Debt sustainability of states in India: An assessment

The debt position of the state governments in India, which deteriorated sharply between 1997–1998 and 2003–2004, has witnessed significant improvement since 2004–2005. Debt sustainability analysis based on empirical estimation of inter-temporal budget constraint and fiscal policy response function in a panel data framework, covering 20 Indian states for the period 1980–1981 to 2015–2016, indicates that the debt position at the state level is sustainable in the long run. The increase in contingent liabilities of states and take-over of large chunk of these liabilities through debt restructuring of State Power Distribution Companies, however, would adversely affect the debt position of states.


Introduction
In line with an overall decentralizing trend, the sub-national governments worldwide have been entrusted with increasing responsibilities towards delivery of public goods and services, and investment in physical and social infrastructure. As the concomitant expenditure requirements generally fall short of own revenue receipts and inter-governmental transfers from the national authorities, the sub-national governments have to depend on borrowed resources to finance such expenditure. However, the borrowing limits of sub-national governments in various countries are subject to either regulatory restrictions or self-imposed fiscal discipline, given the underlying requirement to ensure debt sustainability at the sub-national level.
In India, the state governments have been playing an important role in discharging various functions assigned to them under the Constitution. As the non-debt 1 3 Debt sustainability of states in India: An assessment relating to the evolution of state government debt in India are presented in Sect. 5. Section 6 presents an empirical assessment of debt sustainability at the state level based on different approaches. The rationale for extending the conventional debt sustainability analysis to include off-budget fiscal position of states in the context of additional debt liabilities which have arisen on account of take-over of debt of state power utilities is explained in Sect. 7. The concluding observations are covered in Sect. 8.

Defining debt sustainability
Debt sustainability is a term that has been used with increasing frequency in the academic literature and multilateral policy discussions, but with different connotations under different circumstances (Balassone and Franco 2000;Chalk and Hemming 2000). How one defines debt sustainability could affect the conclusion one arrives about the sustainability or otherwise of debt in an economy. In the pioneering work on debt sustainability, based on the post-Second World War US data, Domar (1944) pointed out that primary deficit path can be sustained as long as real growth of the economy remains higher than the real interest rate. Buiter (1985) suggested that sustainable fiscal policy is one that is capable of keeping the public sector net worth to output ratio at its current level. Blanchard (1990) provided two conditions for sustainability viz., (a) the ratio of debt to GNP should eventually converge back to its initial level, and (b) the present discounted value of the ratio of primary surpluses to GNP should be equal to the current level of debt to GNP. Buiter (1985), Blanchard (1990) and Blanchard et al. (1990) considered debt level as sustainable if a country's debt to GDP ratio remains stable, and if the economy generates debt stabilising primary balance to cover that debt in future.
In terms of the standard definition of fiscal sustainability, the ratio of outstanding debt and debt servicing to GDP, in a steady state, should not increase over time (World Bank and IMF 2010). The focus in this approach is on stabilising the debtto-GDP ratio. International Monetary Fund (2011) considers a set of fiscal policies as sustainable in case a borrower is able to continue servicing its debt without an unrealistic large future correction to its income and expenditure.
Typically, conventional debt sustainability analysis is an accounting-based approach linked to the inter-temporal budget constraint as follows: which states that public debt at the beginning of the period t + 1 i.e., (B t+1 ) equals past period debt including interest payments but adjusted for primary balance, depending on whether there is primary surplus or deficit. Recursively solving (1) with time period (t) starting at 0 and extending up to infinity, we get (1) (2) B 0 = ∞ ∑ t=1 PS t ∕(1 + r) t + Lim t→∞ B t ∕(1 + r) t Fiscal policy is said to be sustainable, if the initial stock of debt is equal to the sum of present discounted values of primary surpluses. Alternatively, the present value of revenues must be equal to the present value of spending including interest on the public debt plus repayment of the debt itself. This is defined as the inter-temporal budget constraint and is satisfied if the discounted sum of end-period debt converges to zero, i.e., Lim B t /(1 + r) t becomes 0. This transversality condition rules out a 'Ponzi' scheme and requires that debt should not grow at a rate faster than interest rate. The solvency condition for government debt implies that future budget surpluses would be sufficient to meet current debt liabilities.
The transversality condition relating to the long-term solvency of public debt, when expressed in terms of GDP ratio, states that the GDP growth rate has to be lower than the interest rate so that the discounted terminal period debt ratio converges to zero. This implies that in case of a positive initial public debt, the sum of the cumulated discounted future public surpluses should exceed the sum of the cumulated discounted future public deficits. However, if the rate of growth of GDP is higher than the interest rate, there would be reverse stabilising effect on the ratio of debt to GDP even if a sub-national government is accumulating primary deficit. It may not always be possible to sustain high growth situation and/or maintain the positive growth-interest differential for all times to come; and a positive primary balance may become necessary to ensure sustainability of public debt and avoid Ponzi scheme.

Review of literature
In the theoretical literature, the rationale for maintaining low/sustainable level of debt is attributed, among others, to the need to ensure sustainability of fiscal policy, provide fiscal space for undertaking counter-cyclical policy, absorbing contingent liabilities without threatening debt sustainability, reducing vulnerability to crises and optimizing growth by reducing the risk of crowding out of private investment, while taking into account concerns relating to inter-generational equity and future spending needs. In the Indian context, there are several empirical studies, which have examined fiscal/debt sustainability of states (Table 1).
Overall, the empirical studies on debt sustainability at the state level in India indicate a mixed picture. While some of the studies point out that the debt position of states is unsustainable, others have drawn attention to the declining debt-GSDP ratios at the state level and attributed this improvement to the strong growth performance and the implementation of fiscal rules during 2003-2012. It is held that a slowdown in growth momentum could pose risk to the achievement of envisaged gross fiscal deficit and debt-GSDP targets under the medium-term scenario.

3
Debt sustainability of states in India: An assessment  Makin and Rashmi (2012) 1990-1991 Fiscal sustainability at the state level While majority of the states have stabilised public debt levels as a proportion of GSDP, the slowdown in economic growth could expose many Indian states to considerable fiscal risk Misra and Khundrakpam (2009) 1991-1992to 2007-2008 Debt sustainability of state governments The liabilities of state governments, based on the Present Value of Budget Constraint, were found to be unsustainable Nayak and Rath (2009) 1991 Debt sustainability of special category states The Domar sustainability condition i.e., real growth should be higher than the real interest rate was achieved in all the states except Arunachal Pradesh, while the solvency condition was satisfied only in the case of Assam Rajaraman et al.(2005) Goyal et al. (2004Goyal et al. ( ) 1951Goyal et al. ( -2000 Debt sustainability of the centre, states and general government After addressing the issue of regime shift, while fiscal stance of the central and state governments at the individual level were found to be unsustainable, it was weakly sustainable for the combined finances of centre and states Dholakia et al. (2004Dholakia et al. ( ) 1988Dholakia et al. ( -1989Dholakia et al. ( to 2003Dholakia et al. ( -2004 Debt sustainability of states Based on a uniform target of debt to GSDP ratio of 35%, it was observed that there was a debt problem of significant magnitude only in about half of the 25 states covered in the study Buiter and Patel (1992) 1971-1989 Debt sustainability of centre, states and public sector undertakings (PSU) Indian public debt was found unsustainable after discounting by various alternative measures of interest rates as all the discounted debt series turned out to be non-stationary

Need for assessment of debt sustainability at the state level
Globally, sub-national governments (SNGs) have assumed importance in the wake of their increasing role in provision of various essential services while also catering to urban infrastructure requirements. In this process, their resource base has also expanded with growing dependence on borrowed funds. However, the borrowing limits of SNGs are, by and large, regulated by the upper tiers of government in countries with a federal system. In countries with 'golden rules' in place, borrowings are required to be authorised, and in some countries (France, Ireland and the UK), the central government could directly restrict borrowings by lower levels of government. In Sweden, it is mandatory for SNGs to balance their budgets by year-end; in case of deficits, balance has to be restored in 2 years. Apart from the imposition of restrictions on borrowing limits, the practice of having explicit co-ordination agreements between different government tiers have also been observed.
In the Indian context, the starting point of the debt sustainability exercise is to examine whether the state governments really face hard budget constraint? Article 293 of the Indian Constitution stipulates that a state may not without the consent of the Government of India raise any loan if there is still outstanding any part of a loan which has been made to the state by the Government of India or by its predecessor Government, or in respect of which a guarantee has been given by the Government of India or by its predecessor Government. This implies that the state governments do not have unrestricted power to borrow as long as they are indebted to the Centre. In addition, states are also prohibited from borrowing abroad with the exception of loans from multilateral financial institutions intermediated by the central government.
In addition to the restrictions under Article 293 of the Constitution of India, the state governments have gone ahead with the self-imposed restrictions through the enactment of FRBM Acts/FRLs. The implementation of a rule-based fiscal discipline mechanism under these enactments since the early 2000s has been marked by a gradual move towards sustainability of their fiscal and debt positions, with majority of the states achieving the thirteenth Finance Commission (FC-XIII) targets as also their self-imposed targets. However, a few states continue to face fiscal stress and their debt positions remain an area of concern. Furthermore, notwithstanding strict monitoring of overall borrowing limits and adherence to various restrictions, the state governments have been able to raise additional 'off-budget' borrowings with guarantees through state-controlled Special Purpose Vehicles (SPVs) and/or state-owned public sector enterprises (SPSEs), which have in-built risks of various kinds. It is against this backdrop that the following Section presents the evolution of debt position of state governments beginning 1980-1981.

3
Debt sustainability of states in India: An assessment 5 Evolution of state government debt in India: some stylised facts The fiscal position of states in India, which had remained comfortable in the first three decades since independence, exhibited signs of fiscal stress since the mid-1980s. The average debt-GDP ratio inched up slightly from 18.3% during the 1980s to 20.8% during the 1990s. The period from 1997-1998 to 2003-2004 was, however, marked by a sharp deterioration in key fiscal indicators of states, which was reflected in an increase of around 6 percentage points in average debt-GDP ratio to 26.8% and further to a high of 31.8% in end-March 2004 (Fig. 1a).
In recognition of the need for fiscal discipline, the state governments adopted a rule-based fiscal framework through the enactment of FRBM Acts/FRLs, which also included stipulation of ceilings on total liabilities and in some cases on debt-service liabilities (Goa, Jharkhand and Odisha). Karnataka was the first state to enact its FRBM Act in September 2002, followed by Kerala (2003), Tamil Nadu (2003 and Punjab (2004). Other states also adopted these legislations to avail of the benefits under the incentive scheme recommended by the FC-XII. The adherence to these legislations was also supported by the implementation of Debt Swap Scheme from 2002-2003-2005and Debt Consolidation and Relief Facility from 2005-2006 by the central government. These two debt restructuring schemes provided debt relief through debt consolidation, Fig. 1 Key Fiscal Indicators of State Governments. 1. Ratios pertaining to 'All States' are as percentage to GDP. 2. NSC and SC refer to non-special and special category states, respectively and reduced interest burden on the states. In addition, a turnaround in interest rate cycle also contributed to a gradual reduction in effective interest rates with debt servicing costs declining over time. Reflecting all these developments, the debt position of the state governments improved significantly, as evident from a decline in the average debt-GDP ratio to 22.2% during 2012-2013 to 2015-2016 from around 31% in the last decade and a half. This is in line with the FRBM Review Committee (2017) recommendation of targeting debt to GDP ratio of 20% for the state governments to be achieved by 2023. However, at a disaggregated level, the debt-GSDP ratio was higher than 30% in Kerala, Punjab, Uttar Pradesh and West Bengal while it was above 25% in Bihar, Goa and Rajasthan in the latest period (Table 2). Odisha recorded a remarkable improvement in its debt-GSDP ratio during the period 2004-2005 to 2015-2016.  (1981-1982 to 1991-1992) (1992-1993 to 1996-1997) (1997-1998 to 2003-2004) (

3
Debt sustainability of states in India: An assessment

Assessment of debt sustainability at the state level in India
In the empirical literature, there are primarily two approaches to fiscal (debt) sustainability. The first approach looks at various indicators of the sustainability of fiscal policy (Miller 1983;Buiter 1985Buiter , 1987Blanchard 1990;Buiter et al. 1993) while the second approach involves empirical evaluation or tests of government solvency (Hamilton and Glavin 1986;Trehan and Walsh 1988;Bohn 1998). The empirical testing techniques include determination of sustainable (long-run and maximum sustainable) level of public debt based on a partial equilibrium framework, a modelbased approach and signal approach to fiscal sustainability. Marini and Piergallini (2007), however, suggest an integration of the results from these two approaches so as to provide additional information on the issue of government solvency. While indicators are said to be forward looking, tests are considered backward looking as they are based on historical data. It is the stability of the parameters of the primary surplus equation, which in fact determines the usefulness of results derived from indicators or from tests in the assessment of the sustainability of public debt. It is held that "without a systematic break in policy, the predictions of tests are more reliable since the results of indicators are likely to reflect cyclical factors". This paper has used both indicator-based approach and empirical testing techniques for an assessment of debt sustainability at the state level.

Indicator-based assessment
Traditionally, debt sustainability analysis, under indicator-based assessment, takes into account credit-worthiness indicators (nominal debt stock/own current revenue ratio; present value of debt service/own current revenue ratio) and liquidity indicators (debt service/current revenue ratio and interest payments/current revenue ratio). These indicators broadly enable an assessment of the ability of a state government to service its interest payments and repay its debt as and when they become due through current and regular sources of revenues excluding temporary or incidental revenues as grants or capital revenue resulting from sale of assets. Alternatively, debt and debt-service indicators are monitored to assess relationship of existing debt to different types of expenditures or as ratios to various fiscal balances so as to gauge sustainability of both debt and fiscal situation. An improvement in fiscal conditions creates fiscal space, and enhances debt repayment capacity, while worsening of fiscal conditions entails higher borrowings, adding to the debt burden. In certain situations, the improvement in debt-servicing conditions could also be policy-induced, as discussed in the earlier section. From an analytical point of view, both trends in various fiscal indicators as also characteristics of institutions matter for an assessment of debt sustainability at the state level. In addition, debt sustainability is also associated with a non-financial dimension about the capacity to plan, organise and implement policies, which may be both budget and debt-related.
An analysis based on various indicators of debt sustainability in different phases during the period 1981-1982 to 2015-2016 (Table 3) reveals that the rate of growth  (1981-1982 to 1991-1992) (1992-1993 to 1996-1997) (1997-1998 to 2003 (1981-1982 to 1991-1992), Phase III (1997-1998to 2003 and Phase V (2012Phase V ( -2013Phase V ( to 2015Phase V ( -2016. However, the Domar stability condition that the real rate of interest on debt (i.e., effective interest rate adjusted for inflation) should be lower than the real GDP growth was fulfilled in all the phases except in Phase III when the real rate of interest was almost equal to the real output growth. Here, effective interest rate represents current interest payments as a per cent of outstanding liabilities of state governments in the previous year. Both primary balance and primary revenue balance remained negative in all the phases, even as there was some improvement in primary revenue balance-GDP ratio in the last two phases. Interest payments (average), which had crossed one-fifth of revenue receipts (considered as a tolerable ratio of interest burden, Dholakia et al. 2004) during Phase III, declined to 16.5 and 11.8% of revenue receipts in Phase IV and Phase V, respectively.
The trend in debt-GDP ratio of all states was influenced by the differential between the GDP growth and effective interest rate during the period under review (Fig. 2).
A state-wise position in respect of debt sustainability indicators for 17 non-special category states is presented in Table 4. It may be seen that the rate of growth of GSDP was higher than the effective interest rate in all the states in the last two phases, even as the gap between the two narrowed down in Phase V (Table 4a). Furthermore, the rate of growth of public debt turned out to be higher than the GSDP growth in several states in Phase V, which is a cause of concern (Table 4b). The debt redemption pressure is also evident from the ratio of debt redemption (principal and interest payments) to total debt receipts, which shot up from 64. 1% during 1981-1982 to 2003-2004 to 79.8% during 2004-2005 to 2015-2016. This is indicative of a smaller proportion of borrowed funds being available for productive uses by the state governments during the latter period.
In addition to the debt sustainability indicators as discussed above, it may also be appropriate to analyse debt profile-linked vulnerability indicators viz., spread on state government debt, average maturity and ownership pattern of debt. These indicators provide an idea about liquidity and pricing risks associated with the level of  Table 4 Indicators of Debt Sustainability. a Rate of growth of GSDP (g) should be higher than effective interest rate i; g − i>0, b Rate of growth of public debt (k) should be lower than growth rate of nominal GSDP (g); k − g<0 Source: RBI, various reports of State Finances a Study of Budgets and authors' calculations State (1981-1982 to 1991-1992) (1992-1993 to 1996-1997) (1997-1998 to 2003-2004) (

3
Debt sustainability of states in India: An assessment debt and its composition. From 1988 to 1989 onwards, the weighted average yield on state government securities has been observed to be marginally higher than that on the central government securities. Before this period, these loans were intermediated by the central government. The ownership pattern of state government securities indicates a pre-dominance of commercial banks, although their share in total outstanding state government securities has declined steadily from 78.5% in end-March 1991 to 61.9% in end-March 2000 and further to 42.1% in end-March 2016. The share of insurance companies has, however, increased significantly during the same period. As the state government securities are eligible for being counted towards SLR requirements of banks, investment in these securities is considered credit-risk free. Higher yield on these securities vis-a-vis central government securities is another attraction for long-term investors. The state-specific fiscal performance related risk factors are presumably not being factored in by the investors. However, this situation may not continue for long in case there is any deviation in the extant institutional arrangement for management of state government debt.

Econometric framework for assessment of debt sustainability at state level
The fiscal/debt sustainability exercise, in the empirical literature, is extended beyond the simple indicator-based assessment to validate whether inter-temporal government budget constraint is satisfied. This entails test of stationarity properties of the government debt stock (in level and first difference), examination of the long-term relationship between government revenues and expenditures, between primary balances and debt, and between capital expenditure and public debt (Bhatt 2011). While confirmation of stationarity of government debt stock (in level and first difference) indicates statistical reversion towards mean value after temporary disturbances, the presence of cointegration between government revenues and expenditures reflects their co-movements and anchoring of fiscal imbalances. State (1981-1982 to 1991-1992) (1992-1993 to 1996-1997) (1997-1998 to 2003-2004) (

Inter-temporal budget constraint
In line with the empirical literature, we have made an attempt to test whether the fiscal policy stance of Indian states is sustainable, i.e., whether it satisfies the intertemporal budget constraint. This test basically examines whether the past behaviour of state governments' revenues, expenditure and fiscal deficit could be continued indefinitely without prompting an adverse response from the lenders/investors from/ to whom they borrow/sell securities to meet their resource gap. The inter-temporal budget constraint, under the assumption that the funding of interest payments are not made from the new debt issuances (i.e., no-Ponzi scheme), imposes restrictions on the time series properties of government expenditure and revenues. This requires that government expenditure, revenues and debt stock are all stationary in the first differences. The stationarity property also restricts the extent of deviation of government expenditure from revenues over time. In case government expenditure and revenues are I (1) and cointegrated, then the error correction mechanism would push government finances towards the levels required by the inter-temporal budget constraint and ensure fiscal and debt sustainability in the long term (Olekalns and Cashin 2000).
In this section, to start with, the stationarity properties of state government debt, revenues and expenditure have been tested in a panel data framework. After having done the stationarity test, we have examined whether a long-run equilibrium exists between government expenditure and revenues through panel cointegration tests.

Data
All data on state government expenditure, revenues and outstanding level of debt have been taken from the 'Handbook of Statistics of the Indian Economy', published by the Reserve Bank of India. As already mentioned, the data covers the period 1980-1981 to 2015-2016 for 20 Indian states. A list of the states selected for the present analysis is presented in Appendix 1. Only those states have been selected, for which data on all the relevant variables are available for the entire time period. In the case of Bihar, Uttar Pradesh and Madhya Pradesh, the data on respective fiscal variables from 2000 to 2001 also include data relating to Jharkhand, Uttarakhand and Chhattisgarh, respectively. This has been done to ensure comparability of data for the entire period covered in the econometric exercise. The variables have been converted into real terms and logarithmic values of the variables have been considered for the analysis.

Unit root analysis
As already mentioned, the stationarity properties of state government debt, revenues and expenditure are tested through panel unit root tests. Panel unit root tests are perceived to be more powerful than the unit root test applied on a single series. This is because the information content of the individual time series gets enhanced by that contained in the cross-section data within a panel set up (Ramirez 2006). There are different methods to carry out panel-based unit root tests. While the panel unit root methodology of Levin et al. (2002) assumes that there is a common unit root process across the relevant cross sections, the tests suggested by Im et al. (2003) and Maddala and Wu (1999) assume individual unit root processes.
The results of panel unit root tests on relevant fiscal variables (debt, total revenues and total expenditure) are furnished in Table 5. It may be seen that the tests (Levin, Lin and Chu; Im, Pesaran and Shin; and Maddala and Wu) failed to reject the null hypothesis of a unit root for government revenues and expenditure in level form. The tests, however, reject the null of a unit root in the first difference. The government debt, on the other hand, was found to be stationary in level as per the Levin, Lin and Chu and Im, Pesaran and Shin tests. As per the Maddala and Wu test, however, the government debt turned out to be stationary only in the first difference. Overall, the results reveal that the three variables viz, debt, total revenues and total expenditure are stationary in first difference.

Panel cointegration
Since log R and log G were found to be I (1), in the next step, an attempt has been made to test, whether there exists a long-run equilibrium (steady state) between government expenditure and revenues through the panel cointegration tests. Panel cointegration technique has an advantage over the cointegration tests for individual series as it allows to selectively pool information regarding common long-run relationships from across the panel while allowing the associated short-run dynamics and fixed effects to be heterogeneous across different series of the panel (Pedroni 1999). In this section, the methodology proposed by Pedroni (1999) has been used to test whether a cointegrating relationship exists between government revenues and expenditure of the selected Indian states under study. This method employs four panel statistics and three group panel statistics to test the null hypothesis of no cointegration against the alternative hypothesis of cointegration. In the case of panel statistics, the first-order autoregressive term is assumed to be the same across all the cross sections. On the other hand, in the case of group panel statistics, the parameter is allowed to vary over the cross sections. The statistics are distributed, in the limit, as standard normal variables with a left hand rejection region, with the exception of variance ratio statistics. The results of the cointegration tests are presented in Table 6.
The test results for both the panel and group statistics reveal strong evidence of panel cointegration. The estimated 'rho' statistics, variance ratio 'V' statistics, Augmented Dickey Fuller 't' statistics and the Phillips and Perron (non-parametric) 't' statistics reject the null hypothesis of no cointegration at 1% level for all the three models: (i) model with no deterministic intercept or trend; (ii) model with individual intercept and no deterministic trend; and (iii) model with individual intercept and individual trend. This implies that the cointegration results are not affected by different modelling assumptions.
The results of the Pedroni test are also supported by Kao residual cointegration test, which rejects the null hypothesis of no cointegration at 1% level (Table 7).

3
Debt sustainability of states in India: An assessment Thus, the overall findings of the panel cointegration tests reveal that the two series, government revenues and expenditure are cointegrated, indicating a long-term comovement between them. The results suggest that the current fiscal position of Indian states is sustainable in the long run.

Fiscal policy response function
Bohn (1998), Adams et al. (2010) and Tiwari (2012) have analysed the response of primary surplus to variations in public debt for the purpose of assessment of fiscal policy/debt sustainability. In case primary surplus (relative to GDP) is observed to be a positive function of public debt (relative to GDP), it implies that rising debt ratios lead to higher primary surpluses relative to GDP, which is indicative of a tendency towards mean reversion and thus fiscal/debt sustainability. We have also used this approach in the following analysis.

Model specification
The following equation is estimated in a panel data framework with annual data from 1980-1981 to 2015-2016. In this equation, GSDP is the gross state domestic product; S is the primary balance to GSDP ratio; D is debt to GSDP ratio; GSDPGAP is the deviation of actual output from the trend; EXPGAP is the deviation of actual primary expenditure from the trend; ε is the error term. The business cycle variable GSDPGAP has been included to account for the fluctuations in revenues. The variable EXPGAP captures the impact of deviations of real primary expenditure from its long-term trend on the primary balance ratio. Here 'β' is the key coefficient, which measures the response of primary balance to debt. A value of this coefficient between zero and unity is consistent with a sustainable fiscal policy response to debt. A negative coefficient implies potentially destabilising response. In addition, allowance has been made in the estimations for the response of primary balance to GSDP ratio to be non-linear and vary with debt levels by introducing a square term of the debt to GSDP ratio as an additional explanatory variable.

Data
As in the earlier empirical exercise, the fiscal policy response function has also been estimated for 20 states, for which data on all the relevant variables are available for the period 1980-1981 to 2015-2016. The data for Bihar, Uttar Pradesh and Madhya Pradesh from 2000 to 2001 also include that relating to Jharkhand, Uttarakhand and Chhattisgarh, respectively. Outstanding liabilities of each state government have been used to represent the level of their debt. GSDPGAP for each state has been worked out by extracting the deviation in real GSDP from its trend through HP-Filter. The deviation is expressed as a per cent of real GSDP. EXPGAP has been calculated in a similar manner using real primary expenditure of the state governments. The pair-wise correlation coefficients between the explanatory variables were found to be statistically insignificant, thus ruling out any multicollinearity problem.

Results
Before proceeding with the estimation, all the series were tested for stationarity. Based on panel unit root tests involving common unit root process (LLC) as well as individual unit root process (IPS), the dependent variable and the explanatory variable series were found to be stationary, i.e., I (0). The results of the panel unit root tests are furnished in Table 8.
To decide on the panel models, i.e., whether it is a fixed effect (FE) model or a random effect (RE) model, Hausman test was conducted for each of the two model specifications (linear and non-linear). The summary results of the Hausman test are furnished in Appendix 2. The results of the Hausman test for both the models indicate that there is a significant difference in the coefficients estimated by the FE and RE models. Therefore, the null hypothesis of correlated random effect is rejected and the alternative hypothesis that individual specific effect is correlated with the explanatory variables is accepted. Accordingly, fixed effect model has been chosen for estimating the two model specifications indicated above.  LLC Levin, Lin, Chu (2002), IPS Im, Pesaran, Shin (2003) *The rejection of the null hypothesis of non-stationarity at 1 per cent level of significance The models have been estimated through generalised least square technique with cross section Seemingly Unrelated Regression (SUR) with a correction for first order autoregressive error term. The models are adjusted for the heteroscedasticity with White cross-section standard errors and covariance method. The empirical results from the panel regression exercise are presented in Table 9. In Model 1 (linear model), the coefficients of all the explanatory variables were found to be significant at one per cent level. Positive coefficient of D indicates that the primary balance of state governments increases in response to rising debt ratios. This implies that the primary fiscal balance in India responds in a stabilising manner to increases in debt. Positive coefficient of GSDPGAP implies that primary balance improves when GSDP is above the trend. The negative coefficient of EXPGAP, on the other hand, indicates that the primary balance declines when primary expenditure is above the trend. These findings are in line with a priori expectations.
In the non-linear equation approach (Model 2), allowance was made for the possibility that the response of the primary balance to debt is better represented in terms of a quadratic function rather than a linear response function. The results suggest that the primary balance function has an inverted 'u' shape, implying that the adjustment parameter first rises and then falls.

Going beyond the conventional debt sustainability analysis
In the empirical literature, several studies have gone beyond the conventional debt sustainability analysis in various ways. This has been done by extending the scope of conventional debt analysis (based on the inter-temporal budget constraint in a static environment) to account for fiscal and economic behaviour in response to shocks (sensitivity analysis), fiscal vulnerabilities (stress-testing exercise) and short-term refinancing risks. The interaction of key variables driving debt dynamics is also factored in debt sustainability exercises. There are other studies which have used a more comprehensive concept of debt, covering not only explicit liabilities but also contingent, implicit and off-budget liabilities. After having examined the debt sustainability issue, based on indicator-based approach, inter-temporal budget constraint exercise and fiscal policy response function of states in the earlier sections, an attempt has been made to examine the impact of contingent liabilities on debt/fiscal sustainability of states in India. Article 293 (1) of the Constitution of India provides that a state government can give guarantees within such limits as may be fixed by the state legislature on the security of the Consolidated Fund of the State. Guarantees issued by states are considered as contingent liabilities on the Consolidated Fund of the State in case of default by the borrower for whom the guarantee is extended. The state governments have generally been conservative in the issuance of guarantees (in respect of loans raised by government departments, public sector undertakings, local authorities, statutory boards and corporations, and co-operative institutions) and follow certain norms, whether stipulated under the State Government Guarantees Act or FRBM Acts/FRLs of states or administrative limits fixed for issuance of guarantees. Under these enactments, limits are fixed on annual incremental guarantees as ratio to GSDP or total revenue receipts (Appendix 3). Apart from the differences across states in terms of guidelines relating to guarantees, there are also sharp differences when it comes to awareness about fiscal risk linked to issuance of these guarantees and the state level efforts to reduce outstanding guarantees as a policy initiative.
The guarantee commitments of state governments in respect of state public sector enterprises (SPSEs) have recently emerged as a major source of potential risk to fiscal and debt sustainability at the state level. While the need for issuance of guarantees to SPSEs arose after 1993-1994, when the practice of allocation of a separate share in market borrowings to these enterprises was discontinued, it assumed further importance in the wake of declining budgetary support to these enterprises for meeting their capital requirements. As borrowing requirements of these entities increased, these were backed by issuance of guarantees in several states, resulting in an increase in explicit contingent liabilities of these states. This problem is more acute in those states, which have not enacted any law or framed any rules for fixing the ceiling on guarantees to be given by the state government. On the other hand, there are a few states (Odisha) which have exercised due precaution in putting in place rules to avoid the spill-over effect of these guarantees to state budgets.
The unbridled growth in guarantees issued to SPSEs, which have large outstanding debt and are also incurring losses, have increased vulnerability of these enterprises with fiscal implications for the state governments. This is evident from the data relating to outstanding debt and accumulated losses/profit of SPSEs in end-March 2015 (Table 10). In some states, the outstanding debt of SPSEs is of much larger magnitude than outstanding guarantees issued to these undertakings. On top of this, many SPSEs have accumulated huge losses, which indicate their poor debtservicing capacity, entailing the risk of default in future.
A state-wise picture of outstanding liabilities and guarantee commitments including guarantees issued to power sector companies is given in Appendix 4. While the guarantees outstanding as a per cent of outstanding liabilities of all states was around 16.1% in end-March 2015, it exceeded the all-states average in seven states. Power . 16 states 1 have signed MoUs to take over 75% of outstanding debt of their DISCOMs under the UDAY over a period of 2 years (and in some cases in 5 years) adding to their liabilities and involving additional interest expenditure over a period of 10 years (Appendix 5). Furthermore, these states are also expected to fund the future losses, if any, of DISCOMs in a graded manner and this liability could be as high as 50% of the previous year's loss in the year 2020-2021. These states have issued bonds to the tune of Rs. 2.32 lakh crore, i.e., around 54% of total outstanding debt of DISCOMs, estimated at Rs. 4.3 lakh crore as at end-September 2015. It is, therefore, imperative that the underlying operational efficiency parameters 2 are achieved within the stipulated time frame to bring about a turnaround in financial position of DISCOMs, and to avoid state government participation in such restructuring exercises in future, which may assume crisis proportion. Majority of the states are yet to implement the recommendation of the Group of State Finance Secretaries on the Fiscal Risk on State Government Guarantees (2002) that appropriate risk weights be assigned to guarantees given by states on the basis of probability of devolvement of guarantees, and adequate budgetary provisions be made for honouring these guarantees in case they devolve on the states. The Group had in fact gone a step further by recommending that "guarantees in regard to liabilities which are clearly intended to be met out of budgetary resources should be treated as equivalent to debt." In case, we take outstanding liabilities of states along with their PSUs, the position of some states turns out to be quite alarming on the back of accumulation of losses of PSUs in these states (Appendix 6).

3
Debt sustainability of states in India: An assessment

Conclusion
In this paper, the debt sustainability of state governments in India was assessed through indicator-based analysis as well as empirical exercises. The indicator-based analysis revealed that while most of the debt sustainability indicators showed significant improvement during 2004-2005 to 2015-2016 compared to the earlier phase (1997-1998 to 2003-2004), debt repayment capacity and interest burden indicators lagged behind their respective performance levels achieved during 1981-1982 to 1991-1992. The estimation results based on a panel data framework covering 20 Indian states for the time period 1980-1981 to 2015-2016 revealed that there is a cointegrating relationship between government revenues and expenditure in India, which tantamount to satisfying the inter-temporal budget constraint. Moreover, the estimated fiscal policy response function showed that the primary balance position of Indian states responds in a stabilising manner to the increases in debt levels. Thus, both the results indicate that the current debt situation at the state level is sustainable in the long run.
Disaggregated level analysis, however, revealed that despite an overall improvement in debt position of the Indian states, some of the states have not been able to achieve their respective FC-XIII targets. Going forward, there are several developments with a bearing on debt/fiscal sustainability of states in India. First, the committed liabilities of states may increase in case they decide to implement the Seventh Pay Commission Award, even as some of them have their own pay panels. Second, the interest liabilities of states that have participated in financial restructuring of DISCOMs would increase besides additional provision to be made by them for extending financial support to these utilities in case they continue to incur losses in future as spelled out in MoUs signed by them. Third, the guarantees given to other SPSEs in some states, which are also loss-making entities, could also give rise to financial burden on the states. These dimensions would assume importance in case there is any deviation from the extant institutional arrangement for management of state government debt.
Overall, the conventional debt sustainability analysis, as attempted in this paper, shows that debt position of states at the aggregate level is sustainable. While we have not analysed the implicit liabilities (linked to PPP projects and unfunded liabilities related to pension), our paper highlights that the explicit contingent liabilities linked to guarantees given by the state governments have assumed significance in the context of debt sustainability exercise at the state level. Given this, any debt/fiscal sustainability exercise, based only on outstanding liabilities of states does not provide a realistic assessment of the situation at the individual state level. We would like to conclude with an observation of the RBI Group which was set up to assess the fiscal risk of State Government Guarantees (2002)