I'm not going to get into correlation does not imply causation bullshit with you. I understand statistics very well. Point is, admission committees don't care about statistical relevance, they care about simple measures. Since when was your GPA between the ages of 14 and 17 a predictor of success?

I'm not talking about correlation vs. causation. Considering correlation alone, "most billionaires are male" does not mean that selecting only males will help maximize the number of billionaires you pick, especially since billionaires are a small group while males are a large one.

Suppose 1% of males are billionaires and 0.1% of females are billionaires. (Then if the total population is half male, this means 89% of billionaires are male.) Eliminating females from your sample will increase your "billionaire rate" from .55% to 1%.

Now suppose that 10% of billionaires are children of billionaires; half of those are male and half female. Let's say 1% of both males and females are children of billionaires. Suppose your college has two applicants, a male child of a billionaire, a female child of a billionaire, and a male child of non-billionaires. Which has a better chance of being a billionaire?

In this (totally contrived) example, "child of a billionaire" is a better predictor for "billionaire" than "male" is (even though "billionaire" a predictor for "male" than for "child of a billionaire"). The college that wants to maximize billionaire alumni should not select only males. Instead, it should choose children of billionaires over all others. Then the rate increases to 5.5%.

Your point is well taken. There are certainly far better predictors than gender. I chose to use male as my example because it is (1) Simple (2) Something the admissions committee has access to (they may not have access to the financials of the parent) and most importantly (3) It fits within the broader discussion of race and gender.

I've talked to a few admissions staff and they want nothing more than to find a better predictor of success than GPA.

It is certainly not perfect, but its still better than nothing, which is why it is usually supplemented with SATs and interviews.