A paper titled Structural Adjustment, Education and Poor Households in India: Analysis of a Sample Survey by Basanta K Pradhan & A. Subramaniam (NCAER 1999), presents the following data.

Here're the conclusions of the paper.
The paper addresses an issue of great concern - despite massive Government efforts, the state of education in India is improving slowly. The literacy rates among poor households are low and across occupational categories, the lowest is among Agricultural and Non-agricultural Wage and Self-employed Farm workers in rural areas. The Non-agricultural Wage earners in the urban areas also have very low literacy rate. Both dropout and non-enrollment rates are also high in either regions. This is because of the lack of expected future earnings. Many studies in the literature attribute such poor performance in the educational sector to poverty, and child labour supplementing family income is considered to be the main cause for such maladies.
However, the MIMAP survey (NCAER, 1998) - an all-India household survey on demand and supply issues in schooling - shows that the major reasons for such poor performance of the educational sector are both unconstrained and constrained demand. Unconstrained demand refers to factors having no constraint but simply lack of willingness and constrained demand factors are constrained by poverty conditions, participation in household economic activity, etc.. In general, unconstrained demand factors stand out as the major reason in rural areas while in urban a larger percentage of poor attribute to constrained demand factors.
Therefore educational reforms in India should basically focus on the demand lead supply transformation towards skill enhancement. This seems to be missing in the government-owned and operated educational sector. In this context, the role of the state should be reduced to provide the basic education. The higher education should be left to the market. However, the government has proposed to move in this direction.
The restructuring of the educational sector will have both short run and long run effects. In the short run, this will not only strip the undue burden which the state carries in terms of high budget deficits but also can provide adequate resources for the fulfillment of the basic objective line 'literacy for all' or mass education, etc. Apart from this, open economy framework in the educational sector is likely to allocate resources efficiently and bridge the gap between demand and supply of skills.
In the long run, the opportunities for employment and income will increase the demand for education not only for secondary and higher education but also for primary education in lieu of future prospects. However, the privatisation of education most often exclude the poor from drawing benefits of education at secondary and higher levels. Here too, the role of the state is important apart from providing free primary education. The state should provide adequate scholarships and assistanceships for the vulnerable, which infact, requires much less government resources than the government owned institutions of higher learning.
Another report on education sheds some more light on household expenditure on education in India.
“Schooling is too expensive” was the first among reasons cited by the study conducted by the Public Report on Basic Education (PROBE) to explain why a child has never been sent to school in rural areas. Survey data confirmed that the cost of sending a child to school is far from negligible. The most costly items are clothing, followed by textbooks and stationery. The average cost of sending a child to school works out to be Rs 318 per year in rural areas. This may not look high but it is a substantial burden on a poor family. The figure reflects actual expenditure, not the expenditure that would be required to cover a pupil’s basic need (in terms of textbook, stationery and so on). The poor are often constrained to produce cash at
short notice to meet immediate needs. This creates a basic liquidity problem, even when the annual expenditure is otherwise affordable.
A study conducted by NCAER in 1994 found that the expenditure per household on educating children aged 6–14 years in rural India was estimated to be Rs 680 per household or Rs 378 per pupil per year. Expenditure on schooling for girls was about 68 per cent of the amount spent on boys. A large proportion of this expenditure was incurred on books, stationery and school uniforms, followed by private coaching and fees. This expenditure increased or decreased with the annual fluctuations in the cost of books, stationery and uniforms.
The study further found that only about 75 girls for every 100 boys go to private schools. The expenditure on girls is much lower, especially in private schooling, whereas parents do not think twice about spending large sums of money on educating boys in private schools. There is an increasing dependence on private schooling in recent years that has affected female education more than male education. The household expenditure was also found to be higher among large landowners, relatively rich and salaried persons.
The patterns of expenditure in rural and urban areas are different. In rural areas, the expenses are high on uniforms, books and stationery. Whereas in urban areas, expenditures are highest on tuition fees and private coaching fees.
The proportion of children going to private schools is highest among the high-income households. It is interesting to note that over 70 per cent of ST and SC households study in Government schools. Only about 43 per cent of Christian and 49 per cent of Muslim children attend Government schools, while most of the remaining Christian children go to Government-aided schools. The expenditure on education is relatively low among STs, SCs and Muslims, amounting to about three per cent of the household income for the respective social group.
Report no. 439 - Attending Educational Institutions in India: its Level, Nature and Cost - part of the study on Consumer Expenditure, Education, Health and Aged in India, National Sample Survey 52nd round (July 1995 - June 1996) supposedly has more data on household expenditure on education.