Dynamics of public debt sustainability in major Indian states

Abstract

This study empirically tests whether the public debt is sustainable or not at 22 major Indian states during 2006–07 to 2015–16. It employs the Bohn model for panel data, five alternative specifications and p-spline technique to analyze the issue at aggregate and disaggregate levels. While the results indicate that the debt is sustainable at the aggregate level, it is sustainable only in about 11 states. The results suggest that the fiscal reaction function is linear and the central grant-in aid is an important and a significant undermining factor of sustainability. If the grant-in-aid is excluded from the primary balance, there remain significant positive responses at the aggregate level. However, at the disaggregate level it is significant in only 11 states. Further, the most sustainable states fail to meet the no-Ponzi condition and so the policy intervention is required to improve the debt situation of the states where debt is unsustainable.

Keywords:

• Public debt
• primary balance
• sustainability
• Indian states
• Bohn model for panel data

Notes

Disclosure statement

No potential conflict of interest was reported by the authors.

Notes

1 The no-Ponzi condition is such that the public debt growth rate has to be lower than the real interest rate. It ensures solvency since the funding of interest payments are not made from the new debt issuances. In other words, the existing debt can be serviced with own revenue effort and hence no additional borrowing is required (Azizi et al. 2012).

2 There are three broad measures of deficit at state level in India viz. the revenue deficit (=excess of revenue expenditures over revenue receipts), the fiscal deficit (=total expenditure – revenue receipts – non-debt capital receipts) and the primary deficit (=fiscal deficit – interest payments).

3 Article 293 of the Indian Constitution stipulates that the state governments do not have unrestricted power to borrow as long as they are indebted to the Centre and also prohibited direct borrowing from abroad with some exceptions.

4 This condition has been extended by adding various other indicators like growth, liquidity, creditworthiness, fiscal burden, fiscal space etc. and renamed as ‘Indicator approach’ (Blanchard et al. 1990; Rajaraman, Bhide, and Pattnaik 2005; Mishra and Khundrakpam 2009; Kaur et al. 2014)

5 The IBC is $d_t^{\mathrm{*}}={\sum }_{j=1}^{\mathrm{\infty }}\frac{1}{{(1+r)}^j}{E}_{t\mathrm{ }}\left[{\mathrm{s}}_{t+j}\right],\mathrm{ }$ where $d_t^{\mathrm{*}}$ = (${1+r}_t$). $d_{t-1}$ is the stock of the debt-output ratio at the beginning of period t, $E_{t\mathrm{ }}\left[.\right]$ denotes the expectation operator conditional on the information available at time t, and ${\mathrm{s}}_{t\mathrm{ }}$ is the primary surplus-GDP ratio. The IBC of the government requires that the present value of public debt asymptotically converges to zero.

6 One can also assess sustainability by seeing the co-integrating relationship between the public revenues and the public expenditure (Hakkio and Rush 1991; Afonso 2005) i.e.${{\mathrm{ }R}_t=\alpha +\beta \mathrm{ }E}_t+\mathrm{ }{\upsilon }_t;$ where R and E are revenues and expenditure respectively and I(1), while $\upsilon$ is I(0).

7 Born (2007) warned against interpreting failure of the stationarity and the co-integration as an evidence of debt sustainability while others criticized as both are not informative about model-based sustainability conditions, statistical failures during structural breaks, ergodicity problem etc.

8 The tax-smoothing hypothesis implies that public deficits are used to keep the tax rates constant which minimizes the excess burden of taxation. Therefore, the regular public expenditure can be financed by government revenues, while the deficits can be used to finance the unexpected expenditures. One can yield primary balance equation from tax- smoothing model by subtracting the primary expenditure to GDP from the Tax revenue to GDP (Barro 1979)

9 Fincke and Greiner (2011b) provide justifications for using the time-varying coefficients as: (i) the true data generating process is unknown and most likely nonlinear and any nonlinear model can be approximated by a linear model with time-varying coefficients which is more robust than the OLS and gives an estimation result that comes close to the true data generating mechanism

10 For estimation purpose, p-spline considers the parametric form: f(d_t) = d_t β_d + Z(d_t)γ, where Z is a high dimensional basis in d (for instance a cubic spline basis) and γ is a corresponding coefficient. The high dimensionality restricts the use of OLS. So it imposes a penalty term on γ, shrinking its value to 0. It obtains estimates by minimizing penalized OLS criterions: ∑{s_t - d_t β_d - Z (d_t) γ}² + λ γ^TPγ; where λ is smoothing the penalty parameter and γ^TPγ is a penalty. P matrix is chosen in accordance with the basis (see Ruppert, Wand, and Carrol 2003 for details). λ basically steers the amount of smoothness of the function (if it is zero, then the model becomes unpenalized OLS). The fitted functions (f*) can be written as f₁* (d) = H(λ) where H is smoothing matrix. To obtain a reliable fit, λ should be chosen data-driven. One possibility is the use of Generalized Cross Validation (GCV) criterion as GGV= $\sum {\left[\frac{\mathit{st}-f(\mathit{dt})}{1-\mathit{tr}(H)/n}\right]}^2$; A suitable choice of λ is achieved by minimizing GCV. This procedure is the same if the time-varying coefficients are estimated (Greiner and Kauermann 2008)

11 There are two-way the response variable is differenced with Davig and Leeper (2011) model: (i) instead of tax to GDP we have used state’s own revenue to GSDP by adding non-tax revenue in the state context; (ii) instead of revenue receipts as such we subtracted tax share and grants from it, in order to account for the own revenue effort of the states.

12 The GSDP data is converted into single base (2011–12) using growth rate splicing method

Additional information

Notes on contributors

P. S. Renjith

P. S. Renjith is Assistant Professor at Christ University Bangalore (India). He has completed his Ph.D in public finance from Madras School of Economics. His main research areas are public finance, development economics and applied econometrics.

K. R. Shanmugam

K. R. Shanmugam is Director and Professor at Madras School of Economics, Chennai (India). He received his Ph.D in Econometrics from University of Madras. His specialization includes public finance, macro-modeling, applied econometrics and environment economics.