Fiscal transfers based on inputs or outcomes? Lessons from the Twelfth and Thirteenth Finance Commission in India

Summary Background There is limited empirical evidence about the efficacy of fiscal transfers for a specific purpose, including for health which represents an important source of funds for the delivery of public services especially in large populous countries such as India. Objective To examine two distinct methodologies for allocating specific‐purpose centre‐to‐state transfers, one using an input‐based formula focused on equity and the other using an outcome‐based formula focused on performance. Materials and Methods We examine the Twelfth Finance Commission (12FC)'s use of Equalization Grants for Health (EGH) as an input‐based formula and the Thirteenth Finance Commission (13FC)'s use of Incentive Grants for Health (IGH) as an outcome‐based formula. We simulate and replicate the allocation of these two transfer methodologies and examine the consequences of these fiscal transfer mechanisms. Results The EGH placed conditions for releasing funds, but states varied in their ability to meet those conditions, and hence their allocations varied, eg, Madhya Pradesh received 100% and Odisha 67% of its expected allocation. Due to the design of the IGH formula, IGH allocations were unequally distributed and highly concentrated in 4 states (Manipur, Sikkim, Tamil Nadu, Nagaland), which received over half the national IGH allocation. Discussion The EGH had limited impact in achieving equalization, whereas the IGH rewards were concentrated in states which were already doing better. Greater transparency and accountability of centre‐to‐state allocations and specifically their methodologies are needed to ensure that allocation objectives are aligned to performance.


| INTRODUCTION
In large, federal, populous countries such as the United States and India, fiscal transfers from central to state governments are an important source of funds for the delivery of public services. 1,2 When designed well, such intergovernmental fiscal transfers can increase the accountability and the effectiveness of public service delivery, which in turn can lead to improved outcomes in areas such as in health. 3,4 Although there is extensive theoretical knowledge about different fiscal transfer mechanisms, there is limited empirical evidence about the effectiveness and efficiency of these fiscal mechanisms especially with regards to improving health outcomes in low-and middle-income countries. 5 Two theoretical approaches have been applied to understanding these intergovernmental fiscal transfers. Traditional fiscal federalist theory assumes a perspective of the social planner, with the planner being the central government, and provides a normative framework for assigning responsibilities to different levels of entities. 6,7 Given the revenues raised through taxation, a given central government entity may face the challenge of distributing these funds across states and other subnational entities. Fiscal transfers can theoretically address two types of imbalances, vertical and horizontal. Vertical imbalances arise from the different ability of states compared to the central government in raising funds and carrying out their functional responsibilities, eg, public service delivery in health and education. Horizontal imbalances arise from the disparities across constituent units in the federation, ie, the states themselves, due to the different revenue-raising abilities. Federal distribution of funds can thus either exacerbate or alleviate these disparities.
A second relevant framework is principal-agent theory. 8 In the case of intergovernmental fiscal transfers, a given central government entity may be seen as the principal and the state government as the agent. The principal (central government) has certain predefined goals, which are delegated to the agent (state government) to execute, and the agent is paid for completing certain activities and reaching these defined goals. Here, the systematic transfer of funds is often based on the receipt or verification of the ex post performance of certain indicators including submission of utilization certificates which may need to be submitted before receiving additional funds. Sometimes, however, the principal is unable to verify the outputs of its payments to the agent, or the outputs are not well defined and can be manipulated, and, consequently, the agent is able to shirk-by accepting funds without actually carrying out the mutually agreed upon activities and instead using the funds to their preferences.
Through the lens of a social planner, it can be argued that central governments have focused on the design and implementation of ex ante formula to allocate funds to its constitutive state treasuries and thereby address these two imbalances as an immediate goal of planning. Funds allocated based on ex ante formula can be channelled as general purpose transfers or unconditional block grants, thereby giving states the autonomy to allocate their budgets and determine their activities and expenditures, or alternatively transfers which focus on addressing vertical and horizontal imbalances. In contrast, principal-agent theory lends itself to the use of conditional block grants or specific purpose transfers, with a focus on ex post conditionalities placed on activities completed or results achieved using the allocated funds. Both theories nevertheless lend to the use of formula or systematic methodology but vary in the timing of their payments and the type of data and information on which payments are determined.
This study adds to both of these literatures by using the case of specific-purpose transfers by the Twelfth and Thirteenth Finance Commissions in India, one transfer using an input-based formula and the other transfer using an outcome-based formula. Finance Commissions are constitutional bodies set up every five years to provide recommendations on the division of tax revenue and grants between the central and state governments (see next section for further background on the Indian intergovernmental fiscal transfer system). For the purpose of this paper, we define inputs as the financial, human, and physical resources required to carry out an activity, whereas outputs and outcomes are measures and goals of an activity. To examine the input-based formulaic approach, we study the case of the Twelfth Finance Commission (12FC)'s use of Equalization Grants for Health (EGH), for which the formula examined the needs and differences across states in their revenue capability (ie, their inputs). To examine the outcome-based formulaic approach, we analyse the case of the Thirteenth Finance Commission (13FC)'s use of Incentive Grants for Health (IGH), for which the formula sought to incentivize states towards better performance for health outcomes (ie, their outcomes).
The goal of this study is to examine the extent to which both of these specific-purpose transfers can be replicated using their predefined methodology and to investigate the advantages and disadvantages of these methodologies which, to our knowledge, has not been analysed or discussed hitherto in the literature. This paper is organized as follows. First, we describe the background to these specific-purpose transfers of the 12FC and 13FC including international comparisons in the UK and US for additional contextual background. Next, we examine the methodologies and describe the data used. Our main results are a simulation and replication of these methods, with a focus on the distribution of funds across states and their equity and potential health implications. We conclude by offering specific policy recommendations to improve these types of mechanisms and policy alternatives.

| Indian intergovernmental fiscal transfer system
Theoretical approaches are useful for providing intuition for prediction, but theories nevertheless should be contextualized to real-world circumstances. As such, we briefly describe the complex intergovernmental fiscal transfer system in India and the political economy of these transfers to help elucidate the weaknesses of theory. Fiscal transfers from the centre to state in India represent a complex relationship between national and subnational governments and "are governed by a complicated mix of constitutional assignments, institutional precedents, discretion and negotiation." 9 The seventh schedule of the Indian Constitution assigns the specific powers and functions of the centre and the states. This list includes the Union list of the exclusive powers of the centre, the state list of exclusive powers of the states, and the concurrent list of those falling under the joint jurisdiction of both the centre and the states. Health in India is a state subject per the Constitution. However, amendments have given the Central government joint responsibility in areas such as population control, prevention of spread of infectious diseases, and regulation of food and drugs. This overlap of responsibility in financing of health between the central and state governments requires an effective and efficient system of fiscal transfers.
There are broadly three main channels of fiscal transfer systems. First, the Finance Commission, a constitutionally mandated body appointed by the President of India every 5 years, is responsible for the distribution of taxes between the centre and states. In addition to tax devolution, the Finance Commission is also required to recommend specific grants to the states in need of assistance.
The second main channel of fiscal transfers was through the now-defunct Planning Commission-a central government body set up through an act of Parliament. Similar to the role of a social planner, the Planning Commission had allocative decision-making power and responsible for dispensing assistance to states to finance development activities. Planning Commission transfers came through various channels: formula-based normal central assistance, additional central assistance (mostly project financing in infrastructure), and special central assistance (on the basis of state-specific priorities). States which were particularly disadvantaged were categorized as special category (as opposed to general category) and received additional funding. Past research found that allocations to states by Planning Commission were politically driven over 1987-1995: states which elected leaders of the same party as the central government tended to receive more Planning Commission funds, whereas allocations by Finance Commission were not susceptible, if not countercyclical, to changes in political power. 10 Finally, the third main channel is through various central ministries which give discretionary grants to their state counterparts, either wholly funded by the Centre (central sector programs) or requiring a matching state contribution known as Centrally Sponsored Schemes (CSS). An example of a CSS for health is the National Health Mission (NHM).

| Indian health system in an international context
India's health system follows the model of the UK National Health Service (NHS) because a large part of the public health infrastructure was created during colonial time and expanded after independence in 1947. India's publicly delivered health services have suffered from underinvestment in personnel, infrastructure, and other inputs, leading to huge variations in service quality across states. 3 The UK health system's use of target setting may provide a comparison for  11 The effect on services excluded from the ratings, however, is unclear, as there have been allegations of data manipulation and "gaming" to meet the prescribed targets. 12 The move from structure and process indicators in the mid-2000s towards outcome indicators at the end of the decade was in part due to the recognition that the latter are more effective at capturing improvements in the quality of services, a key goal of any health sector reform.
In contrast, financing for the US health system, specifically the Medicaid program, health insurance for low-income households, provides a comparison for India's input-based fiscal transfers such as the Twelfth Finance Commission's EGH and the Centrally Sponsored Scheme called the NHM. US Medicaid is financed using a formula called the Federal Medical Assistance Percentage (FMAP), for which the federal government pays a fixed share of all eligible Medicaid spending and is also called the "match rate". The formula depends on the state's income level, which is a proxy for the state's ability to pay and raise revenues as well as the state's level of eligibility for the program, and is determined annually by comparing state's average per capita income level to the national average income level. According to this formula, the federal government shares range from 50% for higher-income states to about 74% for lower-income states, with the remainder covered by either state or local government. 13,14 Local government can include Medicaid providers (eg, a county hospital or school district) which either transfer funds to the state Medicaid agency through an intergovernmental transfer, or certify expenditures incurred to provide Medicaid services or administration, known as a certified public expenditure. The latter practice is analogous to the NHM for which states are expected to contribute 40% of total resources with centre providing 60% as of financial year 2015-16. NHM payments, however, are not linked to insurance payments, similar to Medicaid certified public expenditures. While both EGH and IGH use a formula, neither EGH nor the IGH require matched state contributions as in Medicaid.

| Twelfth Finance Commission's Equalization Grants for Health
Low and unequal public spending on social sectors including the health sector has long been identified by Government of India as a challenge. 15 Indeed, CSS in India (such as NHM) were established precisely to increase spending by both state and central governments. The EGH established by the 12FC were also intended to address this problem of low public spending. See Annex 1 for per capita total public health expenditure prior to 12FC.
For the period 2005-10, the 12FC was guided by "[a]n equalization principle in determining service-specific grants can play an important role in a situation where, while the average expenditure on health and education may grow covering all states, for some states where service provisions are below average, expenditure on these heads needs to grow faster than the average, if they are to catch up." 16 Thus, the 12FC recommended two equalization grants to be distributed among states whose per-capita expenditure was below the national average-one grant for the education sector amounting to Rs 10,171.65 crore (1 crore = 10 million) distributed to eight states and the other grant for the health sector amounting to Rs 5,887.08 crores for seven states.
Full capacity equalization would be constrained by a number of factors, including the sheer quantum of grants that would be required to achieve full equalization as well as presence of multiple channels of fiscal transfer, large and growing differences in state capacities, and problems regarding availability of accurate, real-time data. Given these constraints, the 12FC focused on providing a partial correction in per-capita expenditures for health and education. This was done in two ways. First, it allowed for expenditure restructuring in favour of each sector by providing higher growth rate for non-salary component in projecting expenditure. Second, they provided separate grants-in-aid for certain states for health and education. Each of these is described in detail in Section 3 on methodology.

| Thirteenth Finance Commission's Incentive Grants for Health
In contrast, the IGH established by the 13FC were intended to spur improvements in health outcomes and specifically further decrease infant mortality. Although formula-based transfers have long been used by both the Finance and Planning Commission, much less is known about the use of formula in which the outputs or outcomes, rather than inputs, are used.
The 13FC guided by the principles of the "need to improve the quality of public expenditure to obtain better outputs and outcomes" recommended six different grants with incentives to improve outcomes as a supplement to its tax devolution and block grants, which totalled Rs 14,446 crore over the 5-year period (2010-15). Of these 6 grants for improving outcomes, one grant was allocated Rs 5,000 crore (35% of the total) for health specifically. This IGH used outcome-based transfers with the goal of reducing infant mortality rates (IMR) in states. This novel use of outcomebased transfers marked a departure from the 12FC which used an ex ante formula based on financial management and expenditure scores of states. , the next step entailed determining the growth rate for the forecast period. It was found that, for the health EGH, on average 75% of expenditures consisted of salaries. In order to ensure that salaries are held constant in real terms, a 5% growth rate was assumed for the salary component of health expenditures. For the non-salary component, a growth rate of 30% was assumed. Combining the growth rates of salary and non-salary components with their respective weights as mentioned previously, a composite growth rate of 11.5% for health was obtained.
These growth rates were then applied to states in order to obtain projected health expenditure for the period 2005-10.

b. Grants-in-Aid for Specific States
The next step was determining the eligibility of states who would receive the EGH and the quantum of grant to be provided. In devising the specific purpose grant for health, a two-step normative approach was adopted. First, given that states accord different priority to the health as expressed in their total public health expenditure per capita, care was taken to not underwrite the deficiency in priority accorded to the sector by the state government by "correcting" for this low expenditure preference-see Table 1.
For this purpose, total expenditure data (both non-plan and plan) for each of the states were examined for 2002-03. The ratio of spending on health as a proportion of total expenditure was worked out for each state according to their categories ("special category" and "general category states"). In explaining the methodology, 12FC observed that "while working out this ratio, expenditure relating to pensions, interest payments and other adjustment items... were excluded from non-plan revenue expenditure for arriving at the 'adjusted' total revenue expenditure." Next, the deviation of the average preference from the corresponding group average was calculated. For states whose proportionate expenditure on health was lower than the group average, the states proportion was replaced by the corresponding average group ratio. Having corrected for low expenditure preference, the next step entailed identifying those states which, even after spending the required percentage, fell short of comparable average percapita expenditure in the sector. This corrected per-capita revenue expenditure for each state for 2002-03 was thus calculated. Here, too, group averages were simultaneously worked out for general and special category states. The 12FC observed that those states whose per-capita levels of public health expenditure were found to be below their group average "were reckoned as needing financial assistance, because their lower expenditure could be on account of low fiscal capacity." 16 For the states that fell in this category, due to limited availability of resources, the grant amount was calculated for covering 30% of the distance by which a below-average state was lagging behind its group average of per capita expenditure. Finally, having determined the amount of grant required in 2002-03 for health (multiplying the per-capita requirement with the population figures), a growth rate equal to the group's TGR for 1993-2003 for non-plan revenue expenditure (NPRE) on health was applied on this amount in order to estimate the quantum of grant in the base year.
The 12FC put the minimum amount to be received by an eligible state to Rs 10 crores per year as grant.

| Data sources
The primary data used for this exercise was the data from Finance Accounts available from state governments. In replicating this exercise, we have also used finance accounts data from 1993-2010. Population data in calculating percapita health expenditure were taken from Census 2001.

| Methodology and data for 13FC IGH
The methodology for the IGH was developed by the Administrative Staff College of India (ASCI) Hyderabad, which was tasked by the Finance Commission to design a formula-based approach that could improve outcomes. ASCI's final methodology recognized that states each had different starting levels of performance, and their methodology used the median of state IMR as a key parameter in its formula. The intention of the formula was to provide greater incentives to states below the median state IMR. Scenario 2. If the state's IMR in assessed year t decreased from its 2009 value (ie, IMR has improved) and is now below the median: We found 4 such states in 2011 and 6 such states in 2012.

Case 2. If the state is below the median value in 2009
Scenario 3. If the state's IMR in the assessed year t remains below the median IMR and drops below their own 2009 value (ie, shows improvement), the state's IC is as follows: (Continued) Scenario 4. If the state's IMR in the assessed year t increased from its baseline 2009 IMR but remains below the median, the state's IC is 100. We found 3 such states in 2011 and 2 such states in 2012.
In addition, there were two additional scenarios in which no incentive was to be given, ie, the state's IMR in the assessed year was worse than the baseline, and its baseline value was already above the median, or if the state's IMR in the assessed year was below the baseline median but its IMR increased above the baseline median.
Step 2. Sum all the Incentive Coefficients for all states.
Step 3. Divide each state's Incentive Coefficient by the sum of all Incentive Coefficients for that assessed year to obtain the state's share.
Step 4. Multiply the total allocation of the IGH for the assessed year by the state's share.
The IGH gave states a lead time in the first 2 years (2010 to 2011) to make improvements, and allocated the Rs  Table 2.

| 12FC EGH results
A comparison of the projected and actual NPRE for the period 2005-10 indicates interstate differences ( In short, we were unable to replicate the grants-in-aid to the specific seven states due to data constraints. The 12FC methodology note mentioned a number of adjustments undertaken in the NPRE. While some of these such as pensions, lotteries, calamity relief, and interest payments were easy to identify, it was difficult to cull out all the adjustments made to the NPRE. As a consequence, only a partial replication could be made, and the analysis has thus been omitted from this study.
Despite the difficulties of replication, the relative amount of allocation to states merits consideration. The EGH for the seven states had a number of conditionalities attached in terms of the NPRE that should be incurred, and consequently not all grants-in-aid were ultimately released (see Annex 3,

| 13FC IGH results
Using the IMR values presented in Table 2, we replicated the 13FC's formula for IGH and calculated each state's expected share of total allocations for the first 2 years of IGH's implementation-see Table 4. Table 4 shows that one state alone, Manipur, received more than a quarter and more than a fifth of the national allocation in the 2 years. Four states (Manipur, Sikkim, Tamil Nadu, and Nagaland) received over half of the national allocation, in each of the 2 years. The formula incentivizes states that are already below the median by offering a higher share of the total allocation. In 2012 and 2013, better-performing states who were already below the median level, received approximately 57% of the national allocation of the total IGH and in 2013 they received 56% of the grant amount.
Whereas the formula was constructed purely based on performance of IMR relative to different standards, the allocation levels were not benchmarked according to either the level of health expenditure or the population. Indeed, given the capture of over half of the winnings of the total national allocation, ie, Rs 1,500 crores for four states in each of the first two years, the question arises regarding the size of these winnings relative to the total state health expenditure budget. Alternatively, the allocated funds can be benchmarked relative to the population, to give a sense of the leverage that the incentive might (or might not) have provided.

| DISCUSSION AND CONCLUDING REMARKS
In this study, we analysed the two cases of the 12FC and 13FC's use of two different specific-purpose transfers for health. We found that EGH placed conditions for releasing funds, but states varied in their ability to meet those conditions, and hence their actual releases varied considerably, eg, Madhya Pradesh received 100% and Odisha 67% of its expected allocation.
We found that the 13FC IGH allocations revealed highly unequal distribution of 'winnings' and raises several questions about the efficacy and implications on equity of the formula. The formula's design gave a particularly large reward to those states which were already doing better, ie, states which were already below the median level for IMR. We found that the implementation of the formula generated a perverse scenario in which the vast majority of the grants went to four states alone. Further, we found that the absolute level of allocations relative to certain benchmarks were be incentivized to increase their eligibility status. The ability to replicate these allocations using a transparent methodology and publicly available data can help to ensure greater integrity and robustness of the allocations and reduce the risk of poor distribution in consequences that were seen. One general recommendation is to encourage formulabased centre-to-state allocations and to make their data and methodology replicable and available to states and the public at large. In general, greater transparency of the methodology of the Finance Commission reports as well as data collected and analysed are needed.
Despite these design challenges in both the 12FC and 13FC, formula-based transfers in India are more important today than in the recent past. A third type of allocation is that which is not based on any formula and often are discretionary to political decisions. Grants determined by the Finance Commission have been predominantly based on formula, either for general purpose by the state treasury "Tax Devolution and Block Grants" or for specific purpose, eg, "Equalization Grants" or "Incentive Grants", with a fraction of funds going for state-specific grants (which are not based on formula). 17 By no means should the results of the 13FC IGH replication be a rejection of all outcome-based fiscal transfers.
Rather, the particular design of this fiscal transfer is clearly problematic and with a severely perverse distribution. The complex formula also presented several highly implausible counterfactual scenarios, eg, providing no incentive to the better-performing states if their IMR increased (highly unlikely, because they were already on a trajectory to reduce their IMR). The lack of population weight or weight by the number of babies born or babies who died was also problematic and contributed to this severely unequal distribution of payments captured by small better-performing northeastern states.
There are many alternatives to structuring the formula of outcome-based fiscal transfers. How the budget envelope is defined (eg, on a per unit basis) and whether the states can collect savings (eg, if states can deliver cheaper programs while maintaining coverage) are critical design parameters. The US Medicare Shared Savings program, for example, incentivizes states in the form of an increased share of the savings and also reflects a global budget payment mechanism potentially linked to specific quality indicators. 19 Despite the perverse distributional consequences of this formula, the 13FC should be lauded for their boldness and willingness to experiment with such an innovative fiscal transfer mechanism and should continue to experiment and learn. 20 In addition, the IGH's use of an independent data source-namely, the SRS-is a substantial improvement over the use of self-reported administrative data, which are subject to bias.
This study's results also raise questions with regards to the efficacy and efficiency of different transfer mechanisms in a context with multiple channels of delivery and usage. For instance, all the states which received the EGH also received additional funds under the National Rural Health Mission being part of the group that was lagging behind in health indicators compared to the rest of the country. In addition, there are general purpose transfers to states both by the Finance Commission and Planning Commission. These multiple sources of funds with their independent criteria and release mechanisms can at times lead to a fragmentation of transfers, which in turn can fragment state's attention. 21 These specific-purpose transfers do not tackle the challenge of fungibility. Past work has shown that central-tostate allocations results in state's lower contributions in that sector. 22 Comparable issues of fungibility and lowering domestic spending in response to international aid have been observed in the context of international-to-national aid relationships. 23 To get better outcomes, international experiences suggest that transfers need to consider at least three key dimensions. First, central government's allocation of national revenues to subnational governments should correspond to population size and needs, however defined. Second, transfers should generate incentives to improve the quality of the spending by subnational governments and their subsequent performance on outcomes. Third, independent systems to monitor, evaluate, and provide feedback data on subnational performance can generate greater accountability to the central government, parliaments, and legislatures, and ultimately greater accountability to the public. Improving performance of intergovernmental fiscal transfers may well require greater transparency and greater accountability.
Beyond the designs of outcome-based fiscal transfers, future work should go further down below the level of state, ie, how to allocate transfers to districts and ultimately households. The Fourteenth Finance Commission, the most recent one, submitted its report in February 2015. 24 In the context of this Finance Commission's recommendations which significantly increased the share of untied transfers to states, both the central government and states could consider the use of such specific-purpose transfers to districts.
Data at the district level is hardly in short supply in India. The District Level Health Survey conducted by the International Institute for Population Sciences collects representative data every five years about reproductive and child health in all Indian districts. Moreover, the newer Annual Health Survey conducted at district level by the Office of the Registrar General and Census Commissioner represents an important source of independent, unbiased, and accurate information about mortality in lower-performing districts. Together, these two sources of information could be utilized to develop district-level outcome-based fiscal transfers.

FUNDING
Bill and Melinda Gates Foundation. improvement over the base year level. The intention is that states that are already at a relatively higher level of performance and are to some extent disadvantaged by the restricted scope for incremental percentage improvement should not stand to lose. Hence, their percentage improvement in performance over the base (or initial) year should be suitably weighted to compensate them for this "inherent disadvantage". It is, therefore, proposed to weight their performance by the distance of their output/outcome indicator from the median (benchmark) as a percentage of the median (benchmark).

For Infant Mortality
Rate there is an inverse relation between the level of the indicator and performance of the state. Ie, a decrease in the indicator will lead to an incentive, while an increase will be penalized. The formula incorporates this requirement.