William M. Briggs

Statistician to the Stars!

FiveThirtyEight Weighs In With Some Odd Affirmative Action Statistics

From FiveThirtyEight; see link below.

From FiveThirtyEight; see link below.

FiveThirtyEight ran the piece “Here’s What Happens When You Ban Affirmative Action In College Admissions” that ran some dicey stats.

Affirmative Action is, as all know, the practice of rewarding somebody for some physical characteristic but where the physical characteristic is beside the point. Implementing it always leads to quotas and to officially sponsored racism. This is always an irony because AA was meant to be the cure of racism. That instead of a cure it is corrosively iatrogenic is never recognized or acknowledged by those pushing it.

Some people do recognize it, however, which is why there have been many lawsuits seeking to ban it, and there will be many more. The Supreme (as they call it) Court is on new case: you can go to FiveThirtyEight and read about it.

One thing AA supporters often claim is that AA doesn’t lead to quotas. Now this is always nonsensical because that is exactly precisely what AA is: mandatory quotas. Don’t believe me, believe 538.

FiveThirtyEight was concerned that officially sponsored racism would be removed by banning AA, and that this removal would “hurt black enrollment”. They produced the figure at the top of this post (which I left in color) as proof of that claim.

It shows the percent of blacks (of a certain age) who live in the same state as one of several colleges by the percent of blacks who are undergraduates in those colleges (they don’t say these are the same age, which is a bias). Equality demands that these points either be the same percentage, or that the latter percentage is higher, which it is (if I counted right) in 16 colleges. (Equality does not mean equal. The dashed line represents equality, not Equality.)

The other two lines are unnecessary, eye-distracting fictions. They are replacements for reality and not reality itself. Ignore them, if you can. Which you probably can’t, because the lines seems to explain the dots, which they don’t. The lines are models which in our day are keen substitutes for actuality. They are not actuality.

Boy do those lines latch on the eye and shake it about, no? Remember our many, many, many lectures about not replacing time series with fiction, i.e. smoothers and the like? Same thing applies here. We do not need a model to tell us what we already know. Models tell us what isn’t true when we know the truth. Why? Because the models aren’t the reality!

Am I getting through to anybody?

Yes, nearly everybody does what FiveThirtyEight did, including professional statisticians. To paraphrase Lina Lamont, everyone’s a dope. Everybody is wrong. Never do this. The dashed line is harmless enough, because it represents a reality, which is where the points would align if equality (not Equality) held. But the other lines are nothing.

Anyway, that’s one error, and in the context a small one. The major one was to use these models to implicitly demand Equality, i.e. quotas. Whether quotas are moral (I have no idea about the law) is the real question, and that question FiveThirtyEight bypassed by jumping right to data that does nothing to answer this question.

If quotas are judged immoral and harmful, those points on the graph may lie anywhere, who cares where. If quotas are judged moral, then it’s a simple matter to ensure, by strict and immediate enforcement, to push all the points north (by north west) of the proportional representation line, and so gain by fiat the desired Equality.


  1. The paper on which the 538 report is based has more detail than can be digested in a quick scan, but I note that the authors “estimate affirmative action for colleges in various selectivity ranges (based on MSATACT)” which is each institution’s median freshman SAT and ACT entrance exam scores. They compare data from 1992 v. 2004 for colleges in “selectivity ranges (e.g., MSATACT >= 1100).” They don’t mentions any correction for the fact that scoring of the SAT changed in 1996 to account for the growing number of high school students of lesser ability taking the test. It was called “re-centered scoring” and had the effect of lowering national scores by 10-20 points. Additionally, the SAT and ACT tests are different, are the primary test in different regions, are scored differently, and scores do not precisely map to the other test. They use other national longitudinal studies to derive estimates and weightings, which may be valid, but complicate the analysis and bring some value judgements into it. Finally, over this time period many schools have de-emphasized test scores in admissions policies, so a basic criterion for the study may not have been stationary.

    So I question whether the data and it’s handling are adequate for their analysis.

  2. Surprising little “trend” between the two trends, actually. I’d take it as evidence that “affirmative action bans” don’t matter very much. Of course, if public opinion ever catches up to the reality of the net disadvantage of higher education, then anything that keeps under-represented minorities out of school should be hailed as heroic.

  3. Gary,

    If what you describe is true, then that’s a real bad sign for the paper.

    I’m still hung up on the charts in the article. No adjustments for income, college costs, admission requirements, number of applicants, etc. For all we know there’s a Simpson’s paradox running around in there. The fitted lines don’t even seem all that much different. Even a ‘first or second way’ analysis isn’t given in the article to help me believe that the lines are significantly different.

    Notice also that it’s “Public Research Universities”. The underlying assumption is that apparently you have to go to one of those schools to be successful. What’s wrong with non-research universities? Can’t they help people improve their lives at all? What about overall enrollment and success rates in these states? Also, just because a state doesn’t have a ban, what about the actual universities themselves? Do they implement AA or not?

  4. Agree with those above, paying attention to the lines, there really is little difference.

    What strikes me is that the “ban” ends where population composition is around 23% for a state. That suggests a couple things as well (both in a positive light and negative light) when a state has few blacks.

    (Maybe there should be enforced population compositions within states!)

  5. One could also take issue with the focus on enrollment. Presumably graduation is the desired outcome, not enrollment per se. Assume for the sake of argument that admission quotas are morally justified. Do they result in similar proportions of graduates? Maybe not.

    Colleges can increase their minority enrollment numbers by admitting unqualified students that fail out later. Colleges have more incentives, and even legal obligations, to admit minorities than to graduate them.

    Some have argued that affirmative action has not increased the *number* of minorities in higher education, only their *position* within higher education. A student who would have graduated from certain level of college may be admitted to and drop out of a college a notch up.

  6. James,
    Yes, nullius in verba. If you have the slightest credulity for Wikipedia see https://en.wikipedia.org/wiki/SAT for confirmation.

    The AA quota idea is becoming passe. Schools are trying hard to boost “diversity” numbers. It’s happening anyway because of changing demographics. For the last decade the buzzword has been “access.”

    I disagree that schools have incentive to enroll minority students and then ignore them. Accreditors and stakeholders are keenly interested in completion rates, the primary indicator of institutional “success.”

  7. Wow. This is the first time I actually visited 538, and boy, for a “statistics” site, I think bright high school students could have made better graphs.

    Even more telling, only about 10% of the comments actually discuss the statistics. At least those folks recognized it as an amateurish attempt to justify a POV with “mathy” pictures.

  8. Well, the only person who seems distracted by trend lines here is you, Briggs. The substance of the matter is that there will be less black and Hispanic students getting higher educations without some kind of AA program.


  9. If that’s true, you sure wouldn’t know it from the argument made at 538. But when you know everything, who needs to competently analyze data?

  10. No JMJ. As previously stated, this discusses enrollment, not graduation. We have no idea whether those enrolled received a higher education.

  11. JMJ: So you are saying black and hispanic students don’t have the grades to get into college without quotas? I also wonder, as did Jerry M, if graduation is included. If not, there’s not much of a point to enrolling students that won’t finish?

  12. Sheri,

    “If not, there’s not much of a point to enrolling students that won’t finish?”

    Sure there is, their tuition subsidizes the successful students. The university need not worry about their lack of success because they get paid anyway. Even if the student is left unemployable, it’s not the university that gets left holding the bag when a student can’t pay back their loans.

  13. In the Other AA, the first step is Admission. The schools seem to have take this to heart by employing an equivocation.
    Got step 2 right as well: belief in a Power greater than ourselves: aka Government.

    Oddly, all of the other 10 steps are followed allowing for equivocation. Maybe this is saying something about the meaninglessness of the 12-step program.

  14. Gosh, I look at the lines that they have so nicely draw for us… I know that our dear professor wants us to ignore the lines… and the first thing I say to myself is that there are very good reasons why the intercepts on these lines should be 0, ban or no ban!

    And then after that, the lines are parallel, there is in fact no difference in the correlation coefficients. My frequentist upbringing screams “statistically insignificant.”

    In fact the orange line is steeper. This suggests that states with a high percentage of black students would better serve the black population if they instituted a ban!

    I will stick with, low signal to noise in this picture, and the lines that are supposed to help us see the signal are in fact muddying it.

  15. Assuming blacks are socially inferior to whites for cultural and historical reasons — which is the only possible, or at least politically acceptable, argument for allowing positive discrimination — does this course of action (granting blacks with less merits than whites favours they would not otherwise obtain based on achievement alone) benefit society in the long run? Probably yes, especially in a welfare based society.

    But is it moral? For some it will be, for others not. It’s at least pragmatic, until it becomes an end in itself.

  16. One thing AA supporters often claim is that AA doesn’t lead to quotas. Now this is always nonsensical because that is exactly precisely what AA is: mandatory quotas. Don’t believe me, believe 538.

    Perhaps if I didn’t know how AA works, I would’ve believed the above.
    Just like, e.g., extra points given to students from the Upper Peninsula of Michigan, extra points assigned to a certain minority group are not quota. There is no denying that the extra points would increase an applicant’s chance of being accepted.

    But no, no evidence of quota at 538 as imagined by Mr. Briggs, who is statistician and should be able to tell his readers correctly what kind of evidence would allow one to support the claim of AA = Quota. No proof either. Simply what JMJ says in his comments.

    Yes, I expect Briggs to say that I am wrong without any arguments again.

    Perhaps the small difference between AA-ban and some non-AA-ban universities imply that not many black students have actually benefited from the extra points. I am just guessing, and cannot claim to have any proof.

    Briggs, your political agenda has made you a biased statistician.

  17. I thought the 538 measure of “success” was “proportional representation”. Colleges that were failing in AA were not at (or above!) the proportional representation line. Is there a difference between “proportional representation” as a “desired outcome” and the definition of “quota”?”

    “a fixed share of something that a person or group is entitled to receive or is bound to contribute”


  18. John M,

    I thought the 538 measure of “success” was “proportional representation”. Colleges that were failing in AA were not at (or above!) the proportional representation line

    Are 538 people the ones who decide the AA policies and how each school should admit students? The 45 degree angle line is a commonly used tool to show the difference between two variables, e.g., black share of state population (18-24) and black share of undergraduate enrollment.

    So what does the evidence that colleges were failing the 538 measure of success tell you? Could it be that quota is not the admission policy? (Of course, we cannot really know from the data analyzed.) Again, race is a factor in some university’s dmission policy. If extra points are added to an black applicant evaluation results, it may make a difference in the admissions decision. It is unlikely that a flagship university would use quota.

    The conclusion is succinctly summarized in the graph: banning affirmative action hurts black enrollment.” How much is the enrollment hurt? I don’t know.

    Even if indeed the 538 measure of “success” is “proportional representation,” (my measure of success is “33% Asian students in Ivy League”), still, no evidence of quota

  19. “Are 538 people the ones who decide the AA policies and how each school should admit students?”

    I thought we were talking about the 538 article. What are you talking about?

    And I can understand why you ignored the literal definition of quota.

    Much easier to just say “no it’s not.”

  20. John M, *sigh*

Leave a Reply

Your email address will not be published.


© 2016 William M. Briggs

Theme by Anders NorenUp ↑