NEOCLASSICAL growth theory predicts that, in the absence of institutional harriers to the mobility of capital and labor across national or regional boundaries and for economics with similar technology and preferences, per capita incomes lend to converge over a period of time during economic growth. The study by Cashin and Sahay (I996) is one of the few attempts to test this prediction at the regional level for a particular country. The study analyzes the growth trends in 20 states in the Indian Union over the period 1961-91 and concludes that (1) “there has indeed been convergence in real per capita incomes across the states of India during the period 1961-91” (Cashin and Sahay. 1996. pp. 163-64); (2) while the cross-sectional dispersion in real per capita incomes increased over the years, the center-state grants ensured that the dispersion in real per capita disposable incomes remained broadly constant over lime; and (3) labor mobility was not an important factor contributing to convergence, particularly because of the existence of barriers.
A close examination of the paper, however, brings out a number of conceptual and methodological issues, both in the tests for convergence and in the analysis of the equalizing impact of intergovernmental transfers. First, the convergence is observed only when the manufacturing output share variable is included in the regression. We argue that, for controlling sectoral shocks, the inclusion of this variable is irrelevant. However, if the manufacturing output variable is taken as a policy variable, the results point toward conditional rather than absolute convergence. Second, the inclusion of four “special category” states makes the sample heterogeneous. Third, as net state domestic product includes the effect of federal grants, including them again to analyze the equalizing impact is conceptually erroneous. Finally, it is incorrect to exclude tax devolution from the analysis of equalization. Each of these points is discussed below.
I. Tests for Absolute Convergence
The absolute convergence hypothesis of the neoclassical growth model is that, for economies with the same steady state, poorer economies grow faster than richer ones, Cashin and Sahay test whether states in India exhibit absolute convergence by estimating a regression of the form
Absolute convergence occurs if (3 is significant and positive. The authors do not find evidence of absolute convergence with the regression as estimated in the form of equation (1) above. They then augment the basic regression with two additional explanatory variables: the share of agriculture and manufacturing in each state’s net domestic product (NDP) in the initial year (AGRi, t-T and MANi, t-T. respectively). The inclusion of these variables is justified on the grounds that, as sectoral shocks may affect individual states or groups of states differentially, failure to introduce these additional variables may lead to biased estimates of β, contingent on the realization of a particular shock. Using these variables, the authors now find evidence of a positive and significant β to confirm absolute convergence among the 20 Indian states.
It is important to note that absolute convergence is found only when MANi, t-T is included in the basic regression along with AGRi, t-T. However, if the purpose of including these variables is to control for sectoral shocks, the inclusion of one of these variables renders the other superfluous in the regression model. In the Indian context, there is a strong correlation between the relative price of agriculture and that of manufacturing (the correlation coefficient between the two in the period 1961-91 is minus 0.94), so that a positive shock to the relative price of agriculture would imply a negative shock to the relative price of manufacturing (and vice versa). Therefore, the inclusion of AGRi, t-T would be sufficient to control for the differential effect of sectoral shocks on states with differing sectoral compositions. A similar argument can be made against the inclusion of MANi, t-T in the regression if its inclusion is meant to capture the sectoral shifts in employment from low-productivity agriculture to relatively high-productivity manufacturing: here again, the inclusion of AGRi.t-T controls for the effect of changes in the industry mix on initial state per capita income. Therefore. MANi, t-T is an irrelevant variable: its inclusion should make no difference to the regression results.
Yet when MANi, t-T is included in the regression, β—the convergence coefficient—is found to be positive and significant. One important reason for this could be that differences in the degree of industrialization across states in India have largely been due to a policy bias in the planning process aimed at directing the regional spread of industrialization by. among other things, directing the How of credits, controlling the location of public sector enterprises, and issuing industrial licenses. Because differences in policies across economic regions in the neoclassical growth model would be reflected in differences in steady states, the variable MANi, t-T may be proxying in part for steady state differences across the 20 Indian states. In this case, the significant and positive β found in the augmented regression when the share of manufacturing in state NDP is included may indicate conditional convergence rather than absolute convergence, contrary to what has been argued by Cashin and Sahay.
II. Inclusion of Special Category States in the Sample
Cashin and Sahay’s pooling of the 20 Indian states in the regression analysis rests on the assumption that the states are homogeneous in technology and preferences, which is certainly not valid in the case of the four hill states of the north and northeast. These states are considered to be distinctly different in both official treatment and academic research: their critical distinguishing feature is the virtual absence of a production base. The only major determinant of net state domestic product in them is government expenditure, as may be seen from Table 1. Not surprisingly, the tax bases in these states are narrow, and much of the economic activity is triggered by central transfers. Therefore, these are considered “special category” stales and are treated preferentially in dispensing the plan funds.1 It would have been more meaningful at least lo have an exercise excluding these states and to compare the results.
|States||Ratio of government|
expenditures to net stale
|Ratio of central transfers to|
|Jammu and Kashmir||77.6||59.0|
III. Equalizing Impact of Federal Grants: A Conceptual Issue
Section V of Cashin and Sahay’s paper deals with the impact of intergovernmental grants on interstate disparities in real per capita incomes. This impact is measured by estimating the unweighted cross-sectional standard deviations in In Yit(σNDP) for the states in each of the years and comparing them with the dispersion of state per capita disposable incomes (σSDI). defined as NDP plus central grants.
An important conceptual issue in this analysis pertains to the definition of disposable income. NDP for the states in India is estimated using the income-originating approach, which includes the effect of federal grants. To add it again to get a measure of disposable income would amount to double counting. Take the simple identity
where Y, C. and I, respectively, denote income, consumption, and investment of the ith state. This can be shown as
because C, = C(p + g)i and Ii = I(p + g)i, where g and p represent government and nongovernment sectors. Assuming that taxes and expenditure spillovers cancel out, the state government budget will be
where T, F, D, and Gp refer respectively to tax collections, federal transfers, borrowing, and transfer payments to the private sector, which is a part of Cpi Substituting (4) in (3), we have
Thus, states’ NDP estimates include the impact of federal grants, and the concept of estimating stale disposable income by adding federal transfers to NDP is erroneous.2 Hence, the conclusion ought to be that despite the equalizing effect of federal transfers the dispersion in per capita NDP has widened over the years, and not that central grants ensured that the dispersion in real per capita disposable income was narrower.
IV. Noninclusion of Shared Taxes in the Analysis of Equalization
Another issue pertains to the exclusion of shared taxes in the analysis on the grounds that the transfers are given for the “purpose of reducing regional income disparities” (Cashin and Sahay, 1996, p. 128), and that the determination of the magnitude of the income-equalizing component lacks transparency. This raises two questions: What are the objectives of transfers? and What items should be included in the analysis of equalization?
In the literature, the economic rationale for federal transfers is given in terms of (1) closing the fiscal gap or offsetting vertical fiscal imbalances arising from tax and expenditure assignments to different levels of government; and (2) providing horizontal equity transfers to enable the residents in every state to enjoy a given level of public service at a given tax price by offsetting revenue (fiscal capacity) and cost disabilities of the states (Boadway and Flatters. 1982). Thus, reducing interstate income disparities is not one of the objectives of federal transfers, although the transfers given to offset disabilities in taxable capacity would be equalizing.
In the Indian context, it is conceptually incorrect to exclude tax devolution from the analysis of equalization because, like grants, shared taxes are used to offset fiscal disabilities. In fact, the Finance Commissions have varied the amount of tax devolution and grants to achieve the intended equalization. The methodology of the commissions consisted of filling the projected budgetary gaps of the states first through tax devolution (based mainly on population and backwardness), and then, for those states still left with gaps, through grants. Stung by the criticism that the commissions acted like “fiscal dentists filling in budgetary cavities.” the Seventh Finance Commission doubled the states’ share of excise duty from 20 percent to 40 percent to leave most of the states in surplus, obviating the need for grants. However, the increased tax devolution had to be targeted to poorer states by increasing the weight assigned to per capita NDP.3 The empirical analysis shows that the equalization impact of tax devolution has been greater than all other forms of transfers, particularly during the period covered by the awards of the Eighth (1984-89) and Ninth (1989-94) Finance Commissions (Rao. 1996). It must also be noted that the commissions explicitly stated the distribution formula, the weights assigned to different factors, and the relative shares of the states.4 Thus, it is conceptually unsound to exclude shared taxes from the analysis of equalization: it is also factually incorrect to assert that there is no transparency in the equalizing component of shared taxes and to exclude them from the analysis on that basis.
BarroRobertJ. andXavier Sala-i-Martin1991“Convergence Across States and Regions,”Brookings Papers on Economic Activity: 1Brookings Institution pp. 107–58.
BoadwayRobin W. and FrankFlatters1982Equalization in a Federal State: An Economic Analysis (Ottawa, Canada: Canadian Government Publishing House).
CashinPaul andRatnaSahay1996“Internal Migration. Center-State Grants, and Economic Growth in the States of India,”Staff Papers International Monetary FundVol. 43 (March) pp. 123–71.
Government of India 1995Indian Public Finances 1995 (New Delhi, India:Government of India Press).
RaoM.Govinda1996“Intergovernmental Fiscal Relations in a Planned Economy—The Case of India,”paper presented at the conference on “Fiscal Decentralization in Developing Countries”Montreal, CanadaSeptember19–20.
RaoM.Govinda andRaja J. Chelliah1996Survey of Research in Indian Fiscal Federalism (New Delhi, India: Indian Council of Social Science Research).
The “Gadgil” formula used to distribute plan transfers does not apply to special category stales. While for the general category states 30 percent of plan assistance is given as grants and 70 percent as loans, the grant-loan ratio is 90:10 for the special category states.
In their study of the U.S. economy, Barro and Sala-i-Martin (1991} exclude federal transfers from the personal incomes of states to measure the equalizing impact of federal transfers.
Tax devolution was made progressive also by choosing the inverse