The paper addresses the question of government expenditure allocation among sectors by extending the model developed by Archana Dholakia (1993) and illustrates its application by using the data on major Indian states from 1971 to 1991. It is argued that on margin, the changes in the expenditure allocation are determined not by the magnitude of the marginal productivities of the government effort (as several people seem to be thinking), but by the behaviour of the marginal returns in relative terms. A lot of emphasis is put on the appropriate measurements of al the variables flowing from the theoretical framework so as to ensure proper interpretation of the coefficients of the model. The dependent variable is the disparity reduction rate (DRR) in the index of basic welfare that is measured by 9 indicators from education, health and nutrition & other sectors. The independent variables are the average per capita annual development expenditures at constant prices by state governments on revenue and capital accounts in different sectors reflecting the change in the government efforts in different directions; and the level of real per capita state income. In the illustrative exercise, the economic (physical capital) sector and social (human capital) sectors are considered. Similarly, two t!me periods -1971-81 and 1981- 91 are considered to examine the stability of the coefficients. Statistical tests of equality of the coefficients of revenue and capital expenditures are also carried out. The coefficients are stable and equality restrictions are valid. The government efforts on the social (human capital) sector show increasing returns whereas on the economic (physical capital) sector show diminishing returns. Based on our findings, it is advisable for the government to spend more on social sector and less on economic sectors than what it is doing in the recent past.