Around 36 per cent of the country's population is defined as belonging to the Other Backward Classes according to the National Sample Survey's 1999-2000 round, and not 52 per cent as defined by the Mandal Commission, a number that most politicians still use while asking for reservation.

If you exclude Muslim OBCs, the proportion falls to 32 per cent according to the NSS, 1999-2000. Indeed, Yogendra Yadav, professor at the Delhi-based Centre for the Study of Developing Studies, who is in favour of reservation for OBCs, agrees that there is no empirical basis to the Mandal figure: "It is a mythical construct based on reducing the number of SC/ST, Muslims and others and then arriving at a number."

One of the reasons for the much higher Mandal number is that it defined OBCs in socio-economic terms, and so included, for instance, "castes/classes considered as socially backward by others". If, to cite another of the 11 criteria used, the percentage of married women below 17 years is a fourth above the state's average in rural areas (and 10 per cent in urban areas), the community is considered to be OBC.

Similarly, if castes/classes where the proportion of working women is 25 per cent higher than the state's average, the castes/classes are considered OBC - today's double-income families, if living in areas where women do not work, would then be considered OBCs using the Mandal definition.

In the NSS case, respondents were asked to indicate their caste, and this was then tallied with the list of castes that each state defines as OBC.

The NSS data is also corroborated by the National Family Health Statistics, a survey conducted in 1998 by the DHS, which has conducted 200 such surveys in 75 countries.

The NFHS data show that the proportion of non-Muslim OBCs is 29.8 per cent, a figure quite close to the NSS' 32.1 per cent. For SC/ST, while the NSS shows this is 28.3 per cent of the population, the NFHS estimates this at 27.9 per cent. The 2001 Census estimated the SC/ST population at 24.4 per cent, though the Census did not canvass any information on OBCs.

The share of the Muslims (including OBC Muslims) in all three data sets is quite similar, ranging from 11 to 13 per cent.

So far, the Supreme Court ceiling of 50 per cent on all reservations has been justified by arguing this covers only the non-creamy layer OBCs, since 22.5 per cent of all reservations are for SC/ST, leaving 27.5 per cent for the OBCs - that is, only around half the OBCs would be entitled to reservations.

If, you use the NSS/NFHS figures, a 50 per cent reservation ceiling will cover three fourths of all OBCs, and if Muslim OBCs are to be kept out of reservations, then 86 per cent of the remaining will get covered by reservations.

Part II: Schooling is an effective job quota