Last night I posted an introduction to the Center for Equal Opportunity’s report on the use of race in assessing applicants to the University of Wisconsin at Madison, and attempted to untangle their preposterously misleading claim that black students have a 576-to-1 advantage in admissions. This morning I’ll be tackling CEO’s claims as to the rates at which student applicants of various racial groups are admitted to UW.
The CEO study’s author, Althea Nagai, claims on the report’s first page that in 2007 and 2008 “UW admitted more than 7 out of every 10 black applicants, and more than 8 out of 10 Hispanics, versus roughly 6 in 10 Asians and whites.” The university’s own public records for those years, however, show admissions rates of just 42% and 55% for black and Latino students, respectively, compared to a 55% rate of acceptance for white applicants and a 56% rate for Asians. According to UW, in other words, white, Asian, and Latino students were accepted at almost exactly the same rates in the period covered by the CEO report, while black students’ admission rates were considerably lower.
The CEO report does not attempt to explain this dramatic discrepancy. It does, however, provide a hint in a footnote, when it gives its raw numbers for total UW applicants and admissions in 2007 and 2008. By CEO’s reckoning, there were 38,476 applicants and 23,769 admittees to UW in those two years, while according to the university, 50,348 students applied for admission in 2007-08, of whom 27,415 were admitted.
CEO’s data set, in other words, is missing some 11,872 students, of whom only 3,646 were admitted. Nearly twelve thousand students — almost a quarter of the total applicant pool — were left out of CEO’s calculations. Why?
To start with, CEO’s sample set excludes “those cases for which race or ethnicity is listed as ‘Other,’ missing, or unknown,” as well as Southeast Asians and “American Indians and Native Hawaiians.” (A more appropriate approach would have been to combine these students into their own category, in order to preserve the integrity of the study’s data set, but we’ll let that pass.)
Aggregating UW’s figures for white, black, Latino, and “Other Asian” applicants for 2007 gives a total of 21,443, of whom 12,261 were admitted. CEO’s sample consists of 19,345 applicants, of whom 12,219 were admitted. That’s a further omission of 2,433 applicants, of whom 42 were admitted.
CEO doesn’t provide a breakdown by race of the data pool it used, so it’s impossible to say with absolute precision what numbers they used. But it is possible to reverse-engineer their dataset by applying the percentage totals they arrived at for each race to the aggregate numbers they provided. When you do that, this is what you get:
Of 854 black applicants, CEO’s sample included approximately 503. Of 769 Latino applicants, they included approximately 600. Of 2,038 non-Southeast Asians, they included approximately 1,528. And of 18,117 whites, they included approximately 16,714.
What does this mean? It means that about 10% of UW applicants who fit CEO’s racial parameters were left out of their study’s sample, with those exclusions coming disproportionately from student of color applicant pools. Only 8% of white applicants were left out of the CEO report, as compared to 10% of non-Southeast Asians, 22% of Latinos, and a staggering 41% of blacks.
So who were these students who were excluded from the sample?
CEO says that “cases with missing academic data were dropped from the statistical analyses,” as were applicants whose inclusion “might lead to the identification of an individual.” It’s clear that most of the applicants excluded from the CEO pool were students who were not admitted to UW, presumably because they submitted incomplete or otherwise unsatisfactory applications. But whatever the reasons for their rejections, they did apply, and CEO has simply erased them from the record as if they never existed.
Another erasure is the exclusion of Southeast Asians from the category “Asians,” noted above. Nearly a dozen times in the first page of the report, CEO compares statistics for Asians with those of other racial groups without noting once that their definition of “Asian” excludes more than fifteen percent of Asian applicants and admittees to the university. (CEO mentions the distinction between Southeast Asians and other Asians only twice in the report — both times in discussions of retention rates.)
CEO claims, as noted earlier, that UW’s admissions rates for black and Latino students are dramatically higher than those for whites and Asians. That claim rests on the exclusion of more than two thousand students of color from their applicant sample, an exclusion with major implications for CEO’s analysis.
5 comments
Comments feed for this article
September 16, 2011 at 11:53 am
Brad
Dr. Johnstone,
I really admire you’re analysis of these data, but you should consult with some admissions professionals regarding these data. COE obviously went in with an agenda that I disagree with. Still, that does not mean that their data or analysis techniques were wrong or that their conceptualization of admissions numbers is somehow inaccurate.
A few notes:
-Admit rates are often calculated from the number of total applications. This decreases selectivity and makes it look harder to get in for publications like US News and World Report. This is why so many schools have the “snap apps” which are already filled in and just require a student to click the mouse to apply. Ever wonder why Tulane receives more applications than Harvard?
-However, modeling admission or enrollment decisions must come from those who completed their applications. There would be no way to understand SAT averages for students who did not submit an SAT score or any other piece of information required for analysis. Obviously it leans the data, but this is a common statistical problem and there is literally a whole discipline in statistics which looks at missing data problems. My guess is that a HUGE bulk of COE’s missing data comes from the discrepancy between UW counting ANY application while COE was only able to analyze completed applications. This is somewhat of a non-factor since I can almost guarantee that UW admitted not a single incomplete app.
-Completion rates can be very, very low. After all, admissions offices recruit heavily to boost the application numbers, but have less incentive to get those applications complete. Also, as it gets easier to apply, more students end up in the pool who have little intention of actually going through the process. Different schools count applications differently, and I have no knowledge of what UW considers it’s “intent to apply.”
-Native Hawaiians and American Indians were probably very small numbers and therefore excluded. The “Other” box can mean anything and was therefore rightfully excluded.
-Excluding those who didn’t respond to the ethnicity question is common in reporting admissions data. There are a number of theories why students check the “Not Responding” box. Arguments have sprung up suggested it can be anything from Asians who are trying to give themselves an edge (based on the well documented statistical bias against that group). It can also be white students who don’t agree with affirmative action or even minority students who are philosophically opposed to affirmative action. Others might suggest it’s for privacy and not relevant. Either way, admissions offices can not impute these data and must take it at face value. The COE study probably could have included each category separately, but their reporting is not too far from the norm.
-Disccussing Asian populations is difficult but certainly deserves some airtime. It is not uncommon to think about Asia as a series of very different countries (which it is). Students from Chinese, Korean, or Japanese descent are not necessarily underrepresented in American higher education and therefore might not be considered for affirmative action initiatives. Students from Laos, Cambodia, Thailand, Vietnam, Malaysia, etc. are often considered as new immigrant populations with fewer resources, less representation, and greater likelihood of asylum or refugee parents. Again, there is no prevailing theme on this. check out this article for a better idea: http://www.insidehighered.com/news/2007/11/20/asians
I hope this helps clarify a bit.
*Views are my own*
September 16, 2011 at 12:08 pm
Angus Johnston
Brad,
Thanks for writing. My objections, as I hope I’ve made clear, are primarily to the ways in which CEO has misrepresented and obfuscated its data. To claim that “Asians” have an admissions rate of X or an average SAT score of Y without ever noting that you are excluding 15% of Asian applicants from your sample is improper. Similarly, to exclude 41% of black applicants from your pool without ever indicating that you have done so is misleading.
There were, comparing UW’s numbers to CEO’s, several dozen admitted students who were excluded from the CEO study on non-ethnic grounds. It appears that most, perhaps all, of these students were students of color. It seems clear that most of those excluded who were not accepted were excluded because of incomplete applications, but that’s only an inference on my part.
CEO has an obligation to properly account for its methodology, and they have not done so.
September 16, 2011 at 12:50 pm
Brad
Respectfully. But what would your response be if you found out that their numbers could be replicated and were right the whole time? What if UW never officially confirms but never officially refutes?
Perhaps the discussion is better served by defending the importance of diversity and the ability for colleges to proudly extend those opportunities even if that number hits 1000:1.
September 16, 2011 at 1:07 pm
Angus Johnston
Don’t we need to do both, though? Don’t we need to be honest about the data and be forthright in defending our values?
By my reading of the statistics, UW admittees are broadly similar across races. Black and Latino students tend to have somewhat lower test scores and class ranks, but compared to their ethnic peers they tend to be somewhat higher achieving. (The average white UW admittee in CEO’s study had SATs at about the 85th percentile of American college-bound whites, for instance. The average black admittee had SATs at about the 95th percentile of college-bound blacks.) And once they arrive, they tend to do similarly well, as indicated by CEO’s own retention data.
So yes, I’m happy to defend UW’s admissions process. It looks to me to be pretty solid, and it seems to me to produce good results. But part of what CEO is doing here is manipulating data to make that process, and the results it achieves, look like something other than what they actually are. And that part of the discussion is also well worth having.
September 18, 2011 at 7:03 am
Sunday Reading « zunguzungu
[…] people: Do Black Students Really Have a 576-to-1 Advantage in University of Wisconsin Admissions?, How Did CEO Arrive at Their Admission Rate Numbers?, and Final Thoughts, For Now, on the CEO Study of Wisconsin […]