While the government has made it clear it has no intention of re-visiting the debate over creamy layers in OBC (other backward castes) reservations, it is making a huge mistake since it is incomes and not caste that are the major determinants of levels of education.

Not just for OBCs or SC/STs, but even for upper caste Hindus. And this is something that is revealed by most surveys, whether it is the National Sample Survey of 1999-2000 or the Reproductive and Child Health Survey, which was conducted over two years from 2002 to 2004 over 600,000 households across the country - the RCHS, in fact, has a sample size that is several times larger than the NSS.

If you look at the broad data, as the government and other pro-reservationists do, the RCHS data show SC children had an average schooling of around 3.2 years, versus 3.9 for the OBCs, and 5.6 for the upper caste Hindus.

The rest follows from this - low years of schooling mean lower possibilities of going to college and hence, a poor probability of getting a good job. So, the case for reservations is self-evident.

Though the government is keen on reservations at the graduate level, the real issue is the lack of adequate schooling since, without enough SC/ST/OBC passing out of high school, they cannot be admitted to college.

Analyse the mass of data, as economic research firm Indicus Analytics has done, and you get a totally different picture, a picture which shows that factors such as income levels, urban versus rural settings and location in different states are perhaps even more important determinants than caste.

Take the poorest income quintile (fifth) of rural population, and you find that there isn't such a significant difference in the education levels across castes - it is 1.6 years for SCs, 1.7 for OBCs and 2.2 for upper caste Hindus.

Take the same castes, the same lowest income quintile, but change the setting to urban areas. And you find that the same SC child now has 2.6 years of education, or two-thirds more; the increase for OBCs and upper castes is equally high.

While the difference between the SC and the upper caste Hindu in the lowest income quintile in rural areas is 37 per cent, the difference between the rural SC and the urban SC in the same quintile is 63 per cent!

Slice the data according to income, and you get a similar result. In the same rural setting, an average SC child in the top-most quintile, or the richest 20 per cent of the population, has 5.1 years of schooling, a figure that's 3.2 times that for the SC in the poorest quintile.

For OBCs, the difference between the top-most quintile and the bottom-most is also 3.2 while for upper caste Hindus the difference is around 2.7 times. In the case of urban populations, similar differences are to be seen across income groups.

The NSS data for 1999-2000, not surprisingly, show similar results - the NSS number-crunching, to present the data in the desired format, has been done by Surjit Bhalla's Oxus Research.

In the lowest income quintile, an upper caste Hindu has 87 per cent more years of education than an SC and 47 per cent more than an OBC. But, an SC in the upper-most income quintile in rural areas has 2.5 times the number of years of education as an SC in the lower-most quintile; for OBCs the difference is 2.2 times.

Similar differences show up between urban and rural areas when you look at SCs or OBCs or even upper caste Hindus in the same income quintile.

Indeed, the NSS shows that if you take the poorest 1 per cent of SCs in rural areas, the average years of schooling is just 1.3 (it's 1.7 for OBCs and 3.5 for Hindus). Now move up to the top 1 per cent, and the average SC has 5.9 years of education (it's 6.3 for the OBC and 7.1 for the Hindu)! That is, the difference between castes shrinks dramatically as income levels rise.

If caste was to be the sole/major determining factor, then surely the difference between the education levels would be a lot higher between castes in the same income group than they are for the same caste across different income groups. The fact that there are such huge differences between rural and urban areas, it would appear, is also due to the income difference between rural and urban areas.

Examine the data across states, and they throw up equally interesting results. The average SC in Bihar has 1.9 years of schooling as compared to 2.9 for OBCs and 4.9 for Hindus (RCHS data). But, in Maharashtra, the figures are 4.4, 5.1 and 5.7 - that is, the relative inter-caste differences don't hold good across states.