Although present day Indians trace their ancestors through waves of migrations across millenia and there’s been plenty of inter-marriage as groups moved across regions and even caste boundaries changed, the fact remains that there are thousands (as many as 30 to 50 thousand, according to Cavalli-Sforza) of endogamous groups (jatis or microraces) in India. That is, these groups, in the time frame of (say) 1000 years, have almost exclusively married within their clan. This fact, together with the effects of natural selection (e.g., famine, assured food supply to a certain person by virtue of his/her talents) may have made these groups diverge.

We have pretty good evidence (a recent paper by Cochran and Harpending) that the highly intelligent Ashkenazi Jews may have, in fact, increased their intelligence over the past 1000 years due to such considerations (because they were banned from most work in european cities and discriminated against, they got into money lending a profession that made a lot of use of higher cognitive abilities.. and a Jew who was clever by random genetic variation was more likely to survive and produce children in these years when life was pretty nasty and short).

The surprising conclusion is that traits such as general cognitive ability aka intelligence can have accelerated selection (as opposed to the idea that these things take tens of thousands of years as in the case of skin color.. the white skin of europeans is a mutation that arose 7000 years ago)

Add to this that our current technological, knowledge based economy puts a premium on general intelligence (verbal and mathematical ability), then these facts, taken together lead us to certain conclusions that people may not want to admit. Namely, it is highly unlikely (actually mathematically impossible) that there wont be significant variation on traits like mean intelligence across the thousands of endogamous groups. Thus the key form of discrimination (occuring over a thousand or more years) is probably marriage within clan.

Imagine the distribution of mean IQ (not individual IQ) across these thousands of groups (no one has bothered, as far as I know , to make some estimate.. but its not hard to do based on data that various govt agencies already have). The difference in the mean difference between the top 10% and the bottom 10% (of groups) is very very likely to be at least a couple of standard deviations or more. But the details of this distribution will have clear implications for policy making (such as identifying environmental causes for depressed intelligence like nutritional deficiencies that may be affecting certain groups but not others). Its a mystery as to why our govt agencies who collate various statistics havent bothered to create something like this (one doesnt actually need to measure IQ, one could use a combination of various proxy measures). It should not be more than a few months worth of data collection and analysis (if various agencies are forthcoming). One likely conclusion would be that reservations at elite institutions is pointless and a waste of efforts.. its much more efficient to have reservations at non-elite institutions and earlier (primary education) via school vouchers etc as some NGOs are advocating. Another would be the isolation of environmental factors and relevant interventions that could actually make a difference to certain groups.