1) This graph uses data from the National Longitudinal Survey of Freshmen (NLSF). This data seems to be easily available to academics, so readers are encouraged to replicate my analysis themselves. Please let us know if you do so. Rory: Does this graph look correct to you?

2) NLSF does not allow users to indicate which data come from which school. So, for example, I am not allowed to report that the average SAT score for Asian-Americans at Williams is X. I can only report the results in groups like “elite liberal arts colleges,” as I have done here.

3) To select my sample, I use “la0102 <= 10″. This restricts the observations to students who attended a liberal arts college ranked in the top 10 by US News in 2001-2002.

4) Within this universe, I look at students marked as “A” (Asian) or “B” (Black/African-American) by NLSF. There are 66 such students in the sample from elite liberal arts colleges: 35 Asian and 31 Black. This is, obviously, not a very large sample and only includes data from two schools. There is reason to believe, based on what I have read elsewhere, that these results are not entirely an artifact of small sample size, but this data alone doesn’t show that.

5) 7 A students and 10 B students are missing SAT scores. For 3 As and 5 Bs, I am able to impute their SAT scores using their ACT results. But removing those observations does not matter to the overall distributions.

6) With the imputed scores, we have a final sample of 31 Asian-American and 26 African-American students. Their SAT scores are used to create the plot.

7) The middle 90% of the distributions do not overlap. The kernel smoothing method I use obscures that fact a bit. In other words, if you delete the three highest B scores (1400, 1420 and 1460) and the three lowest A scores (1200, 1340 and 1350), there is no overlap between Black and Asian SAT scores at places like Williams.

8) Poking around the NLSF for other schools, it seems like the A/B gap of around 250 points is larger in elite liberal arts colleges than anywhere else.

Among elite universities (say, top 10 in US News), the gap is about 100 points smaller because African-American (and Hispanic) students have significantly higher scores. This is consistent with what I have heard elsewhere. Places like Williams lose many/most/all of the African-American (and Hispanic) students they most want to larger universities. If you want to attend college at a place where SAT scores for Asians and African-Americans are only (?) 150 points different, don’t go to an elite liberal arts college like Williams.

Why is it that – for the most part – the highest scoring African-American high school students have such a strong aversion for the elite liberal arts schools (or such a strong preference for the research universities)? Is is principally a name recognition factor? But why would that effect be different for African American students as compared to white students?

Do we have any readers who work with NLSF data? Tell us what you think!

Special thanks to Rory for pointing out the existence of the NLSF and encouraging/shaming me into taking a careful look at it.

Facebooktwitter
Print  •  Email