Sustainable Debt Policies of Indian State Governments

Abstract

This study analyses the public debt sustainability issue of 20 major Indian states using the Bohn framework for panel data from 2005–2006 to 2014–2015. It employs regular panel data estimation procedures and the penalized spline (p-spline) technique. The results indicate that the primary balance of state governments responds positively to high public debt, so debt policies are successful in sustaining the debt situation of Indian states as a whole. However, at the individual level, debt is sustainable only in 12 states; in 8 states, debt is unsustainable and so these states require corrective action. These findings may be useful to policymakers and other stakeholders to formulate appropriate strategies to improve the debt situation of Indian states.

JEL Classification: E62, H63, H72, H740

Keywords

Primary Balance Sustainable Debt Indian States Bohn Framework

1. Introduction

After the seminal contribution of Hamilton and Flavin (1986), several empirical studies have been conducted on the sustainability of public debt. Three empirical approaches have evolved over the traditional indicator approach. They are the: (a) unit root approach (Caporale, 1995; Trehan & Walsh, 1991; Uctum, Thurston & Uctum, 2006); (b) cointegration approach (Hakkio & Rush, 1991; Jha & Sharma, 2004); and (c) Bohn’s model-based approach (Abiad & Ostry, 2005; Bohn, 1998). In the first approach, the debt series needs to be a mean-reverting process, while in the second approach, the public revenue and public expenditure series need to be cointegrated. The Bohn model suggests that primary surplus relative to GDP is a positive and at least linearly rising function of the public debt-GDP ratio.

The economic intuition behind the Bohn approach is that if governments run into debt today, they have to take corrective actions in future by increasing the primary surplus, otherwise, public debt will not be sustainable (Greiner & Fincke, 2009). This model has been widely used to test whether public debt policies in different countries are sustainable or not (Abiad & Ostry, 2005; Fincke & Greiner, 2011; Greiner & Kauermann, 2008; Haber & Neck, 2006; Kaur, Mukherjee, Kumar & Ekka, 2014; Mahdavi, 2014).¹

The debt sustainability issue is relevant not only for national governments but also for sub-national governments (states) because in countries like India, both governments borrow to finance their deficits. During 2006–2007 and 2007–2008, the centre and states in India (combined) had a primary surplus of 1 per cent of GDP, but in the following year (2008–2009), the combined primary deficit was −3.25 per cent of GDP; in 2009–2010, it rose to −4.53 per cent. The combined debt also increased and reached 88 per cent of GDP (of which the centre’s alone was 61 per cent) in 2005–2006. The aggregated debt of all states rose from ₹11,477 billion in 2005–2006 to ₹27,038 billion in 2014–2015.

In the Indian context, studies, such as Rangarajan, Bhide and Pattnaik (1989), Moorthy, Singh and Dhal (2000) and Pattnaik, Misra and Prakash (2003), used the traditional indicators approach, which basically uses the popular Domar condition.² Studies, such as Buiter and Patel (1990) and Pradhan (2014), employed the unit root approach, while Jha and Sharma (2004) and Tronzano (2013) used the cointegration approach. Tiwari (2012), Kaur and Mukharjee (2012), Jose (2014) and Shastri and Sahrawat (2015) applied the Bohn framework. Only a limited attempt has been made to analyse the issue at the state level.

However, most of the studies have used the Domar condition (Dholakia, Mohan & Karan, 2004; Maurya, 2015; Misra & Khundrakpam, 2009; Rajaraman, Bhide & Pattnaik, 2005). Only Kaur et al. (2014) employ the Bohn framework and show that debt is sustainable for Indian states as a whole. But the question remains whether debt is sustainable for each Indian state, as the aggregate picture may hide specific individual status. Therefore, this study attempts to use the Bohn methodology for panel data to test whether debt is sustainable in 20 major Indian states from 2005–2006 to 2014–2015. It also uses the penalized spline (p-spline) estimation procedure to estimate the time-varying coefficients or reaction coefficients, showing how coefficients associated with the debt-GDP ratio evolve over time across states.³

Thus, this study differs from past studies in the following two respects: (a) This is the first study using the Bohn model for panel data to test the sustainability of debt in 20 major Indian states, individually and provide the state-specific policy suggestions on whether a given debt policy of a state can go on or it needs to be changed; and (b) this is the first study employing the p-spline estimation procedure for a panel of Indian states to show that the reaction coefficient has not stayed constant, but varying across states and time. The rest of this study proceeds as follows. The first section provides a brief note on public debt scenario in India. The third section presents a brief review of literature. While the fourth section explains the data, the model and the variables used in this study, the fifth section presents and discusses the empirical results. The final section gives the policy implications and concluding remarks of the study.

2. Indian Public Debt Scenario

The Indian Constitution (1950) has provided for a two-tier system of government: the centre and states, each with separate tax powers and expenditure functions. As all mobile and buoyant taxes are assigned to the centre, the states generate their own revenues from less buoyant taxes (like state VAT) assigned to them and non-tax and non-debt capital receipts sources to meet their expenditure commitments. At the same time, states are assigned more expenditure functions than the centre. In the centre, the excess federal revenues relative to its responsibility and a corresponding deficit in states’ accounts where expenditures exceed revenues is referred to as the vertical fiscal gap (Rangarajan & Srivastava, 2008). Recognising this vertical imbalance, the Constitution provides for transfer mechanisms to transfer resources from the centre to the states in the form of tax devolution, grant-in aid and centrally sponsored schemes (Rao, 2005).

Both governments borrow if their revenues cannot meet their expenditure needs. The centre in general borrows from internal and external sources,⁴ while states’ internal debts include market loans and bonds, ways and means advances from the central bank (the Reserve Bank of India, popularly known as the RBI), loans from banks and other institutions, provident funds, etc. States’ external debts are subject to a ceiling and approvals from the centre.

Table 1
Centre, States and Combined Primary Balance and Public Debt Ratios, 2004–05 to 2015–16

Year Primary Balance-GDP Ratio
Public Debt-GDP Ratio

Centre States Combined Centre States Combined

2004–2005 0.04 −0.52 −1.25 61.5 26.30 87.90

2005–2006 −0.37 0.03 −0.83 61.2 26.80 88.00

2006–2007 0.18 0.36 0.29 59.11 25.49 84.60

2007–2008 0.88 0.39 1.00 56.90 23.73 80.62

2008–2009 −2.57 −0.53 −3.24 56.11 23.56 79.67

2009–2010 −3.17 −1.22 −4.53 54.49 23.24 77.73

2010–2011 −1.79 −0.39 −2.29 50.6 21.64 72.24

2011–2012 −2.78 −0.36 −3.26 51.71 21.18 72.89

2012–2013 −1.78 −0.45 −2.27 50.95 20.76 71.71

2013–2014 −1.14 −0.71 −1.92 50.30 20.63 70.93

2014–2015* −0.80 −1.34 −2.21 49.62 20.77 70.39

2015–2016# −0.71 −0.79 −1.43 48.73 20.78 69.51

Year	Primary Balance-GDP Ratio	Public Debt-GDP Ratio
2004–2005	0.04	−0.52	−1.25	61.5	26.30	87.90
2005–2006	−0.37	0.03	−0.83	61.2	26.80	88.00
2006–2007	0.18	0.36	0.29	59.11	25.49	84.60
2007–2008	0.88	0.39	1.00	56.90	23.73	80.62
2008–2009	−2.57	−0.53	−3.24	56.11	23.56	79.67
2009–2010	−3.17	−1.22	−4.53	54.49	23.24	77.73
2010–2011	−1.79	−0.39	−2.29	50.6	21.64	72.24
2011–2012	−2.78	−0.36	−3.26	51.71	21.18	72.89
2012–2013	−1.78	−0.45	−2.27	50.95	20.76	71.71
2013–2014	−1.14	−0.71	−1.92	50.30	20.63	70.93
2014–2015*	−0.80	−1.34	−2.21	49.62	20.77	70.39
2015–2016#	−0.71	−0.79	−1.43	48.73	20.78	69.51

Source (Basic data): Indian Public Finance Statistics (various years).

Notes: * and # are the revised and budget estimates, respectively; (−) represents the primary deficit and (+) represents the primary surplus. GDP at market prices (2004–2005 series) was used to compute ratios.

In India, three broad deficit measures are used—the revenue deficit (excess of revenue expenditures over revenue receipts), the fiscal deficit (primary deficit + interest payments = net borrowing) and the primary deficit (the excess of residual non-interest expenditures over total non-debt receipts). Since state governments have constraints on their borrowing sources, they face an inconsistency between their borrowing requirements and debt servicing, but have annual debt requirements in the form of interest obligations on accumulated debt. The extent of these commitments every year is reflected in the primary balance. It is basically the amount of borrowings that are required to meet expenses other than the interest payments (primary deficit) or the pressure of the government on the interest commitments on previous borrowings to borrow (primary surplus). Therefore, the primary balance is the source of all types of deficits and it improves or worsens the fiscal situation due to total debt requirements.

Governments face two types of motivations to expand their activities above trend levels: to minimise the impact or volatility of the cyclicality of growth and to finance their political agendas (Srivastava, 2012). Trends in the primary deficit relative to the GDP and public debt relative to GDP since Independence indicate the cyclical nature of the former and the secular upward nature of the later (Rangarajan & Srivastava, 2005). Since the second half of 1990s, there has been a sharp deterioration in the debt-deficit situation of both centre and states. To reduce debt to a sustainable level, the centre adopted a rule-based fiscal framework (FRBM) in 2003 and most states enacted FRBM rules in 2005–2007,⁵ with some initial success.⁶

Figure 1

Source: Author’s construction.

The situation has worsened since the global slowdown in 2007–2008: fiscal consolidation has been completely reversed due to a combination of spending measures, soaring subsidy bills and fiscal stimulus packages in response to crises, etc. (Tiwari, 2012) (Table 1).

The initial minimal combined primary deficit became surplus in 2006–07 and 2007–08, mainly because of the FRBM Act and other fiscal consolidation measures like the debt swap scheme (in which the 12th Finance Commission allowed states to swap high-cost loans against open-market borrowings and small savings in 2002–04), and the debt write-off scheme (in which the 13th Finance Commission offered a full write-off if states had a zero revenue deficit in 2008–2009 on their debt repayment to the centre and a concession on their interest rate). However, after the global crisis, the combined primary deficit reached an alarming 4.53 per cent in 2009–10, of which the centre’s primary deficit alone was 3.17 per cent.

As most states managed to fulfil their FRBM norms, their debt-GSDP (gross state domestic product) ratios declined from 26.8 per cent in 2005–2006 to 20.8 per cent in 2015–2016 but remain over 25 per cent in seven states, including Jammu and Kashmir (54.9%), Himachal Pradesh (41.9%) and West Bengal (34.7%) (Figure 1). If the current trend continues, many states will undergo adverse conditions. As implementations of the Fifth and Sixth Pay Commission recommendations have increased the revenue deficits and debt positions of many states, the proposed Seventh Pay Commission recommendations may be expected to add fuel to the debt positions of many state governments (Kurian, 1999; Rajaraman et al., 2005). Further, the rolling out Goods and Service Tax (GST) may result in a revenue loss in some states, which will in turn affect their debt sustainability.

Given the above trends in debt, this study considers the period from 2004–2005 to 2014–2015 because this is the post-FRMB period as well as represents the fiscal control era (by 12th and 13th FCs). In addition, various socio-economic-political and dynamic factors emerged in this period. Debt is in fact an accumulation of fiscal deficit every year and so the recent trend on debt is more concern than the past trend.

3. Brief Review of Literature

Debt sustainability is maintained as long as debt does not accumulate at a rate far exceeding the government’s capacity to service it. Unsustainable debt levels can lead to major disruptions in economic activity and reorientation of priorities in an economy. There are three theoretical views on debt financing: the neoclassical view is that a fiscal deficit is detrimental to investment and economic growth, while the Keynesian stance is that a deficit has a growth-stimulating effect. Debt does not pose a problem if the governments run into debt in the home country, as resources are not lost and public deficits imply a reallocation of resources from tax payers to bond holders (Greiner & Fincke, 2009). The third view, the Classical Ricardian equivalence theorem asserts that fiscal deficits do not matter, except for smoothing adjustments to expenditure or revenue shocks. Their stance is that budget deficits today will require higher taxes in future if government cuts taxes without changing present or future public spending. Given that households are forward looking, they realise they need to pay higher taxes in future so their total tax burden remains unchanged, and they reduce their consumption and increase savings to meet their future tax burden. This theorem is based on the inter-temporal budget constraint of the government and on the permanent income hypothesis. Thus, there is no consensus among economists on whether deficit financing is good, bad or neutral (Rangarajan & Srivastava, 2005).

Since the Keynesian view dominated in the 1970s, the public debt rose considerably over the period and was often was even higher than the GDP growth in many countries, so that the ratio of public debt to GDP increased, too. This raised the question of sustainability of the time-path of public debt,⁷ and several studies emerged to address this question starting with the seminal paper by Hamilton and Flavin (1986). Earlier studies had mostly employed the Domar (1944) condition, according to which three conditions emerge from the basic debt accumulation equation:

d_{t} = p_{t} + d_{t - 1} [(1 + i_{t}) / (1 + g_{t})]

(1)

where d_t is the debt-GDP ratio in period t; g is the nominal economic growth rate; i is the nominal interest rate; p is the primary deficit relative to GDP in period t; and the conditions are g_t = i_t, g_tt and g_tt. Fiscal policy is unsustainable when g_t= i_t or g_t_t, because d_t grows linearly when g_t =i_tand explosively when g_t _t. Debt is sustainable when g_t_t. The last condition is considered a necessary condition for sustainability, based on the assumption that the faster income grows the lighter will be the burden of debt.

This approach was extended with additional indicators (growth, liquidity, credit worthiness, fiscal burden, fiscal space, etc.) and renamed the ‘indicator approach’ (Blanchard, Chouraqui, Hagemann & Sartor, 1990; Kaur et al., 2014; Maurya, 2015; Misra & Khundrakpam, 2009; Pattnaik et al., 2003; Rajaraman et al., 2005). As it applied the condition every year and is not sufficient enough to validate whether the inter-temporal budget of the government is satisfied or not, many proposed an econometric or statistical validation to substantiate the sustainability conditions.

Hamilton and Flavin (1986) carried out the first study using the unit root test to check whether the public debt series (D_t) in the USA is stationary or not, followed by Trehan and Walsh (1991). They use augmented Dickey–Fuller (ADF) statistics to test the hypothesis that the given series is non-stationary (H₀) against the alternative hypothesis that it is stationary (H_a). Trehan and Walsh (1991) used another test to analyse whether a quasi-difference of public debt (D_t – vD_t_–1) with 0 ≤ v < 1 + r (where r is the interest rate) is stationary, and whether public debt and primary surpluses (S_t) are cointegrated. If government debt is quasi-difference stationary and public debt and primary surpluses are cointegrated, then the public debt is sustainable (Greiner & Fincke, 2009). (See Afonso (2005) for a brief survey of studies employing these procedures.)⁸

These approaches have been criticised for their limitations: (a) the unit root test is very sensitive to structural breaks and the results could be misleading (Uctum et al., 2006); (b) it is very difficult to reject a unit root in real debt or in the debt-GDP ratio; and (c) rejection of sustainability based on these two test are invalid because the inter-temporal budget condition (IBC) may well be satisfied, even if the components of the budget are not cointegrated and even if debts or deficits, revenues or spending are difference stationary (Bohn, 2007).

The IBC is $d_{t}^{*} = \sum_{j = 1}^{\infty} \frac{1}{{(1 + r)}^{j}} E_{t} [s_{t + j}],$ where $d_{t}^{*}$ = (1 + r_t). d_t_–1 is the stock of the debt-output ratio in the beginning of period t, E_t[.] denotes the expectation operator conditional on the information available at time t, and s_t is the primary surplus-GDP ratio. The IBC of the government requires that the present value of public debt asymptotically converges to zero, and the interest rate r is resorted to discount the stream of public debt and plays an important role. Thus, both the unit root and cointegration tests are independent of the interest rate.

Bohn (1998) proposed a model-based approach to test whether the primary surplus-GDP ratio (s_t) is a positive and, at least, a linearly rising function of the debt-GDP ratio (d_t):

s_{t} = α_{0} + α_{1} d_{t} + ε_{t}

(2)

If this property holds, debt is sustainable. That is, if α₁ > 0 and statistically significant, debt is sustainable, which means that the initial stock of debt is equal to the sum of the present discounted values of the primary surpluses. The IBC is satisfied if the discounted sum of end-period debt converges to zero. The positive reaction coefficient, α₁ensures this convergence.

Bohn (1998) utilises Barro’s (1979) tax-smoothening hypothesis according to which public deficits should be used in order to keep tax rates constant which minimises the excess burden of taxation. Hence, normal expenditure can be financed by regular revenues, and deficits will be incurred to finance unexpected spending. He derives the following fiscal rule or reaction function:

s_{t} = α_{0} + α_{1} d_{t} + ϕ_{1} Y V A R_{t} + ϕ_{2} G V A R_{t} + ε_{t}

(3)

where YVAR and GVAR are business cycle indicators. YVAR accounts for fluctuations in revenues and it gives the deviation of real GDP from its trend, computed using the Hodrick–Prescott (HP) filter. Positive values for YVAR indicate booms and negative values indicate recessions. GVAR gives deviation of real primary spending from its normal value with positive values indicating expenditures above the normal level and vice versa (Greiner & Fincke, 2009).

As the relationship may not be linear and in order to bring this about, the above Bohn model is modified as:

s_{t} = α_{0} + α_{1 t} d_{t} + ϕ_{1} Y V A R_{t} + ϕ_{2} G V A R_{t} + ε_{t}

(4)

where the reaction coefficient α₁ is time-varying. It is, however, noticed that any non-linear model can be approximated by a linear model with time-varying coefficients. The approximation is good if it changes smoothly. So empirical estimation resorts to spline (a type of smoothing technique that allows us to analyse the data in a more flexible way).⁹ The functional form or smoothness is shaped by deviations on individual points (i.e., changing points which are termed knots). A p-spline estimation technique is used to estimate the Equation (4). To avoid the endogeneity issue, d_t is replaced with d_t−₁; α₁ is the average coefficient and the actual coefficient is the sum of α₁ and the deviation, which is given by the smooth function, sm(t). If α₁ is positive, there is an indication of debt sustainability, and the time-varying values indicate a change in the response coefficient over the years.

Bohn (1998) is a pioneering study capturing the significance of reaction by analysing the behaviour of the US public debt and deficit (using the Ordinary Least Square Method (OLS) procedure). Abiad and Ostry (2005) extend the basic Bohn framework by adding the extra determinants of a primary balance ratio and panel data of 31 emerging market countries from 1990 to 2002. They have used the panel random effects (RE) procedure, including the spline for debt at a threshold level of 50 per cent of GDP. Haber and Neck (2006) investigate the sustainability of Austrian fiscal policy from a political economy perspective by incorporating certain political variables. Greiner and Kauermann (2008) incorporate the time-varying parameters in the regression (i.e., spline technique) for the European context. Later many studies have used the spline procedure to analyse the debt sustainability issues of various countries (see Fincke & Greiner (2011) for a review of these studies).

In the Indian context, Kaur and Mukharjee (2012) have estimated the fiscal policy response function using the OLS procedure and found evidence of a positive response of the primary surplus ratio to the increasing debt ratio for the combined (centre and states together) data from 1980–81 to 2012–13. By employing the central government’s primary balance and public debt data for the period during 1983–2010 and an OLS procedure, Jose (2014) has generated a similar result. Shastri and Sahrawat (2015) use all three approaches (unit root, cointegration and Bohn) and find that the central government’s revenue and expenditure are not cointegrated. Tiwari (2012) is the only study employing the Bohn framework with spline methodology for national-level (combined) data from 1970 to 2009, but was unable to find clear-cut evidence on sustainability of the public debt.

Table 2

Summary Results of Important (Selective) Studies on Debt Sustainability

Study	Country	Data Period	Variables/Methodology	Is Debt Sustainable
(i) Unit Root Approach
Hamilton and Flavin (1986)	USA	Annual; 1992–1984	Public debt, deficit	Sustainable
Trehan and Walsh (1991)	USA	Annual; 1890–1983	Deficit, public debt	Sustainable
Caporale (1995)	EU countries	Semi-annual and annual; 1960–1991	Deficit, public debt	No
(ii) Cointegration Approach
Hakkio and Rush (1991)	USA	Quarterly; 1950:II to 1988:IV	Revenue and expenditure	No
Payne, Mohammadi and Cak (2008)	Turkey	Annual; 1990–2008	Revenue and expenditure	No
Tronzano (2013)	India	Annual; 1950–2010	Revenue and expenditure	Weakly sustainable
Kaur et al. (2014)	20 Indian states	Annual; 1980–2013	Revenue, expenditure and debt (panel test)	Sustainable
(iii) Bohn Model
Bohn (1998)	USA	Annual; 1916–1995	OLS	Sustainable
Haber and Neck (2006)	Austria	Annual; 1962–2000	OLS	Sustainable
Doi, Hoshi and Okimoto (2011)	Japan	Quarterly (1980:I to 2010:I)	Markov-Switching model	No
Shastri and Sahrawat (2015)	India	Annual; 1980–2013	OLS	No
Greiner and Kauerman (2008)	Germany and Italy	Annual; 1960–2003 and 1975–2003	p-spline	Sustainable (only Germany)
Fincke and Greiner (2011)	Euro countries	Annual; 1971–2009	p-spline	Sustainable (except for Greece and Italy)
Tiwari (2012)	India	Annual; 1970–2009	p-spline	No
Abiad and Ostry (2005)	31 emerging countries	Annual; 1990–2002	Panel GLS, Arellano Bond	Sustainable
Adams, Ferrarini and Park (2010)	33 Asian countries	Annual; 1990–2008	Panel GLS	Sustainable
Mahdavi (2014)	48 US states	Annual; 1961–2008	Panel FE	Sustainable
Kaur et al. (2014)	20 Indian states	Annual; 1980–2013	Panel FE	Sustainable

Source: Author’s construction.

A few studies have dealt with debt sustainability at the sub-national level. For instances, Fincke and Greiner (2011) use the Bohn framework and spline technique to test the debt sustainability of each state in Germany. Employing a panel version (FE model) of the Bohn framework, Mahadavi (2012) analysed the debt sustainability position of 48 US states from 1961 to 2008 and showed that their debt is sustainable. Kaur et al. (2014) used all three empirical models for a panel of 20 major Indian states from 1980–1981 to 2012–2013 and found evidence of sustainability of the aggregate debt position in the long run.

Table 2 provides summary results of a few, selective studies on the topic. It is noticed that none of existing studies in Indian context use the Bohn model to test sustainability at the individual state level. The present study is an attempt to fill this gap.

4. Model, Data and Estimation

In order to test the sustainability of public debt in Indian states, this study employs the following extended version of the Bohn framework for panel data:

s_{i t} = ϕ_{0} + ψ d_{i t - 1} + ϕ_{1} y {var}_{i t} + ϕ_{2} g {var}_{i t} + λ_{i} + μ_{t} + i t

(5)

where s_it is the primary surplus-GSDP ratio for the ith state in tth time period; d_it_–1is the debt-GSDP ratio for the ith state in t − 1 period; and yvar_it and gvar_it are business cycle variables to account for fluctuations in GSDP and primary public spending, respectively. They are calculated by subtracting the long-term trend of the GSDP (real) from its realised values and the long-term trend of primary spending of the government (real) from its realised values. The long-term trends of the respective variables are computed using the HP filter. λ_i and µ_t are the individual (states’) effects and time effects (year), respectively. It is noticed that the lagged debt ratio is used to take into account the endogeneity issue. If ψ > 0 and statistically significant, debt is sustainable.

Table 3

Descriptive Statistics of the Study Variables (2005–06 to 2014–15)

Definition	Variables	Mean	Standard Deviation
Primary balance (` billion)	S_it	−18.169	36.550
Primary balance-GSDP ratio (%)	s_it	−0.368	1.499
Public debt (` billion)	D_it	869.943	696.781
Debt-GSDP ratio (%)	d_it	30.516	10.953
Nominal GSDP (` billion)	ngsdp_it	3189.352	2825.196
Real GSDP (` billion)	rgsdp_it	2111.696	1713.796
Real GSDP gap (` billion)	yvar_it	0.00005	4105.348
Primary expenditure (` billion)	pe_it	469.516	370.504
Real primary expenditure (` billion)	rpe_it	308.417	212.585
Real primary expenditure gap (` billion)	gvar_it	0.00002	1626.539

Source (basic data): CSO and CAG reports (computed by the authors).

Table 4

Results of Panel Unit Root Tests

Variables	LLC t statistics	IPS w statistics
s_it	−14.966*	−2.353*
d_it	−5.364*	−3.374*
yvar_it	−7.934*	−2.549*
gvar_it	−7.392*	−3.099*

Source: Author’s construction.

Notes: * indicates rejection of the null hypothesis of the panel containing unit roots (non-stationarity) at the 1% level of significance; LLC—Levin, Lin and Chu test; IPS—Im, Pesaran and Shin test.

Table 5

Panel Model Estimation Results of Bohn Framework for Indian States (Dependent Variable: Primary Surplus to GSDP Ratio)

Variables	Model 1	Model 2
yvar_it	0.00002 (0.880)	0.00002 (0.830)
gvar_it	−0.0002 (−4.670)	−0.0002 (−4.690)
d_it– ₁	0.1195 (8.050)	—
d_it_–1× Dummy for Andhra Pradesh (K₁)		0.0320 (0.360)
d_it_–1× Dummy for Assam (K₂)		0.3698 (4.420)
d_it_–1× Dummy for Bihar (K₃)		0.0577 (1.970)
d_it_–1× Dummy for Chhattisgarh (K₄)		0.2213 (3.310)
d_it_–1× Dummy for Gujarat (K₅)		0.0877 (0.940)
d_it_–1× Dummy for Haryana (K₆)		0.5428 (4.660)
d_it_–1× Dummy for Himachal Pradesh (K₇)		0.1648 (4.740)
d_it_–1× Dummy for Jammu Kashmir (K₈)		0.2192 (1.790)
d_it_–1× Dummy for Jharkhand (K₉)		0.2575 (2.100)
d_it_–1× Dummy for Karnataka (K₁₀)		−0.1634 (−0.960)
d_it_–1× Dummy for Kerala (K₁₁)		0.4052 (2.430)
d_it_–1× Dummy for Madhya Pradesh (K₁₂)		0.1111 (1.400)
d_it_–1× Dummy for Maharashtra (K₁₃)		−0.0239 (−0.350)
d_it_–1× Dummy for Odisha (K₁₄)		0.2516 (4.340)
d_it_–1× Dummy for Punjab (K₁₅)		0.1587 (2.890)
d_it_–1× Dummy for Rajasthan (K₁₆)		0.0634 (1.590)
d_it_–1× Dummy for Tamil Nadu (K₁₇)		0.3133 (1.880)
d_it_–1× Dummy for Uttar Pradesh (K₁₈)		0.0232 (0.470)
d_it_–1× Dummy for Uttarakhand (K₁₉)		0.1366 (2.540)
d_it_–1× Dummy for West Bengal (K₂₀)		0.0532 (0.740)
State effects	Included	Included
Adjusted R²	0.427	0.530
F statistics	7.740	6.470
Hausman statistics	52.600	47.040

Source: Estimated by the authors.

Notes: Total no. of observations (N): 200; figures in parenthesis are t values.

Table 6

Penalized Spline Estimation Results of Bohn Model for Indian States

Variables	Panel within Estimation	Spline Estimation
Variables	(1)	(2)
$d_{i t - 1} - {\overset{─}{d}}_{i}$	0.1195 (8.050)	0.0927 (2.811)
$y {var}_{i t} - {\overset{─}{y var }}_{}$	0.00002 (0.880)	0.00002 (0.948)
$g {var}_{i t} - {\overset{─}{g var}}_{}$	−0.0002 (−4.670)	−0.0002 (−4.922)
sm(it)	—	edf: 2.577
ρ-value		0.0451
F Statistics	29.11	2.901
Adj-R²	0.3304	0.333

Source: Author’s construction.

Note: t values are in parentheses.

Equation (5) can be estimated using the standard panel data methodologies: fixed effects (FE) and RE. The former posits that the unobserved heterogeneity factors, λ_i_, and time effects, µ_t, are correlated with other X variables in the equation, while the latter assumes that they are not. The choice of a relevant model depends on the Hausman statistics. If it supports the FEs model, then OLS can be used to estimate equation (5) by incorporating λ_i and µ_t with the state and year dummies. If the time dummies are jointly zero, then the model is a one-way FE model. If the Hausman supports the RE model, the GLS estimation procedure can be used.

Figure 2

Source: Author’s construction.

In order to estimate the time-varying (and state-specific) estimates, this study also uses the p-spline estimation.¹⁰ This allows the reaction coefficient ψ to be a function of time, showing how that coefficients evolve over time and across states. The study uses the following within-estimation specification to employ the p-spline estimation:

s_{i t} - \overset{─}{s_{i}} = ψ (i t) (d_{i t} - \overset{─}{d_{i}}) + \emptyset_{1} (y {var}_{i t} - \overset{─}{y {var}_{i}}) + \emptyset_{2} (g {var}_{i t} - \overset{─}{g {var}_{i}}) + \in_{i t}

(6)

where $s_{i t} - \overset{─}{s_{i}}$ is the difference of the primary surplus and the GSDP ratio of each state i in year t from its mean value for state i. The mean differences of d_it−₁, yvar and gvar of each state are also constructed using a similar procedure. This ‘within-estimation’ procedure wipes out the individual (λ_i) and time (µ_t) effects. If the reaction function is positive and significant, the debt of the Indian states is sustainable as a whole. Both these procedures would reveal whether the debt is sustainable or not for Indian states as a whole. However, setting the within model estimates using the p-spline method would reveal in addition how the debt ratio coefficients evolve over time and across states. That is, ψ (it) is both state specific and time specific.

To check whether debt is sustainable in each sample state, we allow d_it₋₁ to interact with each of the state dummies (K_i) in Equation (6) to get:

S_{i t} = ϕ_{0} + Σ_{i} ψ ​_{i} K_{i} * d_{i t - 1} + ϕ_{1} y {var}_{i t} + ϕ_{2} g {var}_{i t} + λ_{i} + μ_{t} + ε_{i t}

(7)

The coefficients associated with these interaction terms (ψs) would directly reveal whether debt is sustainable in each state.

To estimate the above equations, the study uses data compiled from various secondary sources for 20 major Indian states: Andhra Pradesh, Assam, Bihar, Chhattisgarh, Gujarat, Haryana, Himachal Pradesh, Jammu Kashmir, Jharkhand, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Odisha, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh, Uttarakhand and West Bengal. State-wise GSDP (real and nominal) is compiled from the Central Statistical Organization (CSO), Ministry of Statistics and Programme Implementation (MOSPI), Government of India website (http://www.mospi.gov.in/data) and all the other fiscal variables from the Comptroller and Auditor General (CAG) of India audit reports and the finance accounts of the state governments for the period from 2004–2005 to 2014–2015. Since the lagged debt ratio is included, the total observations in the final analyses sum up to 200 (20 states × 10 years). Table 3 presents the descriptive statistics of the study variables. Table 4 shows the panel unit root tests using the Levin, Lin and Chu (LLC) (2002) hypothesis and the Im, Pesaran and Shin (IPS) (2003) results. Both test results confirm that all the variables used in the study are stationary, that is, they are I(0).

5. Empirical Results

5.1 Bohn’s Sustainability Analysis Results

Table 5 presents the estimation results of Equation (5) (Model 1). The Chow test and Hausman statistics support the one-way FE model. The business cycle variable yvar is positive as expected, but not statistically significant even at the 10 per cent level. The primary expenditure gap variable gvar has a negative coefficient and is statistically significant at the 1 per cent level, implying that primary spending above its normal value has reduced the primary surplus ratio. The variable of interest is d_it–₁. Its coefficnt is positive and statistically significant at the 1 per cent level, indicating the sustainability of public debt in Indian states as a whole. It is noticed that the dependent variable is the primary balance (which may be positive or negative) and when on average the debt-GDP ratio increases by 1 unit, the primary balance-GDP ratio increases by 0.1195 unit. Thus, the latter is a linear and positive function of the former.

5.2 Penalized Spline Estimation Results

Column (1) of Table 6 shows the panel within estimation results for comparative purposes. These results are exactly the same as the FE model results shown in Model 1 of Table 5. The p-spline estimation results are shown in Column 2 of Table 6. The estimated parameter of interest associated with the debt ratio, ψ(it) represents the mean of this coefficient, and the smooth term sm(it) shows the deviation from that mean over individual and time-varying coefficients. The results indicate that for Indian states, the reaction coefficient has been positive on average and statistically significant at the 1 per cent level so that the public debt is sustainable. The yvar is not statistically significant even at 10 per cent level; gvar has a negative and significant parameter, implying that public (primary) spending (real) above its normal value has reduced the primary surplus ratio.

The estimated degrees of freedom (edf) of sm(it) provides information on possible time and state dependencies. As the estimated value of edf = 2.58 and the smooth term sm(it) is significant at the 5 per cent level, we may conclude that the reaction coefficient has not stayed constant across states and over time. Figure 2 shows the path of the smooth term; the curve is drawn such that values larger (smaller) than zero indicate that the coefficient was above (below) its average value shown in Column 2 of Table 6.¹¹

5.3 Debt Sustainability Results for Individual States

In order to check whether public debt is sustainable in each state, we estimate Equation (7) by allowing the d_it–1 variable to interact with each state dummy. The Chow test and Hausman statistics support the one-way FE model. The estimated results are shown in Table 5 (Model 2). As in Model 1, yvar is not statistically significant and gvar has a negative and significant coefficient. The coefficient of the lagged debt to GSDP ratio and state dummy interaction term ψ is positive and statistically significant for Assam, Bihar, Chhattisgarh, Haryana, Himachal Pradesh, Jharkhand, Kerala, Odisha, Punjab and Uttarakhand, indicating that their public debt is sustainable. For Jammu and Kashmir and Tamil Nadu, the coefficients of debt interaction term are positive but statistically significant only at the 10 per cent level implying that debt is sustainable for these two states too. For Andhra Pradesh, Gujarat, Madhya Pradesh, Rajasthan, Uttar Pradesh and West Bengal, the debt interaction coefficient is positive but not statistically significant even at the 10 per cent level, and for Karnataka and Maharashtra the coefficient is negative but not significant. Therefore, debt is not sustainable in these eight states, so they need corrective actions to make their debt sustainable.

6. Summary and Conclusion

This study has analysed empirically debt sustainability in 20 major Indian states from 2005–06 to 2014–15, using the extended Bohn sustainability framework for panel data. It has employed three estimation methods: (a) the panel FE model to test whether debt is sustainable in the states as a whole; (b) the panel FE model, including the debt-state dummy interaction term, to test whether debt is sustainable in each state; and (c) the p-spline method to obtain state-specific and time-specific response effects of the debt ratio.

The results indicate that the primary balance of state governments in India reacts (responds) to high public debt as predicted in the Bohn framework. This means that debt policies from 2005–06 to 2014–15 were in general successful in sustaining the debt situation of states as a whole. However, the situation differs by state. Only in 12 out of 20 states, namely, Assam, Bihar, Chhattisgarh, Haryana, Himachal Pradesh, Jammu and Kashmir, Jharkhand, Kerala, Odisha, Punjab, Tamil Nadu and Uttarakhand, is public debt sustainable. The remaining eight states need to take corrective action to tackle their debt situation. One similarity we find in the eight states where debt is unsustainable is that their absolute amount of public debt was relatively high compared to the other states, and ranged between ₹1,600 billion to ₹3,200 billion in 2014–15. We hope these results are useful to policymakers, academicians and other stakeholders in taking appropriate measures to improve the debt situation of states with unsustainable levels of debt.

Footnotes

Notes

References

Abiad

M. A.

, & Ostry

M. J. D.

(2005). Primary surpluses and sustainable debt levels in emerging market countries (IMF Policy Discussion Paper No. 05(6)). Washington, D.C.: Research Department, International Monetary Fund.

Adams

, Ferrarini

, & Park

(2010). Fiscal sustainability in developing Asia (Asian Development Bank Economics Working Paper Series No. 205). Philippines: Economics and Research Department, Asian Development Bank.

Afonso

(2005). Fiscal sustainability: The unpleasant European case. FinanzArchiv: Public Finance Analysis, 61(1), 19–44.

Barro

R. J.

(1979). On the determination of the public debt. Journal of Political Economy, 87(5, Part 1), 940–971.

Blanchard

, Chouraqui

J. C.

, Hagemann

R. P.

, & Sartor

(1991). The sustainability of fiscal policy: New answers to old questions. Economic Studies (OECD), 15(Autumn), 1–36.

Bohn

(1998). The behavior of US public debt and deficits. Quarterly Journal of Economics, 113(3), 949–963.

Bohn

(2007). Are stationarity and cointegration restrictions really necessary for the intertemporal budget constraint? Journal of Monetary Economics, 54(7), 1837–1847.

Buiter

W. H.

, & Patel

U. R.

(1992). Debt, deficits, and inflation: An application to the public finances of India. Journal of Public Economics, 47(2), 171–205.

Caporale

(1995). Bubble finance and debt sustainability: A test of the government’s intertemporal budget constraint. Applied Economics, 27(12), 1135–1143.

10.

Dholakia

R. H.

, Mohan

T. R.

, & Karan

(2004). Fiscal sustainability of debt of states. New Delhi: Study sponsored by the Twelfth Finance Commission.

11.

Doi

, Hoshi

, & Okimoto

(2011). Japanese government debt and sustainability of fiscal policy. Journal of the Japanese and International Economies, 25(4), 414–433.

12.

Domar

E. D.

(1944). The ‘burden of the debt’ and the national income. American Economic Review, 34(4), 798–827.

13.

Fincke

, & Greiner

(2011). Debt sustainability in selected euro area countries: Empirical evidence estimating time-varying parameters. Studies in Nonlinear Dynamics & Econometrics, 15(3), 1–21.

14.

Greiner

, & Kauermann

(2007). Sustainability of US public debt: Estimating smoothing spline regressions. Economic Modelling, 25(6), 350–364.

15.

Greiner

, & Kauermann

(2008). Debt policy in euro area countries: Evidence for Germany and Italy using penalized spline smoothing. Economic Modelling, 25(6), 1144–1154.

16.

Greiner

, & Fincke

(2009). Public debt and economic growth (Vol. 11). Germany: Springer Science and Business Media.

17.

Haber

, & Neck

(2006). Sustainability of Austrian public debt: A political economy perspective. Empirica, 33(2–3), 141–154.

18.

Hakkio

C. S.

, & Rush

(1991). Is the budget deficit too large? Economic Inquiry, 29(3), 429–445.

19.

Hamilton

, & Flavin

(1986). On the limitations of government borrowing: A framework for empirical testing. American Economic Review, 76(4), 808–819.

20.

K. S.

, Pesaran

M. H.

, & Shin

(2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115(1), 53–74.

21.

Jha

, & Sharma

(2004). Structural breaks, unit roots, and cointegration: A further test of the sustainability of the Indian fiscal deficit. Public Finance Review, 32(2), 196–219.

22.

Jose

(2014). Sustainability analysis of public debt in India. DEECEE School Journal, 5(1), 101–109.

23.

Kaur

, & Mukherjee

(2012). Threshold level of debt and public debt sustainability: The Indian experience (RBI Occasional Papers No. 33(1&2)). Mumbai: Reserve Bank of India.

24.

Kaur

, Mukherjee

, Kumar

, & Ekka

A. P.

(2014). Debt sustainability at the state level in India (RBI Working Paper Series No. 07). Mumbai: Department of Economics and Policy Research, Reserve Bank of India.

25.

Kurian

N. J.

(1999). State government finances: A survey of recent trends. Economic and Political Weekly, 34(19), 1115–1125.

26.

Levin

, Lin

C. F.

, & Chu

C. S. J.

(2002). Unit root tests in panel data: Asymptotic and finite-sample properties. Journal of Econometrics, 108(1), 1–24.

27.

Mahdavi

(2014). Bohn’s test of fiscal sustainability of American state governments. Southern Economic Journal, 80(4), 1028–1054.

28.

Maurya

N. K.

(2015). Debt sustainability of a sub-national government: A case study of Uttar Pradesh in India. Journal of Economic Policy and Research, 11(1), 126.

29.

Misra

B. M.

, & Khundrakpam

J. K.

(2009). Fiscal consolidation by central and state governments: The medium-term outlook. Mumbai: Reserve Bank of India Staff Studies.

30.

Moorthy

, Singh

, & Dhal

S. C.

(2000). Bond financing and debt stability: Theoretical issues and empirical analysis for India (Development and Research Group Studies Series). Mumbai: Reserve Bank of India.

31.

Pattnaik

R. K.

, Misra

B. S.

, & Prakash

(2003). Sustainability of public debt in India: An assessment in the context of fiscal rules (6th Workshop on Public Finance). Italy: Bank of Italy.

32.

Payne

J. E.

, Mohammadi

, & Cak

(2008). Turkish budget deficit sustainability and the revenue-expenditure nexus. Applied Economics, 40(7), 823–830.

33.

Pradhan

(2014). Is India’s public debt sustainable? South Asian Journal of Macroeconomics and Public Finance, 3(2), 241–266.

34.

Rajaraman

, Bhide

, & Pattnaik

R. K.

(2005). A study of debt sustainability at the state level in India. Mumbai: Reserve Bank of India staff studies.

35.

Rangarajan

, Basu

, & Jadhav

(1989). Dynamics of the interaction between government deficit and domestic debt in India. RBI Occasional Papers, 10(3), 163–220.

36.

Rangarajan

, & Srivastava

D. K.

(2005). Fiscal deficits and government debt: Implications for growth and stabilization. Economic and Political Weekly, Special article, 40(27), 2919–2934.

37.

Rangarajan

, & Srivastava

D. K.

(2008). Reforming India’s fiscal transfer system: Resolving vertical and horizontal imbalances. Economic and Political Weekly, Special article, 43(23), 47–60.

38.

Rao

(2005). Changing contours in fiscal federalism in India. New Delhi: National Institute of Public Finance and Policy.

39.

Ruppert

, Wand

M. P.

, & Carroll

R. J.

(2003). Semiparametric regression. Cambridge, UK: Cambridge University Press.

40.

Shastri

, & Sahrawat

(2015). Fiscal sustainability in India: An empirical assessment. Journal of Economic Policy and Research, 10(1), 97–112.

41.

Srivastava

D. K.

(2012). On the political economy of fiscal imbalances in India. In Srivastava

D. K.

& Sankar

(Eds), Development and public finance: Essays in honour of Raja J Chelliah (pp. 125–143). New Delhi: Sage.

42.

Tiwari

A. K.

(2012). Debt sustainability in India: Empirical evidence estimating time-varying parameters. Economics Bulletin, 32(2), 1133–1141.

43.

Trehan

, & Walsh

C. E.

(1991). Testing intertemporal budget constraints: Theory and applications to the US federal budget and current account deficits. Journal of Money, Credit and Banking, 23(2), 206–223.

44.

Tronzano

(2013). The sustainability of Indian fiscal policy: A reassessment of the empirical evidence. Emerging Markets Finance and Trade, 49(1), 63–76.

45.

Uctum

, Thurston

, & Uctum

(2006). Public debt, the unit root hypothesis and structural breaks: A multi-country analysis. Economica, 73(289), 129–156.