The actual disparities between SAT scores of admitted black students and admitted white and Asian students are actually likely significantly larger than implied.
The presentation of SAT data was made by Arcidiacono on p. 26 of his initial report linked in my post #1937 above. It is only pictorial, so we can only estimate, but sometimes a picture is worth a thousand words. The SAT scores presented for black admits over the 18 years or so shown are consistently lower than those of applicants of any other racial groups, except Hispanic applicants. Black admit scores are far lower than those for any other admitted group.
Moreover, note footnote 34 in the document. These are not even superscored SAT scores; they are a composite of SAT math plus the maximum of SAT reading and SAT writing. (Apparently, most of the data are from the 2400 scale SAT; presumably, the data pre-2009 or so reflect pre-2005 scores and are therefore just superscored just V+M 1 on the earlier 1600 scale.) I have no idea how ACT scores fit in.
Using a maximum of two available scores will have the tendency to compress the distribution and range of differences. As blacks are the lowest scoring applicant group, they should benefit most from compression. (Superscoring will magnify these tendencies as well.)
Also, with respect to white admits, note that the scores will disproportionately include those of the weaker legacies (~20% of the admits generally but presumed skewed white) and (especially) donor class and special interest admits (~14% of admits generally, again presumably skewed white).
The statement of undisputed facts that I referenced in post #1988 above (see Paragraph 682 on p.146 is also useful for SAT information by race in that part of the applicant pool deemed to be “standard strong” by the adcoms at Harvard. Here are the SAT Verbal and Math scores by race for those candidates, together with the share of applicants from such race who are deemed to be standard strong:
White: 758V 750M (12.0% of white applicants)
Asian: 759V 770M (15.1% of Asian applicants)
Hispanic: 685V 734M (3.6% of Hispanic applicants)
Black: 615V 625M (1.0% of black applicants)
I was struck not by the gaps, which are to be expected, but by the extraordinarily poor performance of blacks on the SAT Math (compare with Hispanic), which might reflect gender effects in the sample (I suspect the standard strong group for blacks is very skewed female, and females generally score lower on math than males). Note, though, Harvard only produced a random sampling of 10% adcoms sheets categorizing standard strong candidates, and so the extraordinarily low representation of blacks (1%) and glaringly low scores may reflect some sampling error (but again compare with Hispanic, also small N at 3.6% representation).
I have no doubt that the black admits to Harvard have the highest scores of the group, but I do suspect that the pictorial representation in the Arcidiacono Report needs to be informed by all the other data that have been released.
Also, again a point I have made on here before, for prospective black students reading this, the above data are consistent with all the other data I have seen: while many on CC will say test scores do not matter “past a certain threshold,” at least for black applicants maniacal test prep would appear likely to be the single most effective strategy to maximize chances of admission at the elites.
