nagarajan, Rajni Khanna, Deepa Sankar, Paul Wade, Neha Kaul, Arpita Chakraborty, Monika Sharma, Jyoti Sriram and Sapna John. The background studies on efficiency of public spending were prepared by Indicus Analytics, Delhi and Indian Institute of Technology, Chennai. Dipak Dasgupta, Peter Berman and Preeti Kudesia provided helpful comments at various stages of the work. 1 FOREWORD The period 2000-07 witnessed the enactment of Fiscal Responsibility legislation by the center and vast majority of states, committing to medium-term targets for fiscal sustainability. The consolidated General Government (central and state) deficit, after having reached a peak of 9.6 percent of GDP in the late 1990s, came down significantly to 6.5 percent in 2006-07. At the state level, the improvement in aggregate hides wide disparities between the states. Some state governments are already on a sustainable path while some others are far from it. Some have complied with their Fiscal Responsibility acts both in letter and spirit, while others have accumulated off-budget liabilities to unsustainable levels, circumventing the FR Acts, which did not specify caps on such liabilities. Both central and state tinances face new risks today, including the impact of (i) oil and food price inflation and consequent subsidy increase, (ii) civil service pay hike following the 6th Pay Commission award, and (iii) potential slow down in economic growth. The degree of risk also varies between the states, given that the strength and quality of fiscal correction since 2000 has been widely divergent, apart from varying economic growth rates. The set of three Policy Notes on the unfinished fiscal reform agenda at the state level, produced during FY07 and FY08, are focused on how different states have performed, where they stand today, and in defining the contours of the unfinished agenda that lies ahead for the state governments. The notes also have messages for the fiscal federal authorities and federal development agencies, such as the Finance Commission and the Planning Commission. Note I, which was delivered as an intermediate product in September 2007, examines progress with overall fiscal correction by states during 2000-06. Note II presents case studies of fiscal adjustment experience in three selected states. Note III illustrates, using studies focused on education and health spending, how public expenditure efficiency could be measured so as to address the challenge of translating outlays to outcomes. The case studies show the difficulty of assessing the quality of fiscal performance based on budgetary statistics alone. For instance, the UP case shows that increased capital spending could be accompanied by reduced efficiency of such spending due to corruption and monopolistic practices in the public procurement process. On the contrary, the Orissa case illustrates how increased efficiency, in terms of speedier completion of investment projects, could accompany, or indeed be motivated by, a squeeze on the capital budget due to overall resource constraints. A specific difficulty in trying to use sector-wise composition of spending through the state budget as an indicator of 'quality' of fiscal performance is that the readily available data provide an incomplete picture. There is significant development spending taking place outside of the state budget, through direct transfers from center to the district level, associated with several co-fmanced flagship Centrally Sponsored Schemes. Unless these are integrated with the state budget statistics, one could reach misleading conclusions. For instance, a state government in acute fiscal stress may deliberately follow a policy of reducing its spending on budget while trying to maximize utilization of central funds available off-bUdget, for essential services such as elementary education, health and anti-poverty programs (some evidence of such behavior is presented in the Orissa case study in Note II). For the reasons explained above, the methodology used in Note I to measure the strength of fiscal performance across states deliberately avoids using any indicator of' expenditure composition', and limits itself to the degree and sources of deficit reduction, using the latter as a proxy indicator of the quality of adjustment. Note III probes deeper into quality issues, illustrating an efficiency frontier approach in two selected sectors, viz., elementary education and health services. 2 EXECUTIVE SUMMARY INTRODUCTION 1. Both central and state governments have carried out significant correction of financial imbalances since 2000, through the implementation of fiscal reforms including enactment of Fiscal Responsibility acts. However, the improvement in the finances of all states in aggregate hides wide disparities between them. Some state governments are already on a sustainable path while some others are far from it. Some have complied with their Fiscal Responsibility acts both in letter and spirit, while others have accumulated off-budget liabilities to unsustainable levels, circumventing the FR Acts, which did not specify caps on such liabilities. ii. Moreover, fiscal correction by an Indian state is only a means to an end. The end that is, achievement of development goals - depends on how fiscal space is used, or how efficiently the money is spent to produce outputs and the desired outcomes. A strong fiscal performance would lead both to an improvement in the overall balance via improved structure of revenue and expenditures, with incentives in place to ensure sustained fiscal improvement, and better outcomes associated with budget outlays. iii. Note I examines progress with overall fiscal correction by Indian states during 2000-06. Note II presents case studies of fiscal adjustment experience in three selected states. Note III illustrates how expenditure efficiency could be measured so as to address the challenge of translating outlays to outcomes. iv. The case studies show the difficulty of assessing the quality of fiscal performance based on budgetary statistics alone. Increased capital spending in one case is accompanied by reduced efficiency of such spending; and on the contrary, increased efficiency in another case is prompted by a squeeze on the capital budget due to overall resource constraints. In addition, sector-wise composition of spending through the state budget provides an incomplete picture due to significant development spending taking place outside of the state budget, through direct transfers from center to the district leveL For these reasons, the methodology used in Note I to measure the strength of fiscal performance across all the major states deliberately avoids using any indicator of'expenditure composition'. It limits itself to the degree and sources of deficit reduction, using the degree to which it is based on revenue enhancement as a proxy indicator of the quality of adjustment. Note III probes deeper into quality issues, illustrating an efficiency frontier approach to measuring expenditure efficiency in two selected sectors, viz, elementary education and health services. v. The main focus in Note I is on progress towards fiscal sustainability. Following the criteria laid out by the Eleventh and Twelfth Finance Commissions, fiscal sustainability has been defined in terms of recommended limits on debt and debt servicing burden relative to revenue, rather than in the standard way it is defined for sovereign nations. The underlying logic is that sustainability for a sub-national government in India is mainly concerned with the adequacy of fiscal space for non-interest expenditures on a wide range of social and economic services for which states are responsible, according to the Constitution of India. It is less about macroeconomic stability and more about containing debt servicing within a tolerable level, so as to have fiscal space for service delivery. This tolerable level is defined in line with the Finance Commission norms; the Twelfth Finance Commission defined interest above 20 percent of revenue as being a condition of 'debt stress', and the same ratio at 15 percent or below as the desired medium-term target for every state. vi. Both central and state finances face new risks today (as of June 2008), including the impact of (i) oil and food price inflation and consequent subsidy increase, (ii) civil service pay hike following 3 the 6 th Pay Commission award, and (iii) potential slow down in economic growth. The degree o f , . risk also varies between the states, given that the strength and quality of fiscal correction since 2000 has been widely divergent. These factors add to the urgency of addressing the unfinished fiscal reform agenda at the state level, in addition to the unfinished agenda at the central level (which is outside the scope of these notes). NOTE I: FISCAL CORRECTION AND THE UNFINISHED AGENDA vii. More than half the 14 major states of India have achieved significant fiscal deficit reduction during 2000-06, with Karnataka, Tamil Nadu, Orissa and Haryana in the lead. However, some high and middle income states including Maharashtra and Gujarat lag behind, continuing to accumulate unsustainably high levels of debt and guarantees. viii.The majority of states have relied more on enhancing their own revenues than on contracting expenditures. A largely revenue based fiscal correction is welcome in the context of the squeeze on infrastructure and social service expenditures that Indian states had experienced over the 1990s. lX. Those that have committed to fiscal responsibility and have achieved significant correction now face the subsequent challenge of translating outlays to outcomes. Those that lag behind must first address the challenge of fiscal discipline, learning from the success of others. In order to induce high-income states to contribute more to fiscal correction, the center would need to move faster towards market based discipline, i.e., requiring states to access the domestic capital market on the strength of their own creditworthiness, without any explicit or implicit central guarantee. x. Going forward, it would be advisable to target end-of-year stock of outstanding debt and guarantees, in addition to current account balance (Golden Rule); monitoring of end-of-year stock rather than annual flows would enhance year-to-year flexibility in states' fiscal management, while remaining within an agreed correction path. NOTE II: LESSONS FROM FISCAL REFORM IN THREE STATES x.l. In all three cases studied - Orissa, Uttar Pradesh and Maharashtra - fiscal correction was prompted by a crisis situation in the state's finances, but the timing and nature of the crisis were different. xii. In Orissa and Uttar Pradesh, the crisis reached its peak at the turn of the century, while it happened three years later in Maharashtra. In the two poor states, the crisis resulted in serious liquidity problems, and the states responded to fiscal incentives from the center in the form of additional grants and/or debt relieflinked to deficit reduction. In the high income state, the crisis appeared in the form of a default in servicing guaranteed debt and consequent downgrading by the credit rating agencies. xiii. Fiscal correction was more lasting and of better quality in Orissa than in the other two states. While deficit on budget was reduced significantly in UP, quasi fiscal risks from off-budget power sector losses and liabilities remains high. Both in UP and Maharashtra the fiscal correction was mainly or completely revenue based. The correction in Orissa was revenue based in the first phase, but also embraced expenditure containment in the second phase. The re-election of the incumbent government in 2004 emboldened it to implement significant attrition based downsizing of the state government establishment. XlV. The UP story shows that improvement in expenditure composition is necessary but not sufficient, as the quality of public spending and what is delivers also depends on the institutional and governance environment. Fiscal space may increase without a proportionate increase in the quantity and quality of goods and services delivered. Even in Orissa, whose fiscal correction was 4 stronger than the other two cases, there has only been limited improvement so far in the efficiency and quality of public spending. Rate of completion of investment projects has improved, and so has the utilization of central assistance to some extent. However, government provided services still suffer from staff absenteeism, lack of access in many villages and hamlets, poor learning and health outcomes, as well as delivery gaps including leakages. This suggests that translating outlays to outcomes is part of the unfinished agenda in almost all states, even those that have achieved a strong fiscal correction. xv. The Maharashtra story highlights the problem of state guaranteed debt, and that not all states have adopted guarantee caps in theirFRBM laws. Neither the 11 th and 12th Finance Commissions nor the central Ministry of Finance has included guarantees in defining the eligibility criteria for states to receive performance based fiscal incentives, in the form of additional grants and/or debt relief. State governments are not even mandated to ensure full disclosure of information on outstanding guarantees and all off-budget liabilities. The absence of mandatory caps and information disclosure on guarantees is a serious gap in the framework of fiscal responsibility a problem that could be addressed by the 13 th Finance Commission, expected to make its recommendations in 2009. NOTE III: MEASURING SPENDING EFFICIENCY IN SELECTED SERVICES xvi. Translating outlays to outcomes is easier said than done. It involves a transition in the entire mindset of the government machinery, which is used to maximizing the size of the total or 'plan' budget, rather than on maximizing results or outcomes. Financial accountability has traditionally been confined to measuring and monitoring inputs and budget compliance that is, whether allocated monies were spent on the inputs as targeted in the budget. The first step towards shifting the focus to results achieved on the ground is to begin measuring outputs in addition to inputs, and attempting to also measure outcomes, which poses a very big challenge. xvii. One of the main reasons why measurement itself is an enormous challenge is that it cannot be carried out at a macro level, by the Finance Ministry or Department alone. What can be measured based on financial information is the composition of spending, across services and across different economic categories such as wages, non-wage operation and maintenance, subsidies and transfers, capital investment and lending. These are at best proxy indicators of spending efficiency. For example, a rise in non-wage to wage expenditure within the current expenditure on roads is an indication that spending efficiency has probably risen, because there is a larger budget to execute per person employed. But the margin of error in drawing such a conclusion is serious, because appropriate input mix is only a necessary but not sufficient condition for ensuring value for money. The degree of competitiveness in works procurement, and the degree of corruption, also matter. These cannot be accounted for without measuring the quantity and quality of roads maintained per rupee spent. xviii. Outputs and outcomes of different services and departments cannot be measured using a common standard or yardstick. The units and even the methodology to be used for measuring what schools are delivering cannot be the same as those for measuring assets built and maintained by the public works department. xix. Note III presents two case studies, pertaining to two selected services: elementary education and health. It focuses on the state level, given that state governments have primary responsibility for delivery of these services. Both services involve multiple inputs, as well as difficulties in going beyond outputs to measure outcomes. In the case of health there are also multiple outputs and multiple outcomes which the same facilities and staff are supposed to deliver. The case studies illustrate one specific methodology for measuring production efficiency in such circumstances, called Data Envelopment Analysis (DEA). The methodology is explained in Annex A. 5 xx. Given that the DEA methodology is data intensive and there are several gaps in the availability and reliability of data on outputs and outcomes in education and health sectors, the results are to be treated with caution. They are illustrative and not conclusive. The objective ofthis Note is to initiate discussion on how this method could be used, along with other complementary analysis, to provide a robust diagnosis fit to be used for policy decisions. Note 1 June 2008 POLICY NOTE I 1 FISCAL CORRECTION AND THE UNFINISHED AGENDA I. INTRODUCTION 1. In its positive assessmenf published recently, the Reserve Bank of India has highlighted that the fiscal deficie of all states, taken together, declined from 4.7% of GOP in 1999/00 to an estimated 3.2% in 2005/06 4 • Their deficit on current account, called Revenue Oeficie, declined from 2.7% to 0.5% of GOP. Close examination of state specific perfonnance reveals that poorer and fiscally more dependent states have, in general, achieved stronger fiscal correction than the high income states. Traditionally strong and leading states such as Maharashtra, Gujarat and Punjab, have lagged behind in fiscal correction. Even after enacting Fiscal Responsibility legislation, these states have continued to accumulate unsustainably high levels of debt and guarantees, using the lacuna in the legislation, viz., the absence of any cap or target for off-budget liabilities such as state guaranteed debt. 2. Fiscal perfonnance by Indian states during 2000-06 is to be seen in the context of the historical backdrop of the 1990s, a decade characterized by serious deterioration in state finances, and ending with a fiscal crisis6 . At the tum of the century, Indian states were experi~mcing unsustainable debt trends and squeeze on resources available for essential infrastructure and social services for which they have primary responsibility under the Constitution of India including policing, public health, school education, road connectivity, irrigation and drinking water supply. The fiscal improvement during 2000-06 is a result of three underlying factors, namely: (i) fiscal correction efforts by the majority of states, (ii) rise in the share of central resources, especially to the poorer states, resulting from the awards of the Eleventh and Twelfth Finance Commissions; and (iii) acceleration of economic growth in India since 2003/04. 3. This Note pays attention to three dimensions, or criteria, of fiscal correction at the state level: (a) the Golden Rule, or achievement and sustenance of current account balance; (b) debt sustainability, to be pursued by achieving and maintaining a primary fiscal surplus7 and by controlling guarantees; and (c) state's own revenue effort, which is one indicator of the quality of fiscal adjustment. A state that does weakly in its revenue effort, relying mainly on expenditure cuts to reduce its deficit, is to be judged as having perfonned poorer than one that does strongly on own tax effort and reduces its deficit mainly on the revenue side, or through a combination of revenue and expenditure measures. 4. Section II examines the sources of correction in the primary fiscal deficit. Section III explains each of the three criteria and lays out the combined measure of the strength of fiscal correction by states during 2000-06. Section IV summarizes some key lessons for the states, going forward. Section V presents the main lessons for the central authorities. II. SOURCES OF FISCAL CORRECTION 5. The primary (non-interest) fiscal deficit of states has declined during 2000-06 largely on account of (i) increase in central resource transfers, and (ii) states' own correction. The states' own effort can in I For details see paper entitled "Indian States' Fiscal Correction An Unfinished Agenda"; V. J. Ravishankar, Farah Zahir & Neha Kaul, Economic and Political Weekly, Mumbai (forthcoming) 2 Reserve Bank ofindia. (2006) 'State Finances study of Budgets 2006-07'. Mumbai 3 Fiscal Balance=Revenue Receipts-Total Expenditure(Revenue + Capital) 4 Throughout this paper correction during 2000-06 has been calculated by comparing outcomes in 2005/06 with the average during 1998/99 and 1999/00, the period when deficits peaked in all the major states. 5 Revenue Balance =Revenue Receipts« Revenue Expenditure). 6 World Bank (2005) 'State Fiscal Reforms in India Progress & Prospects', Macmillan Press. 7 Primary Balance= Revenue Receipts - Non-Interest Expenditure 7 .tVo/1.!. j June 2008 tum be further decomposed into (a) increase in own revenues, and (b) contraction in non-interest expenditure. 6. The majority of states have relied mainly on enhancing their own tax revenues to reduce deficits, rather than contracting expenditures (Figure A). The largest reduction in primary deficit during 1999­ 2006 has been achieved in Orissa, with significantly larger state's own effort as compared to enhanced central resources. Orissa also stands out as the only state that has gained significantly from both revenue enhancement and expenditure contraction, in addition to enhanced central transfers. Figure A: Most states have achieved revenue based correction Average annual change in % GSDP 2.00 A / 1.50 n I 100 Ii 0.50 -1.00 -1.50 L .-.---.-.-~ .. -.~------ • . . . . . ---.---, c:::=::J Central Resources _ States ON n Resources _ Contraction of !Ibn-Interest expenditure - -.. A-imary Correction 7. A state like Bihar, on the other hand, shows moderate improvement in the primary deficit despite a huge increase in central transfers. This is because it showed a significant increase in non-interest expenditure. Along with Gujarat it is the only state to have registered a decline in own resources over this period. Karnataka shows the highest increase in own revenues among all states but it is also the only state to show a decline in central transfers (Fig. B). This method of fiscal adjustment through fiscal empowerment (larger dependence on own resources) adopted by Karnataka is the most sustainable form of correction as it reduces any vulnerabilities arising from high dependence on central resources. Own Revenue growth(2000-06) 1998100 2005/06 1" ~ 14 Q oJ en oJ ,.I 12 - . ..'" .. .. (!! ... i... 10 8 C ~ . 15,000 crore, including wage arrears. While the FRBM Act requires strengthening, including adoption of a mandatory ceiling on guarantees, such a legislative amendment is unlikely to be pursued prior to the elections. There is also the risk that subsidies may rise in this election year, 'Four lessons from three case studies 67. In all three cases, fiscal correction was prompted by a crisis situation in the state's finances, but the timing and nature of the crisis were different. In Orissa and Uttar Pradesh, the crisis reached its peak at the turn of the century, while it happened three years later in Maharashtra. In the two poor states, the crisis resulted in serious liquidity problems, and the states responded to fiscal incentives from the center in the form of additional grants and/or debt relief linked to deficit reduction. In the high income state, the crisis appeared in the form of a default in servicing guaranteed debt and consequent downgrading by the credit rating agencies. This suggests that poor andfiscally more dependent states are more likely to respond to central fiscal incentives while less dependent high income states respond more to market perception offiscal risk. This is lesson number one. 68. Fiscal correction was more lasting and of better quality in Orissa than in the other two states. Both in UP and Maharashtra the fiscal correction was mainly or completely revenue based. The correction in Orissa was revenue based in the first phase, but also embraced expenditure containment in the second phase, including significant attrition based downsizing of the state government establishment. An important development that enabled Orissa to sustain and strengthen the fiscal reform program was the re-election of the incumbent party and coalition in 2004. This suggests that political stability is an enabling factor for strong and sustained fiscal reform. This is lesson number two. 69. The UP story shows that improvement in expenditure composition is necessary but not sufficient, as the quality of public spending and what it delivers also depends on the institutional and governance environment. Fiscal space may increase without a proportionate increase in the quantity and quality of goods and services delivered. Even in Orissa, whose fiscal correction was stronger than the other two cases, there has only been limited improvement so far in the efficiency and quality of public spending. Rate of completion of investment projects has improved, and so has the utilization of central assistance to some extent. However, government provided services still suffer from staff absenteeism, lack of access in many villages and hamlets, poor learning and health outcomes, as well as delivery gaps including leakages. This suggests that translating outlays to outcomes is part ofthe unfmished agenda in almost all states, even those that have achieved a strong fiscal correction. This is lesson number three. 70. The Maharashtra story highlights the problem of state guaranteed debt, and that not all states have adopted guarantee caps in their FRBM laws. Neither the 11 th and 12th Finance Commissions nor the central Ministry of Finance has included guarantees in defming the eligibility criteria for states to receive performance based fiscal incentives, in the form of additional grants and/or debt relief. State governments are not even mandated to ensure full disclosure of information on outstanding guarantees and all off-budget liabilities. The absence ofmandatory caps and information disclosure on guarantees is a serious gap in the framework offIScal responsibility - a problem that could be addressed by the 13th Finance Commission, expected to make its recommendations in 2009. This is lesson number four. 26 Note III June 2008 POLICY NOTE III 29 MEASURING SPENDING EFFICIENCY IN SELECTED SERVICES Introduction 71. At the national level as well as in several states where fiscal pressures eased and budget space increased for developmental spending in recent years, especially during 2004-07, there has been increasing talk oftranslating outlays to outcomes. Fiscal adjustment or correction, after all, is only a means to an end. The end that is, achievement of development goals depends on how fiscal space is used, or how efficiently the money is spent to produce outputs and what is the final outcome of public spending. 72. Translating outlays to outcomes is easier said than done. It involves a transition in the entire mindset of the government machinery, which is used to maximizing the size of the total or 'plan' budget, rather than on maximizing results or outcomes. Financial accountability has traditionally been confined to measuring and monitoring inputs and budget compliance that is, whether allocated monies were spent on the inputs as targeted in the budget. The first step towards shifting the focus to results achieved on the ground is to begin measuring outputs in addition to inputs, and attempting to also measure outcomes, which poses a very big challenge. 73. One of the main reasons why measurement itself is an enormous challenge30 is that it cannot be carried out at a macro level, by the Finance Ministry or Department alone. What can be measured based on [mancial information is the composition of spending, across services and across different economic categories such as wages, non-wage operation and maintenance, subsidies and transfers, capital investment and lending. These are at best proxy indicators of spending efficiency. For example, a rise in non-wage to wage expenditure within the current expenditure on roads is an indication that spending efficiency has probably risen, because there is a larger budget to execute per person employed. But the margin of error in drawing such a conclusion is serious, because appropriate input mix is only a necessary but not sufficient condition for ensuring value for money. The degree of competitiveness in works procurement, and the degree of corruption, also matter. These cannot be accounted for without measuring the quantity and quality of roads maintained per rupee spent. 74. Outputs and outcomes of different services and departments cannot be measured using a common standard or yardstick. The units and even the methodology to be used for measuring what schools are delivering cannot be the same as those for measuring assets built and maintained by the public works department. The devil is in the details. 75. This Note presents two case studies, pertaining to two selected services: elementary education and health. It focuses on the state level, given that state governments have primary responsibility for delivery of these services. Both services involve multiple inputs, as well as difficulties in going beyond outputs to measure outcomes. In the case of health there are also multiple outputs and multiple outcomes which the same facilities and staff are supposed to deliver. The case studies illustrate one specific 29Based on the following background papers: lndicus Analytics Report (2007) "Efficiency in Public Spending Data Envelopment and Econometric Analysis" lIT Chennai Project Report (2008) "Output and Outcome Efficiency of Health Expenditure" Wade, Paul (2008) "Efficiency of Public Education Spending in Indian States", World Bank (mimeo). 30In social sectors, there is a time lag in translating inputs into outcomes. Different states /districtslsub-districts have different starting points of inputs and hence at a point of time, the funding composition of different states/districts may show these disparities. 27 Note III June 2008 methodology for measuring production efficiency in such circumstances, called Data Envelopment Analysis (DEA). The methodology is explained in Annex A. 76. Given that the DEA is a non parametric approach, it has no way of accounting for random noise, which is one limitation of this methodology. The results are also highly sensitive to the quality of data. Given that there are several gaps in the availability and reliability of data on outputs and outcomes in education and health sectors, which are noted in each case below, the results are to be treated with caution. They are illustrative and not conclusive. The objective of this Note is to initiate discussion on how this method could be used, along with other complementary analysis, to provide a robust diagnosis suitable for feeding into policy decisions. ELEMENTARY EDUCATION 77. This section presents an attempt to develop a comparative measure of the efficiency of public expenditure on elementary education across the states of India. It is based on input and output data that cover both public and private schools; only unrecognized private schools are excluded. This is not too serious a limitation because unrecognized schools generally account for less than 5 percent of enrollment at elementary level (with the exception of Punjab and Haryana where the ratio is estimated to be much higher, and in whose cases the efficiency scores are probably over--estimates). The more serious limitation is that the 'efficiency' scores do not correct for locational disadvantages such as hilly terrain or popUlation density, whiCh issue is discussed towards the end of this section (See para 82). Table 3.1 Status of Elementary Education: Some key indicators AVG ~~ Average; MAX = Maximum among among <~'-"----------r;::;;- i Zvalue and i Coeffident l L ! significance ~ndom Effect ! Monthly per capita expenditure . (1.75)* t ,00027 I i I : Population Density ><----~--"--~-"--·----i--~-(3-~74)*** --+- ------- .00049 ~ ! Simplicity of rules and procedures (Index) L ,---1- (0.96) ----t-'-" ,I ,10248 ~ I 1 Corruption Index (2.58)***: 19510 - I ! ------~____" _ _ _ _ _ I --11---,-:--::-,--,----11 I i YearDummy2 1 (-2.62)*** -.15331; IConstant --------- (-1.52) -.59246 ,.~ i L_____ _._ _ _,__ ___. ~. _ _ _ _..L... _______----1 I Data on learning outcomes based on PRATHAM survey are available only at one point of time. '1 34The regression employs the random effects model, with 50 observations of 25 states at two time points - The Hausman test returns a Prob>chi2 0.7356, which indicates the random effects model to be a relatively efficient model under the circumstances. 32 Note JIl June 2008 * Significant at 10%; ** significant at 5%; ** * significant at 1% 85. Efficiency scores can in general be affected by factors outside the control of schools and the education departments - such as population density, income level in the district, degree of corruption in the state, etc. The results of a regression analysis, carried out to separate the effects of such factors are presented in Table 3.4. The effects of population density and corruption show up as the most statistically significant. 86. After separating out the effect of the explanatory factors mentioned above, the residual can be normalized and be interpreted as "implementation efficiency" of the public elementary school system. The estimated values of implementation efficiency, calculated in this way, are shown in Table 3.5. T a bIe 35 C omparlson 0fImplemen t ation Em' - Itlenty, 200405 Implementation I r Efficiency Efficiency I State (Actual) (predicted) Residual Efficiency i Andhra Pradesh 1.00 0.75 0.25 I 1.60 l°rissa I 0.87 0.67 0.20 1.31 i Karnataka 0.91 0.75 0.16 1.02 t Tamil Nadu i 1.00 0.84 0.16 1.01 ; Assam I 0.73 0.58 0.15 0.98 Maharashtra 0.89 0.78 0.11 0.74 ! Manipur 0.73 0.62 0.1-1 0.72 Madhya Pradesh 0.68 0.57 0.11 0.69 Arunachal Pradesh 0.70 0.60 0.10~ Gujarat 0.88 0.81 0.07 0.49 Mizoram 0.76 0.68 0.07j . 0.46 ! Himachal Pradesh 0.79 0.73J 0.05 0.35 I West Bengal 1.00 0.96 0.04 0.27 I Kerala 1.00 1.00 : 0.00 -0.03 I Punjab 0.84 0.85 -0.01 : -0.09 I Bihar 0.88 0.91 -0.03 -0.17 I I Haryana Janunu & Kashmir I 0.76 0.70 0.84 0.79 -0.08 I -0.09 -0.55 ~ -0.60 , ~ I Naga1and 0.66 0.76 -0.10 -0.67 I Goa : Rajasthan 0.72 0.57 0.83 0.68 -0.11 -0.11I ~71 -0.73 . I Tripura : Uttar Pradesh 0.54 0.57 0.69 1 0.84 -0.15 I -0.27 -0.97 -1.78 : Sikkim 0.35 0.64 -0.29 -1.90I . Meghalaya I 0.33 0.65 -0.32 -2.06 . 87. Based on these estimates, Andhra Pradesh and Orissa come out on top in terms of implementation efficiency, followed by Tamil Nadu and Karnataka. The most inefficient in implementation, among the major states, include Uttar Pradesh, Rajasthan, Bihar and Haryana. 88. As mentioned at the outset, the analysis presented here is meant only to be an illustration of how spending efficiency may be measured, and the kind of analysis that can feed into policy decisions. The results themselves are not robust enough at this stage to fonn the basis for policy decisions. This is 33 NOle IIi June 2008 because of some serious weaknesses in the database, especially the lack of any reliable comparable data on private or NGO run unaided schools, recognized and unrecognized. HEALTH 89. This section presents the results of applying the DEA methodology to district public health systems in two states: Tamil Nadu and Orissa. It illustrates how DEA can help analyze the utilization of existing resources and in rationalizing them to improve spending efficiency. 90. The analysis uses data on inputs and outputs of public health facilities including primary health centers (PHCs), community health centers (CHCs), taluklnon-taluk hospitals, district hospitals and medical college hospitals. in each of the 29 districts of Tamil Nadu and 30 districts of Orissa. 91. Before presenting the DEA results, it is useful and necessary to examine some general features of the health systems and health seeking behavior in the two states, so that the results can be interpreted properly. Data from the National Sample Survey (60 th Round, 2004) reveal the following key features: .:. In Tamil Nadu, nearly 35% of rural households and 25% of urban households that sought health care used public facilities, while in Orissa the corresponding ratios were 62% of rural and 58% of urban households (Table 3.6). The remaining used private facilities . •:. Among those who did not seek care from government providers, more than 25% were "not satisfied with the treatment" at public facilities. In Orissa, lack of access was also cited as a major reason (Figure 3.3) . •:. In Tamil Nadu, out of pocket expenditure per visit for outpatient care in a public facility was Rs. 43 on average, while it was Rs.284 in a private facility. In Orissa, on the other hand, the figures were Rs.262 and Rs.236 respectively - that is, visiting a public facility was costlier than visiting a private facility . •:. The overall loss of income per episode in Orissa was much greater (nearly 6 times) than that in Tamil Nadu . •:. For inpatient care (hospitalization), average out of pocket expenditure in private facilities was Rs. 11,130 in Tamil Nadir, compared to Rs. 928 in the case of public facilities. In Orissa, these figures were Rs.8342 and Rs.3311 respectively . •:. In Tamil Nadu, 74% of those in the poorest quartile who got hospitalized spent at least 10% of their total annual household expenditure per hospitalization episode; and 57% of them spent at least 25% of their annual household expenditure. In Orissa, more than 50% of in-patients in most quartiles spent at least 10% of their total annual household expenditure. Table 3.6 Health seeking behavior in Tamil Nadu and Orissa -.-~- ~:::~::'I Totru i ~=I um"~Totru i i Self reported spells ofail~ent' 11182 113':n 125231 I 776TT82 19581 ~-.. ! Sought care on medical advice I I 966 I'~ ,1132 I I 2098 I!i 597 I 153 I I ~ 1 750 I rsoughtcarefrOmpu~acili~l~_~~.l~~ ~_181 I 412 I th Source; NSSO 60 Round 34 Note III June 2008 Figure 3.3 Reasons for not seeking treatment Reason for not seeking Public Facility 45.00 40.00 35.00 30.00 ~ -- 25.00 DTN ! 20.00 .Oris~ 15.00 10.00 5.00 0.00 IV II) c: 8. j II) o ~ 0 z 92. The NSS data reveal that private health care is more developed in Tamil Nadu. A greater proportion of households visit private facilities than public, and spend more per episode in private as compared to public, unlike the case of Orissa. Given that the availability of public facilities is driven by standard population norms for all states, it can be expected that under-utilization of public facilities is a bigger problem in Tamil Nadu than in Orissa. Given the low population density in the interior districts of Orissa, it is not surprising that access (distance to nearest facility) is a bigger problem in that state (Fig 3.3). Rural patients in Orissa, irrespective of the type of facility visited, incur greater out of pocket expenses than do their urban counterparts. 93. Out of pocket health expenditures by poor households due to hospitalization is catastrophic in nature in both Tamil Nadu and Orissa, with majority of households losing more than 10 percent of annual purchasing power per episode. This is indeed a serious public policy issue which deserves immediate attention. An apparently obvious conclusion is that there is need for the government to spend more in the financing of health care. However, public financing need not necessarily mean public provision. Before stepping up resources provided through public facilities, it is essential to examine how efficiently the resources already available with the public facilities are being spent. This is where the DEA methodology helps, even though it uses public facilities in the best performing districts as the benchmark, rather than private facilities as the benchmark (which is ruled out due to lack of detailed and comparable data on private facilities). 94. The analysis uses three inputs and five outputs. Hospitals systems provide three major services: outpatient services, in-patient and Reproductive and Child Health (RCH) services. Given this homogeneity in types of services provided, and the number of cases treatedlhandled under each category, six outputs were selected. These are: number of inpatients treated (IP), number of out patients treated (OPD), number of sterilizations done (STR), number B.C.G, TT and Measles given. The inputs considered are number of beds (BED), number of nurses (NUR), and number of Doctors employed (DOC). 35 Note III June 2008 95. District-wise technical and scale efficiency scores for the public health facilities in Tamil Nadu are presented in Table 3.7. Of the 29 districts, 16 (55%) were technically efficient as they had relative efficiency score of l.00; they lie on the efficiency frontier. The remaining 13 (45%) districts were technically inefficient with scores ranging from 0.66 (Ramanathpuram) to 0.98 (Dindigul). The average technical efficiency score among the inefficient districts was 85% (Standard Deviation = 0.09), which means that in these districts public resources spent on health inputs could be reduced by 15 percent without any loss of output. Table 3.7 Efficiency of public health facilities in TN, 2006-07 93.79% 85.22% .?.:?_:g~.;_. 100.00% ',...... ............•..... ..." .... ......•............ ...... 100.00% ., .,."., ; " , .. " ...." .... 77.46% • ." ..."." .................................." " ... ., ...,,. 65.52% ; 99.01% • 99.15% ' ".,j. +... "........"...~.~:Q. ?~~., ."._.""""...."...."" .• . 100.00% : i 22 I L~} . . ! 24 I" : 25 Tirunelveli 1-·" .~.~ •• "' _".~ •• _ . _ . , , _.... .'. ".. 1-'" , 27 Vellore [~i"'" , 29 96. Decreasing returns or diseconomy of scale indicates that the concerned facility is too large for the volume of activities it conducts. In contrast, a facility with increasing returns or economy of scale indicates that it is too small for its scale of operation. Similar implications could be drawn for district health systems. Twelve of the districts (41 %) in Tamil Nadu had Scale Efficiency of 100%, which implies that they had the most productive size for that particular input-output mix. The other 59% of the 35 CRS ~" Constant Returns to Scale; IRS Increasing Returns to Scale; DRS Decreasing Returns to Scale. 36 Note III June 2008 districts were scale inefficient. Seven of the 17 scale-inefficient districts had decreasing returns to scale (DRS), while ten of them revealed increasing returns to scale (IRS). To operate at the most productive scale size (MPSS), a district exhibiting DRS should scale down both its outputs and inputs. On the other hand, if a district displays IRS, expansion is justified. Table 3.8 Efficiency of public health facilities in Orissa, 2006-07 97. District-wise technical and scale efficiency scores for Orissa are presented in Table 3.8. Of the 30 districts, only 13 (43%) were technically efficient and, therefore, lie on the efficiency frontier. The remaining 17 (57%) districts were technically inefficient with scores ranging from 0.17 (Cutttack) to 0.99 (Kendrapara). 98. In Tamil Nadu the inefficient districts taken together had 468 beds in excess, 45 surplus doctors and 194 underutilized nurses, removal of which would not alter the existing output mix. In Orissa, on the other hand, the inefficient districts had 574 beds in excess, 23 surplus doctors and 129 underutilized nurses. However, in Orissa only two districts (Rayagada and Baragarh) had surplus doctors; doctors are 37 NOle III June 20V8 scarce in all the remaining districts. In general, excess medical personnel can be redeployed in districts that are technically inefficient and experience increasing returns to scale (IRS). 99. The study shows that in Tamil Nadu there are 12 districts having the Most Productive Scale Size (MPSS)). District such as Kanyakumari, Sivagangai, Nilgiris are experiencing IRS. This means that they need addition to the existing capacity. Dindigul district health system is experiencing decreasing returns (DRS), implying that this district could reduce the level of inputs/resources, without altering existing output level. 100. In Orissa, thirteen technically efficient districts were also scale efficient, indicating that they were having the most efficient productive size, given the input-output mix. One interesting fmding is that in Orissa the districts having Medical Colleges are less efficient than others. These districts also have the maximum slacks in number of beds and nurses. More recently established districts are more efficient than the older districts. 101. It is necessary to undertake follow up studies. This study has shown that districts with comparable inputs have performed differently. Some have performed much more efficiently than others. Naturally, the question arises: why have these district health systems performed more efficiently than the others? More in-depth case studies of selected district health systems could throw more light on the factors that contribute to efficiency of the public health system. CONCLUSION 102. This analysis reveals that if adequate and appropriate data are available, great insights can be generated using the DBA technique. It can be used as a tool for line ministries/departments to reallocate available resources in an efficiency enhancing way, and to target additional resources at specific states/districts where it would have maximum impact on outputs and outcomes. It could also be used by fmance and planning ministries/departments to link additional resource allocations to evidence of efficiency improvements. Overall, this technique could play a useful role along with other analytical methods to pursue development goals within a sustainable fiscal framework. 38 June 2008 AnnexA DEA Methodology The ongm of the modem discussion of efficiency measurement dates back to Farell (1957), who identified two different ways in which productive agents could be inefficient: one, they could use more inputs than technically required to obtain a given level of output, or two, they could use a sub-optimal input combination given the input prices and their marginal productivities. The first type of inefficiency is termed technical inefficiency while the second one is known as allocative inefficiency. These two types of inefficiency can be represented graphically by means of the unit isoquant curve in Figurel. The set of minimum inputs required for a unit of output lies on the isoquant curve YY'. An agent or decision­ making unit's (DMU) input-output combination defined by bundle P produces one unit of output using input quantities Xl and X2. Since the same output can be achieved by consuming less of both inputs along the radial back to bundle R, the segment RP represents the inefficiency in resource utilization. The technical efficiency (TE), input-oriented, is therefore defmed as TE = OR/OP. Furthermore, the producer could achieve additional cost reduction by choosing a different input combination. The least cost combination of inputs that produces one unit of output is given by point T, where the marginal rate of technical substitution is equal to the input price ratio. To achieve this cost level implicit in the optimal combination of inputs, input use needs to be contracted to bundle S. The input allocative efficiency (AE) is defmed as AE = OS/OR. Figure 1 Technical and Allocative Inefficiency y o X1,Y It should be noted that both output and input-oriented models identify the same set of efficient/inefficient producers or DMUs. The two methods of measuring efficiency produce the same estimates under constant returns to scale but give different values under variable returns to scale. A number of techniques have been developed over the past decades to tackle the empirical problem of estimating the unknown and unobservable efficient frontier (in this case the isoquant YY"). These may be classified using several taxonomies. The two most widely used catalog methods into parametric or non-parametric, and into stochastic or deterministic. The parametric approach assumes a specific functional fonn for the relationship between the inputs and the outputs as well as for the inefficiency term 39 June 2008 incorporated in the deviation of the observed values from the frontier. The non-parametric approach calculates the frontier directly from the data without imposing specific functional restrictions. Figure 2 Data Envelopment Analysis (DEA) production possibility fromier CRS F Output y .a VRS o x Iuput The Data Envelopment Analysis (DEA) is a non-parametric method; it avoids assuming specific functional forms for the relationship between inputs and outputs. DEA assumes that linear combinations of the observed input-output bundles are feasible. Hence it assumes convexity of the production set to construct an envelope around the observed combinations. Figure 2 illustrates the single input-single output DEA production possibility frontier. The DEA frontier is a piecewise linear locus connecting all the efficient decision-making units (DMU). The feasibility assumption, displayed by the piecewise linearity, implies that the efficiency of C, for instance, is not only ranked against the real performers A and D, called the peers of C in the literature, but also evaluated with a virtual decision maker, V, which employs a weighted collection of A and 0 inputs to yield a virtual output. DMU C lies below the variable returns to scale (VRS, further defined below) efficiency frontier, XADF, by DEA ranking. The input­ oriented technical efficiency of C is defined by TE YV/YC. If constant returns to scale (CRS) characterize the production set, the frontier may be represented by a ray extending from the origin through the efficient DMU (ray OA). By this standard, only A would be rated efficient. The important feature of the XADF frontier is that this frontier reflects variable returns to scale. The segment XA reflects locally increasing returns to scale (IRS), that is, an increase in the inputs results in a greater than proportionate increase in output. Segments AD and OF reflect decreasing returns to scale. It is worth noticing that constant returns to scale technical efficiency (CRSTE) is equal to the product of variable returns to scale technical efficiency (VRSTE) and scale efficiency (SE). Accordingly, DMU 0 is technically efficient but scale inefficient, while DMU C is neither technically efficient nor scale efficient. The scale efficiency of C is calculated as YN/YV. 40 June 2008 BmLIOGRAPHY Acharya, S (2001) 'India's Macroeconomic Management in the Nineties', Indian Council for Research on International Economic Relations, New Delhi ~~-~~. 2002. 'Macroeconomic Management in the Nineties', Economic and Political Weekly, April. Ahluwalia, M.S (2000) 'Economic Performance of States in Post-Reforms Period', Economic and Political Weekly, May 6, pp. 1637-48. Bagchi, A (2001) 'Perspectives on Correcting Fiscal Imbalance in the Indian Economy: Some Comments', ICRA Money and Finance Bulletin Volume 2, Numbers 4-5, Jan-June. Callen, Reynolds and Towe. (2001). 'India at the Crossroads: Sustaining Growth and Reducing Poverty', Washington D.C., IMF Chelliah, R (2001) 'The Nature Of Fiscal Crises in the Indian Federation and Calibrating Fiscal Policy', ICRA Money and Finance Bulletin Volume 2, Numbers 4-5, Jan-June. Devarajan, S. (2005). 'South Asian Surprises', Economic and Political Weekly, September 10, pp. 4013-14. Farell, M (1957). 'The Measurement of Productive Efficiency', Journal of the Royal Statistical Society, Series AVolume 120, No.3, pgs 253-290. Favaro, EM. and Lahiri, A.K (2004) 'Fiscal Policies and Sustainable Growth in India', Oxford University Press. Ghosh, M (2006) 'Economic Growth and Human Development in Indian States' ,Economic and Political W eeldy, Vol. 40, No. 30. Government of India (2005) Ministry of Finance 'Report of the Twelfth Finance Commission (2005-10),. - - - . 2005-06. 'Economic Survey'. - - - . 2006-07. 'Economic Survey'. Government ofIndia (2006) Planning Commission 'Towards Faster and More Inclusive Growth: An Approach paper to the Eleventh Five year Plan' . Herrera, Santiago and Gaobo Pang (2005), "Efficiency of Public Spending in Developing Countries: An Efficiency Frontier Approach"., World Bank. Indian Institute of Technology, Chennai Project Report (2008) 'Output and Outcome Efficiency of Health Expenditure' . Indicus Analytics Report (2007) 'Efficiency in Public Spending Data Envelopment and Econometric Analysis' Jalan, B (2005) 'The Future of India: Politics, Economics and Governance', Penguin, India, New Delhi Jbingran and Deepa Sankar (2006), 'Education Development Index: Targeting Disparity Reduction in Elementary Education' , mimeo. Kathuria, Vand Sankar, Deepa (2005) 'Inter-State Disparities in Health Outcomes in Rural India: An Analysis Using a StochasticProduction Frontier Approach', Development Policy Review, Blackwell Publishing. 41 June 2008 Kopitz, G (2001) 'Fiscal Policy Rules for India?' Economic and Political Weekly, March 3, Nagarajan Mohan (2008) 'Maharashtra's Fiscal Sustainability" mimeo, World Bank Rakshit, M (2001) 'Some Public Economics of Food Subsidy and Buffer Stock Operations in India: Part I', {CRA Money and Finance Bulletin Volume 2, Numbers 4-5, Jan-June. Rae, M Govinda, (2002), 'State Finances in India: Issues and Challenges', August 3,2002, Economic and Political Weekly Ravishankar, V.J, Rajni Khanna and Arpita Chakraborty (2007)'Uttar Pradesh: Fiscal Improvement and the Unfinished Agenda, Policy Note, W orid Bank. Reserve Bank ofIndia. (2004) 'State Finances study of Budgets 2004-05'. Mumbai 2004. 'Annual Report 2004-·05'. Mumbai '-'-"'-'-. 2005. 'State Finances study of Budgets 2005-06'. Mumbai 2006. 'State Finances study ofBudgets.2006-0T Mumbai 2007. 'Macroeconomic and Monetary Developments'. 2007-08. Mid-Term Review. Mumbai Pinto, B, Zahir, F (2004) 'Why Fiscal Adjustment Now', Economic and Political Weekly, March 6, pp.1039-1048. Pinto, B, Zahir, F and Pang, G. (2006) "From Rising Debt to Rising Growth in India: A case of Microeconomic Dominance", India Development Policy Review, Macmillan India Ltd, New Delhi. Pinto, B and Zahir, F (2003) 'India's Fiscal Policy (Chapter 2)' in India: Sustaining Reform, Reducing Poverty, World Bank Development Policy Review, Oxford University Press, New Delhi. Srinivasan, T.N.(2001) 'India's Fiscal Situation: Is a Crisis Ahead?', Working Paper 92. Stanford University, Center for Research on Economic Development and Policy Reform, Palo Alto, Ca. World Bank, 2006 'India Inclusive growth and Service Delivery: Building on India's successes', Macmillian Press 2005 'State Fiscal reforms in India: Progress and Prospects', Macmillan Press. ~-'----. 2003, 'India Sustaining Reform, Reducing Poverty', Oxford University Press: New Delhi World Bank Report (2008) 'Orissa in Transition-From Fiscal Turnaround to Rapid and Inclusive Growth' 2008. 'Accelerating Growth and Development in the Lagging Regions ofIndia'. Wade, Paul (2008) 'Efficiency of Public Education Spending in Indian States', World Bank (mimeo). Zahir, F (2005) 'An Analysis of Fiscal Situation in West Bengal', South Asia Poverty Reduction and Economic Management, mimeo, World Bank. Zahir, F and Marina Wes (2004) 'India: States Fiscal Crisis: Is it receding? , State Fiscal Reforms in India: Progress and Prospects, Macmillan India Ltd, New Delhi. Zahir, Farah (2002) "India- Trends and Pattern of Public Spending in Uttar Pradesh", Policy Note, Uttar Pradesh Fiscal Restructuring and Governance Loan, World Bank. 42