This seems fine if you wanted to use the SAT to measure an outcome for a given level of inputs. If all you actually care about is the actual outcome though, then this doesn't actually help.
I guess there could be both use cases for test scores, given that higher education might actually find itself more concerned with showing improvement as an important metric. They might prefer students that show more aptitude for a given level of input, moreso than the total aptitude achieved.
I constantly hear employers mentioning that same desire - to have "lifelong learners" are employees. I have to believe at some level that there is some desire for a measure of absolute capability, too.
Past adversity is in the past. So if someone grew up in a poor environment but you're thinking of admitting them to a good college on full scholarship, they'll soon be progressing at the rate they're capable of without the adversity.
So the best predictor of future success in a good environment probably does include a correction for past adversity.
1) Why do you think their circumstances will change once they're in University?
2) There is the assumption that the environment in which you grew up in does not permanently affect the way you learn, while many studies have proven this to be untrue. The most critical years shaping the way you learn is in the younger years. Once you're past that, there is no way you can "make up" for it just by going to a better college.
A radical solution I believe is to completely subsidize all education and child rearing costs, such that every child will at least have the necessary basic conditions needed for them to excel. Of course, parents are always going to try and give their kids an edge, but at some point there'll be diminishing marginal returns. To relieve pressure on governmental funding, another radical idea is to couple this with an upper limit on the number of children you can have, so people will not "overburden" the system by having too many children.
1) For most selective colleges, they'll be living in a different neighborhood or different city, in a college dorm instead of their house, with classmates instead of parents & siblings. So almost everything is different.
2) Sure. Environment has both short-term and long-term effects. So a fraction of the environmental effects wear off. I dunno what the fraction is, but when you do the regression to find the weighting of SAT test score and adversity score that best predicts college achievement, it should find the right balance.
I'm not trying to say there's no correlation from some subgroups. Obviously there is.
I'm trying to say that if you're trying to predict performance in a new environment (college) you can improve the prediction by correcting for factors (bad home environment during high school) that affect current test scores but won't be active in the future.
How large the correction should be can only be answered by large population studies.
Good point. It would be extremely useful to be able to pinpoint factors that are transient vs. those that stay with someone and continue to affect future performance..
Correlation between two factors doesn't tell you how to compare two sets of values. Kids who have to work after school until bedtime won't get as good grades in HS, but if no longer made to do so, may perform substantially better in college.
And that's the key. You're positing that the disadvantage magically stops in university.
I work in law school admissions prep, and you see these disadvantages all the way down the line: SAT, college grades, LSAT, law school grades, Bar exam passage rate
There are some cases where it stops. Maybe the college board can identify which variables may indicate a poor correlation. There might be some factors that indicate transient issues and some that indicate worse lifelong expectations.
But I'd be cautious about too readily assuming that a cause of a lower SAT will vanish in later life.
Actually, I should have been specific: you see disadvantages related to ethnic group down the line. If a group has a lower SAT score, they generally have the same gap on the LSAT, GRE, GMAT etc.
This is heavily related to socioeconomic status in the US. IIRC, in the UK for example you don't see these same gaps between ethnic groups. So it's something specific to US social circumstances, and not racial.
I don't know so much about particular circumstances, I just know what the high level stats say, and am inferring from that. I'm not American, so I don't have lived experience of the class structure there.
There are probably studies about income, but the test makers tend to only collect official data about ethnicity. It's commonly used as a proxy for social class, but I wish the test makers had better data about the elements the College Board is trying to capture with this new policy.
It's about one standard deviation between high scoring groups and low scoring groups across the tests.
LSAC no longer has theirs posted publicly, but it was much the same.
So, to the extent ethnicity is a proxy for social class in the US, we can say that the disadvantage these groups have at SAT time doesn't vanish by the time they take graduate level exams.
I wish I had some data for you that was based on social circumstances other than race, but I don't have deep knowledge in the domain, only what I've seen looking at the reports produced by the test companies recently.
How does one determine how large the correction should be? Since the system is all opaque and political, what prevents one from setting a correction factor of 10x, making sure no child of --bourgeois-- middle-class parents will ever set foot in an university? Such a policy was implemented in Eastern Europe under occupation by Stalin's armies in the aftermath of WWII. The economic and social results became quite apparent a few decades later, when the whole system collapsed.
I guess there could be both use cases for test scores, given that higher education might actually find itself more concerned with showing improvement as an important metric. They might prefer students that show more aptitude for a given level of input, moreso than the total aptitude achieved.
I constantly hear employers mentioning that same desire - to have "lifelong learners" are employees. I have to believe at some level that there is some desire for a measure of absolute capability, too.