Relationships between Household Consumption and Inequality in the Indian States

Abstract

Current evidence on the relationships between growth and inequality is predominantly based on cross-country data sets or panel data sets covering a small number of time periods. But these relationships, being fundamentally dynamic in nature, need to be considered over a much longer time horizon. Available state level results from the National Sample Surveys in India provide such an opportunity. This article uses this unique data set to examine the interrelationships between average consumption and inequality within states, and test for causality. Distributional patterns of growth vary, but there is strong evidence in many instances of a strong negative effect of initial inequality on subsequent growth.

Notes

This would also mean that cross-sectional relationships based on this data set would not suffer from the serious problem of the potential non-comparability of data which affects cross-country estimates. However, they would still suffer from several of the other problems of cross-sectional studies of this issue (see above) and so this article focuses predominantly on the time series relationships.

It may be that because household consumption expenditure accounts for a high proportion of GDP, similar relationships may apply between the level or change of household income or consumption and its distribution, as between GDP growth and the distribution of household income or consumption. But this is an empirical issue.

In a slightly different vein, Das and Barua [ 1996 ] examine the effect of changing trade regime on inter-regional inequality in India, with special reference to Kuznets type relationship. As indicated, our study is more general in that it also considers the effect of initial inequality on subsequent growth.

Note that the Gini coefficients were estimated using parameterised Lorenz curves, using two alternative specifications: the general quadratic and the beta specificiation. The preferred specification varied by state and by urban/rural areas within the state [ Datt and Ravallion, 1992 ]. No other inequality estimates were available for this paper.

Given that we use annual data to study the relationship between consumption (or its changes) and inequality (or its changes), this is perhaps not a surprising result.

This corresponds to government efforts to reduce credit market segmentation where a significant proportion of total credit is still supplied by the informal agencies including the moneylender, who charge an exorbitantly high rate of interest. Although interest rates are generally lower for the credit offered by the formal sector (for example commercial banks), formal loans generally require some acceptable form of collateral, which in turn restrict the access of assetless poor households to formal credit.

We would like to thank Professor Tim Besley for allowing us to use this additional information.

While it is unfortunate that the periods are of differing length, the variations are quite small and this is not a particularly serious issue because the data set focuses on initial values and averages over the period only.

Given the Indian government's emphasis on poverty alleviation programmes, government regularly undertakes redistributive expenditure to provide social, economic and community services targeted in favour of the poor.

This captures an important aspect of segmented credit markets especially prevalent in rural India.

When the dependent variable is the growth of average consumption, this captures the Barro convergence hypothesis: after controlling for all other factors, states with lower initial consumption will experience a higher rate of growth and vice versa.

Here rural output is measured by the share of agricultural output while urban output by that of manufacturing output.