An affirmative action thought experiment

[Toni]

But the link also stated that Asian students are more likely than white students to drop out for ACADEMIC reasons. Please note that reasons for dropping out are self reported. I haven't seen data suggesting that there are objective measures saying that black students or Asian students actually perform less well than do white students. We don't know how well any group evaluates their own performance.

I have stated multiple times that I am not confident that the difference between 0.9% and 1.1% is statistically significant. But even if it were, I never claimed MCAT score was the only factor that influenced dropout rate. It would be madness for me to claim it, since I read my own source and I can see and have said that SES also affects dropout rate, even when MCAT is (very crudely) held constant.

As for objective data about performance, I would welcome it. I would be astonished to the core, however, if it revealed anything contrary to established facts, like that aptitude predicts academic performance, and therefore I would expect the higher MCAT scorers to do better, on average, than lower MCAT scorers.

Lost a post with lots of links and don't have time to re-do it all tonight.

BUT:

If you look carefully at the data of GPA/MCAT scores and acceptance rates by race, you will notice that there are black applicants who have GPAs and MCAT scores in the highest categories. You will also note that the acceptance rate for exceptionally high achieving black students are lower than the acceptance rates of exceptionally high achieving Asian students.

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

 

[Jolly_Penguin]

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

Toni, why do you keep equating socioeconomic status and privilege with race? If socioeconomic status is the issue then why not address it directly?

If you want to address racial discrimination and barriers people face simply due to race then address that.

It is legitimate to bring forward racial barriers, but you don't get to assume a given black candidate is poor and a given Asian rich just because of their races, and give out perks and penalties accordingly, without being called out for racism.

That is the crux of the opposition to affirmative action.

 

[Metaphor]

If you look carefully at the data of GPA/MCAT scores and acceptance rates by race, you will notice that there are black applicants who have GPAs and MCAT scores in the highest categories.

Of course there are. No-one ever claimed otherwise.

You will also note that the acceptance rate for exceptionally high achieving black students are lower than the acceptance rates of exceptionally high achieving Asian students.

I am hesitant to call the acceptance rate between groups statistically significantly different at the high end, because of the small sample sizes involved.

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

I don't know what data you had linked, but if you are referring to an overprediction error (that is, Asians do not do as well as predicted on USMLE 1, given their MCAT score), this is a study that looks at MCAT score and predicted versus actual USMLE 1. If you look at Table 5, I'd say the situation were dire for your arguments: only 69.3% of Black students pass, whilst 93.7% of Asians, and 96.8% of Whites, pass.

MCAT is a biased predictor alright -- but biased in the opposite direction for your argument to work. MCAT overpredicts the performance of African-Americans on USMLE 1 -- based on MCAT scores alone, nearly 25% of Black students were predicted to pass USMLE 1, but actually failed. If anything, the data show that we should be requiring higher MCAT scores from Black students than we require from White or Asians students, not lower.

The truth might be unpalatable but it does not stop it being the truth.

 

[Toni]

Of course there are. No-one ever claimed otherwise.

You will also note that the acceptance rate for exceptionally high achieving black students are lower than the acceptance rates of exceptionally high achieving Asian students.

I am hesitant to call the acceptance rate between groups statistically significantly different at the high end, because of the small sample sizes involved.

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

I don't know what data you had linked, but if you are referring to an overprediction error (that is, Asians do not do as well as predicted on USMLE 1, given their MCAT score), this is a study that looks at MCAT score and predicted versus actual USMLE 1. If you look at Table 5, I'd say the situation were dire for your arguments: only 69.3% of Black students pass, whilst 93.7% of Asians, and 96.8% of Whites, pass.

MCAT is a biased predictor alright -- but biased in the opposite direction for your argument to work. MCAT overpredicts the performance of African-Americans on USMLE 1 -- based on MCAT scores alone, nearly 25% of Black students were predicted to pass USMLE 1, but actually failed. If anything, the data show that we should be requiring higher MCAT scores from Black students than we require from White or Asians students, not lower.

The truth might be unpalatable but it does not stop it being the truth.

Sorry--a lot on my plate today and I don't have time to re-construct or go into more detail at this moment.

I am not at an academic institution and I am not able to read the entirety of the 1998 article you linked because I do not have access. The only portion I am able to read is the abstract. The quote below is from that abstract:

Results: The patterns of prediction errors observed across the groups indicated that, on average, (1) no difference between the sexes in prediction errors was evident; (2) performances of the three racial/ethnic minority. groups tended to be overpredicted, with significant findings for Asians and Hispanics; and (3) Caucasians' performances tended to be underpredicted, although the magnitude of this underprediction was quite small. When USMLE Step 1 scores were the criterion for success in medical school, the majority of errors were overprediction errors.

My emphasis in the above quote.

So, if I am reading this correctly: performance of Asian students (and Hispanic) was over predicted to a significant degree. While black students were not specified, it is a reasonable assumption that they are the third minority which was overpredicted but apparently not to a significant degree. Or maybe I am misunderstanding the authors of the study you linked.

I cannot view the tables or the body of the article so I must rely on the authors' statements.

The statement I had made in the previous post was based upon statements made by the author of another open access article,

 

[Metaphor]

Of course there are. No-one ever claimed otherwise.

You will also note that the acceptance rate for exceptionally high achieving black students are lower than the acceptance rates of exceptionally high achieving Asian students.

I am hesitant to call the acceptance rate between groups statistically significantly different at the high end, because of the small sample sizes involved.

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

I don't know what data you had linked, but if you are referring to an overprediction error (that is, Asians do not do as well as predicted on USMLE 1, given their MCAT score), this is a study that looks at MCAT score and predicted versus actual USMLE 1. If you look at Table 5, I'd say the situation were dire for your arguments: only 69.3% of Black students pass, whilst 93.7% of Asians, and 96.8% of Whites, pass.

MCAT is a biased predictor alright -- but biased in the opposite direction for your argument to work. MCAT overpredicts the performance of African-Americans on USMLE 1 -- based on MCAT scores alone, nearly 25% of Black students were predicted to pass USMLE 1, but actually failed. If anything, the data show that we should be requiring higher MCAT scores from Black students than we require from White or Asians students, not lower.

The truth might be unpalatable but it does not stop it being the truth.

Sorry--a lot on my plate today and I don't have time to re-construct or go into more detail at this moment.

I am not at an academic institution and I am not able to read the entirety of the 1998 article you linked because I do not have access. The only portion I am able to read is the abstract. The quote below is from that abstract:

Results: The patterns of prediction errors observed across the groups indicated that, on average, (1) no difference between the sexes in prediction errors was evident; (2) performances of the three racial/ethnic minority. groups tended to be overpredicted, with significant findings for Asians and Hispanics; and (3) Caucasians' performances tended to be underpredicted, although the magnitude of this underprediction was quite small. When USMLE Step 1 scores were the criterion for success in medical school, the majority of errors were overprediction errors.

My emphasis in the above quote.

So, if I am reading this correctly: performance of Asian students (and Hispanic) was over predicted to a significant degree. While black students were not specified, it is a reasonable assumption that they are the third minority which was overpredicted but apparently not to a significant degree. Or maybe I am misunderstanding the authors of the study you linked.

I cannot view the tables or the body of the article so I must rely on the authors' statements.

The statement I had made in the previous post was based upon statements made by the author of another open access article,

You should be able to access the full text; I was able to do so from home. Just click on the pdf link on the right.

The reference in the abstract is to prediction of the first two years of medical school grade point average, from MCAT and adjusted GPA. For half of Black students, grades were overpredicted (actual grade was less than predicted grade) and for the other half of Black students, underpredicted. White students had a slight underprediction, and Latinos and Asians slight overprediction.

My reference was to MCAT as a predictor of USMLE 1 pass or fail. Nearly a quarter of the Black sample failed USMLE 1 on their first attempt, but were predicted to pass based on MCAT score. (Two thirds passed and were predicted to pass, 6% of the sample failed and were correctly predicted to fail, and 2.4% passed but were predicted to fail).

As I said earlier, this direction of bias actually overpredicts Black performance and if you wanted to bring in race as an additional predictor, you should require Black students to have a higher MCAT score (if anything). But as the authors point out, the precision of MCAT would not be enough to justify such a measure.

Actuarially speaking, it would be justified to apply negative points to Black students in selection for admission to medical school, because their MCAT scores overpredict their success in passing USMLE 1. But I don't think anyone should be penalised for belonging to a particular group with lower (or higher) achievement, because I think discrimination based on race is generally harmful.

 

[Toni]

Of course there are. No-one ever claimed otherwise.

You will also note that the acceptance rate for exceptionally high achieving black students are lower than the acceptance rates of exceptionally high achieving Asian students.

I am hesitant to call the acceptance rate between groups statistically significantly different at the high end, because of the small sample sizes involved.

Further, Asian students--in the U.S., typically with well educated parents of high socioeconomic status (that is: privilege) seem to show peak performance with the MCAT scores but perform less well than expected on the USMLE 1. On the other hand, students from more humble circumstances improve when given the opportunity and perform well on the licensing exam.

I don't know what data you had linked, but if you are referring to an overprediction error (that is, Asians do not do as well as predicted on USMLE 1, given their MCAT score), this is a study that looks at MCAT score and predicted versus actual USMLE 1. If you look at Table 5, I'd say the situation were dire for your arguments: only 69.3% of Black students pass, whilst 93.7% of Asians, and 96.8% of Whites, pass.

MCAT is a biased predictor alright -- but biased in the opposite direction for your argument to work. MCAT overpredicts the performance of African-Americans on USMLE 1 -- based on MCAT scores alone, nearly 25% of Black students were predicted to pass USMLE 1, but actually failed. If anything, the data show that we should be requiring higher MCAT scores from Black students than we require from White or Asians students, not lower.

The truth might be unpalatable but it does not stop it being the truth.

Sorry--a lot on my plate today and I don't have time to re-construct or go into more detail at this moment.

I am not at an academic institution and I am not able to read the entirety of the 1998 article you linked because I do not have access. The only portion I am able to read is the abstract. The quote below is from that abstract:

Results: The patterns of prediction errors observed across the groups indicated that, on average, (1) no difference between the sexes in prediction errors was evident; (2) performances of the three racial/ethnic minority. groups tended to be overpredicted, with significant findings for Asians and Hispanics; and (3) Caucasians' performances tended to be underpredicted, although the magnitude of this underprediction was quite small. When USMLE Step 1 scores were the criterion for success in medical school, the majority of errors were overprediction errors.

My emphasis in the above quote.

So, if I am reading this correctly: performance of Asian students (and Hispanic) was over predicted to a significant degree. While black students were not specified, it is a reasonable assumption that they are the third minority which was overpredicted but apparently not to a significant degree. Or maybe I am misunderstanding the authors of the study you linked.

I cannot view the tables or the body of the article so I must rely on the authors' statements.

The statement I had made in the previous post was based upon statements made by the author of another open access article,

You should be able to access the full text; I was able to do so from home. Just click on the pdf link on the right.

The reference in the abstract is to prediction of the first two years of medical school grade point average, from MCAT and adjusted GPA. For half of Black students, grades were overpredicted (actual grade was less than predicted grade) and for the other half of Black students, underpredicted. White students had a slight underprediction, and Latinos and Asians slight overprediction.

My reference was to MCAT as a predictor of USMLE 1 pass or fail. Nearly a quarter of the Black sample failed USMLE 1 on their first attempt, but were predicted to pass based on MCAT score. (Two thirds passed and were predicted to pass, 6% of the sample failed and were correctly predicted to fail, and 2.4% passed but were predicted to fail).

As I said earlier, this direction of bias actually overpredicts Black performance and if you wanted to bring in race as an additional predictor, you should require Black students to have a higher MCAT score (if anything). But as the authors point out, the precision of MCAT would not be enough to justify such a measure.

Actuarially speaking, it would be justified to apply negative points to Black students in selection for admission to medical school, because their MCAT scores overpredict their success in passing USMLE 1. But I don't think anyone should be penalised for belonging to a particular group with lower (or higher) achievement, because I think discrimination based on race is generally harmful.

Yes, I did click on the PDF but can only get the abstract.

The authors' conclusion is that the over-prediction for Hispanics and for Asians is significant. It sounds as though the over prediction and under prediction cancel each other out in the case of black students, from what you stated above.

 

[Metaphor]

Yes, I did click on the PDF but can only get the abstract.

The authors' conclusion is that the over-prediction for Hispanics and for Asians is significant. It sounds as though the over prediction and under prediction cancel each other out in the case of black students, from what you stated above.

No. The abstract refers to predicting medical school grades using MCAT and adjusted GPA. The authors presented these results first.

The authors also created a model using MCAT to predict USMLE 1 pass/fail predictions. It is this model that overpredicted Black success by quite a large margin.

 

[Jolly_Penguin]

Happened upon this while watching other debates by the same group, "Intelligence Squared" so thought I'd post it: Does Affirmative Action do More Harm Than Good? debate at Intelligence Squared

http://www.youtube.com/watch?v=Bm5eph0g_Ag

 

[Loren Pechtel]

Happened upon this while watching other debates by the same group, "Intelligence Squared" so thought I'd post it: Does Affirmative Action do More Harm Than Good? debate at Intelligence Squared

http://www.youtube.com/watch?v=Bm5eph0g_Ag

Note how those who say it's doing harm go for facts, those who support it are arguing emotions and evading the point.