Hacker News Reader

Comment by rdtsc | Hacker News Reader

The linked report (exhibit A) is interesting to read:

http://samv91khoyt2i553a2t1s05i-wpengine.netdna-ssl.com/wp-c...

It would seem at first that Harvard wants to help poor students and race just happens to be a proxy for socioeconomic status, but it turns out that's not what they are doing:

> Harvard’s preferential treatment of African-American and Hispanic applicants is not the result of efforts to achieve socioeconomic diversity. Rather, preferences for African Americans and Hispanics are significantly smaller if the applicant is economically disadvantaged.

It seems they want to have their cake and eat it too. They want to appear diverse based on physical features. So if someone takes a picture of the class and publishes it everyone would say "Look how diverse they are". But they also don't want poor students because they want donations to keep streaming in.

View on HN

Yep. Affirmative action has biases of its own, should be blinded to prevent it. Let it be based on family income only.

In Australia university admissions are based pretty much solely on one number [1] that represents your high school results (unless you didn't go to high school in Australia), plus bonus points if you are disadvantaged (for which race is not relevant unless you are an indigenous Australian or Torres Strait islander).

No interviews, no essays, just high school results plus some data on disadvantage. And handled by a separate organisation, not by the university.

It seems like a pretty good system. Loads of disadvantaged people coming through my university all the time (...I've been here far too long).

Perhaps hard to implement in the US since there is no national high school curriculum and the country is large and diverse making implementing one difficult.

[1]: https://en.wikipedia.org/wiki/Australian_Tertiary_Admission_...

> race is not relevant unless you are an indigenous Australian or Torres Strait islander

So... race is relevant, then? I mean, that's a pretty big "unless".

I imagine there’d be a lot less opposition to affirmative action in the US if it only applied to descendants of Native Americans and Black people whose ancestors had been held in slavery in the US. Affirmative action as reparations plays a lot better than a spoils system which is what they have now.

Aboriginals and Torres Strait Islanders make up only 3% of the Australian population, so it's not that big.

Well, sure. Which sort of reduces to "Australia doesn't have a problem with affirmative action because it very racially uniform". It's not like[1] the US can adopt these policies by deporting it's minority population in services to it's approaching-mere-plurality-status white cohort.

[1] No matter how hard the current administration seems intent on trying.

same deal in canada although for doctors and lawyers the bamboo ceiling is real...

The Australian National University, headed by the Nobel Prize winning Brian Schmidt is going to change their admissions policy to be more like the US.

http://theconversation.com/anus-new-entrance-criteria-wont-d...

Amongst Australian academics there is a push to change ATAR allegedly because it reflects socio-economic status (SES) too much.

https://theconversation.com/should-we-scrap-the-atar-what-ar...

Andrew Norton, a researcher on academic policy points out however, that ATAR is a very good predictor for success at University.

https://andrewnorton.net.au/2018/06/14/does-atar-measure-mor...

FWIW having been through the Australian university system and having worked at US universities it seems that the Australian system actually rates pretty well. We don't have researchers of the quality of institutions that pay a huge amount but there is far less grade inflation (really compression) than in US universities because you're graded on a curve in the end. In other words, only 5% or whatever of a subject are expected to get HDs (As in US terms).

Replacing ATAR with more holistic admissions is like the campaign to get rid of the GRE in graduate admissions, a combination of the ignorant and venal. The well connected, rich and prestigious will get ahead just fine in a system where academic merit is not the one and only measure.

Amusing that ANU is changing to keep out Asians the same way the Ivy League did to keep out Jews.

The ATAR isn't broken as such, the process of getting one is.

For a NSW student a good ATAR means you can ace a test after 2 years of study, where as Canberra/ACT focuses a lot more on project work and has much more flexibility in subjects.

The difference of experience between Canberra/ACT high school leavers and NSW school leavers is staggering, especially when it comes to project work.

The Canberra secondary school system is incredible, and the rest of Australia could learn a lot from it. Happy to elaborate if anyone is interested.

Is grading to a curve the norm at Australian Universities?

They don't seem to do that at UNSW except in special circumstances.

It's quite strange coming from the Canberra secondary school system where everything was scaled to a consistent mean and stdev.

I think Canberra's secondary school system (years 11 & 12, 2 years before Uni) is light years ahead of other Australian states. It's great how you can focus on individually driven project work rather than studying state-wide standardized tests.

This is largely true, but not entirely. UNSW definitely had a program for accepting engineering students who didn't get the appropriate UAI, based on other criteria. I don't remember what it was called, but I definitely applied for it because I spent a good chunk of school fucking around and wasn't sure how badly I had messed up my grades.

Might be Faculty of Engineering Admissions Scheme (FEAS)?

Though I believe there are also bonus ATAR points for students from regional schools.

https://www.engineering.unsw.edu.au/study-with-us/feas

> In Australia university admissions are based pretty much solely on one number [1] that represents your high school results (unless you didn't go to high school in Australia), plus bonus points if you are disadvantaged

How does this work with universities that have different focuses?

From the information at [1] and at your Wikipedia link, my understanding is that the way this number works is as follows.

• There is a large list [2] of standard courses offered by high schools that can count toward your ATAR score. Each course is either 1 or 2 "units". These include courses in English, foreign languages, science, math, history, business, arts, and others.

• This list is divided into two sections. The second, smaller section is courses that are lighter academic rigor and less demanding. The courses in the first section are called Category A course, and those in the second section are Category B courses.

• Your ATAR is calculated from your best 2 units in English, and your best 8 units from the rest, which can include no more than 2 units of Category B courses.

Suppose you have three students, all with the same ATAR number.

Student #1's score came from English, Biology, Chemistry, Mathematics, and Physics.

Student #2's score came from English, Dance, Drama, Music 1, Music 2, and Visual Arts.

Student #3's score came from English, Ancient History, Latin, Greek, and Modern History.

Do the universities these students apply to just see the ATAR number, or do they also see which courses went into its calculation?

If they all applied to whatever Australian school best corresponds to MIT or Caltech, for example, would the school be able to see that student #1 is probably the one they want from these three? If they all applied to Australia's most Julliard-like school, would it be able to see that student #2 is the one they want?

[1] https://www.uac.edu.au/

[2] https://www.uac.edu.au/future-applicants/atar/atar-courses

Specific university courses will have additional requirements, for example my science course required an ATAR of 95, but also required a minimum score in a maths unit, and required that you completed some number of science units (I don't exactly remember). So based on the requirements, you couldn't ace English and then enrol in a science degree, for example.

In practice the minimum scores required in specific units are more easily achievable than the ATAR, so the ATAR ends up being the limiting factor most of the time.

Here is an example:

https://www.monash.edu/study/courses/find-a-course/2018/scie...

You can see the scores required in individual units are all 30. 30 is the mean by definition, whereas an ATAR of 95 means you're in the 95th percentile, so obviously if you're in the 95th percentile it's pretty likely you were above the mean in the relevant units.

> Do the universities these students apply to just see the ATAR number, or do they also see which courses went into its calculation?

I've been told by an administrative person in my department that there is in fact no system in place to verify that the students have completed the prerequisite high school courses or their scores. This is surprising to me but there you go.

My memory is unreliable so this might not be quite right but I think I heard from the head a department that they just get ATARs and whether the applicant is from a socioeconomically disadvantaged background. So if that's the case and is true for other universities, then the advertised subject requirements etc mostly have the effect of encouraging students to self-select. Presumably after the fact, the university could ask a failing student to provide their high school transcripts, but basically they're not enforcing it up-front automatically.

But I also get the impression things are just not as competitive as in the US. Most people live in the major cities and just attend the best university that their score allows, out of the universities in their own city. They don't care enough about going to the best school to move cities for it. The differences between the top 8 universities are not great enough for it to be worth moving when there is likely one in your city already.

I always assumed there are racial disadvantage for an Asian male applicant, but it seems to be worse than I thought.

> Race plays a significant role in admissions decisions. Consider the example of an Asian-American applicant who is male, is not disadvantaged,3 and has other characteristics that result in a 25% chance of admission. Simply changing the race of this applicant to white—and leaving all his other characteristics the same—would increase his chance of admission to 36%. Changing his race to Hispanic (and leaving all other characteristics the same) would increase his chance of admission to 77%. Changing his race to African American (again, leaving all other characteristics the same) would increase his chance of admission to 95%.

There could be a case that it is not just for photographs but it is actually the marketing pitch for the value proposition of Harvard over, as the Tyler Cowen mentions, UC Irvine.

If you are trying to train and prepare a student to lead global teams, set global government policy, guide diverse groups toward a common solution, etc., you will need exposure to a diverse set of ideas and a diverse future network of fellow alumni to debate ideas with during your formative undergraduate years.

Otherwise, might as well save the $0.25 million in tuition and indeed go to UCI instead because otherwise it really is the same lectures, reading material and tests!

>If you are trying to train and prepare a student to lead global teams, set global government policy, guide diverse groups toward a common solution,

That's a pretty big if, people in power like to keep it in their network.

Ah - maybe a better way to phrase it is, if Harvard is pitching itself as a place a person can learn to lead global teams, set optimal global government policy, guide diverse groups toward a common solution, then the student body should probably not just be high test scorers pulled from high test affluent suburbs. It should be pulled from a variety of sources with a diverse set of socioeconomic, academic, artistic and cultural experiences.

I think that's actually a secret of Silicon Valley. If Apple needs to ask how a UI should be setup to handle the nuances from a certain country, they can go ask a person from that country. It'd be much harder trying to do that in Asia or even Europe.

Is that controlled for measures of merit? Or is that the accidental effect of correlation between ex. academic performance and socioeconomic status?

They build a succession of models each one in the sequence with more controls than the previous. As examples see tables B.6.1 and B.7.1. I think merit would fall under the 2nd model (SAT scores, academic index, GPA, ...).

Reply on news.ycombinator.com