Debt tolerance: a model and its application to Indian states

Abstract

A number of studies on the fiscal situation in India have noticed the deterioration in finances of the state’s governments in India. The deterioration has translated itself into burgeoning debts, the sustainability of debts in states, therefore, is in question. Sustainability of public debt has been a subject matter of many studies among the developed nations. However, only a few studies have examined the question in emerging markets, especially India. Moreover, the models developed to assess debt sustainability are more appropriate for a national economy, not for subnational governments, which may have different characteristics & motivations for fiscal behaviour. In the Indian context, the differences arise because of the following reasons. First, The financial dependence of States on the Centre is expected to create a moral hazard problem when States can afford to be profligate, expecting a bailout in times of crises. Second, States are not exposed to risks on external debt. Third, they also do not face the tradeoff between money and bond financing. Lastly, the base variable in most sustainability studies, namely the debt-to-income ratio uses Gross Domestic Product (GDP), which may be appropriate only for a national government. The use of Gross State Domestic Product (GSDP) at State level has many limitations, as noted in some studies. The present study argues that interest payment to revenue receipts (IP/RR) ratio is an appropriate measure of fiscal stress. The Eleventh Finance Commission (EFC) had specified a limit of 18 percent. In the present study, we examine and derive a toleranceievel for the ratio with more rigour. The tolerance level is derived from the condition that the present value of the partial revenue impact of the present expenditure should be more than or, at the margin, equal to the present value of its partial expenditure impact. In order to segregate these partial impacts, the relationship between State's own revenue (SOR), transfers from the Centre (TRC) and the primary (non-interest) expenditure (PE) is examined. We model the relationship in a Vector Autoregressive (VAR) form, separating the Category (SC) States using a dummy. The equations are estimated for lags varying from one to eight, using Generalized Method of Moments (GMM). Model specifications are tested by Saga- n test of over - identifying restrictions. Using the Difference Sargan (DS) statistic, equations with lags of 6 3 and 8 years are selected for SOR, PE and TRC respectively. The coefficients of NSC and SC States are not found to be statistically different. However, errors in estimation are higher for SC States. Pair-wise Granger causality tests establish bi-directional Granger-causality in all cases except that SOR is not found to Granger-cause PE. Another important observation is that SOR and PE together, and not separately, Granger-cause TRC. The model for tolerable interest to revenue receipts ratio is tested for robustness under forecast errors. Its stability is also tested by applying it in different base years and for different periods. These analyses show that the model is fairly robust and stable. As expected, the tolerance levels so established are found to vary among States. When we take 30-year yield on Government bond (7.1 percent in January 2005) as the discount rate, and compare the tolerance levels with the actual values of the ratio, we find that Orissa, Punjab, Rajasthan, Uttar Pradesh and West Bengal among the NSC States & Himachal Pradesh among the SC States lie above the tolerance limit. The second part of the thesis builds fiscal scenarios for all the States during 2005-06 to 2009- 10, the Twelfth Finance Commission {TFC) period. Since the differentia! of growth in income over interest rate has a benign impact on the underlying fiscal parameters, a model is developed for forecasting the effective interest in the future. In this model, interest rate is endogenously determined by various factors including the fiscal deficit, and its sensitivity to the underlying factors is also computed. We find that effective interest rates are most sensitive to the relative proportion of various debt- components in the total debt portfolio. It is observed that maximum reduction in interest rates can be obtained if the proportion of Loans & Advances from the Centre & Internal debt {net of market loans & WMA) is reduced and this reduction is compensated for by an increase in the proportion of market loans. The recommendations of the TFC, which are on these lines, are vindicated by the study. Falling interest rates on all components of debt are likely to keep the effective interest rates down for at least a decade and the process of reduction in interest rates will be accelerated if the recommendations made by the TFC are implemented. If States exercise reasonable fiscal prudence, the situation will improve impressively for them. The incentive scheme by the TFC for reduction of revenue deficit gives some confidence that the expectation will be met.