Jason Richwine writes in Forbes:
But education and culture can take people only so far. To be a great speller—or, more importantly, a great doctor or IT manager—you have to be smart. Just how smart are Indian Americans? We don't know with much certainty. Most data sets with information on ethnic groups do not include IQ scores, and the few that do rarely include enough cases to provide interpretable results for such a small portion of the population.
The only direct evidence we have comes from the 2003 New Immigrant Survey, in which a basic cognitive test called "digit span" was administered to a sample of newly arrived immigrant children. It is an excellent test for comparing people with disparate language and educational backgrounds, since the test taker need only repeat lengthening sequences of digits read by the examiner. Repeating the digits forward is simply a test of short-term memory, but repeating them backward is much more mentally taxing, hence a rough measure of intelligence.
When statistical adjustments are used to convert the backward digit span results to full-scale IQ scores, Indian Americans place at about 112 on a bell-shaped IQ distribution, with white Americans at 100. 112 is the 79th percentile of the white distribution. For more context, consider that Ashkenazi Jews are a famously intelligent ethnic group, and their mean IQ is somewhere around 110.
Given the small sample size, the rough IQ measure and the lack of corroborating data sets, this finding of lofty Indian-American intelligence must be taken cautiously. Nevertheless, it is entirely consistent with their observed achievement.
The New Immigrant Survey is being run out of Princeton. It's one of those massive projects like the 1979 National Longitudinal Study of Youth. It only applies to legal immigrants. The first round of data was collected in 2003, and papers have been coming out for a couple of years. Parts of the Woodcock Johnson IQ test were given to immigrants, as well as the digit span test. The data is available to registered users.
If you know more about this research, such as what other legal immigrant groups scored, please let me know.
I'm guessing that Richwine crunched the numbers himself, because I can't find anything published on them, and he is writing his Ph.D. dissertation at Harvard on IQ and immigration nexus. (Here's Marcus Epstein's VDARE blog post on Richwine's forthright statements at an AEI conference.) The SPLC is already on Richwine's case for uttering hatefacts.