Wouldn't this mean if you compared two groups of kids from the same income quintile one with a very low SAT score and one with a very high one they'd have the exact same life outcomes? I find this hard to believe.
Also this study disagrees that coaching has much of an impact on SAT scores.
https://www.dropbox.com/s/h2pva2jus90c1el/2012-montgomery.pd...
Two wealthy kids, raised on the same neighborhood/conditions, measured with each other, SAT will give you a proxy for academic ability/IQ.
That comparison might break down when measuring A poor kid, with a wealthy kid. Or a ESL student (english as foreign language, to a native speaker).
The wealthy kids, probably went to better schools, had taken PRE-SATS since 8th grade, has tutor, and everything in between to get better. Poor kids are probably sharing a room with their parents, live in crappy conditions, and have to start working as teens to help out the family economically...
All upper middle class folks know this: the more testing practice you do, the better you become at that test...
SAT and IQ tests are a valid tool of measurement when two people have had the same conditions and training.
Let me translate into engineering terms: Two engineers go to interview, one is good and has some great experience, the other one is average, but just LEETCODES for months before the interview.
The average engineer gets the job, while the better one might not, as he probably didn't solve some Leetcode style of problem fast enough.
Im real life scenarios, the good engineer would have outperformed the average one.
Testing prep just doesn’t matter in aggregate, because admissions rarely have a cliff between marginal score and slightly better marginal score. Upper middle class people spending money on it doesn’t mean it’s money well spent; like many things rich people spend money on, it’s largely signaling to peers.
Also, your caricature of why people get hired for engineering jobs bears very little relationship to what I and my colleagues look for in engineering interviews.
Being able to solve a few simple algorithm problems is table stakes. Being able to talk about what you’re doing and why is what’s being measured.
People who crammed leetcode but don’t actually understand the material are obvious 5 minutes into the interview; people who freeze up occasionally on the algorithm bits but can talk about why they’re doing what they’re doing generally get hired.
Yeah, I'm not sure I've ever even heard of anyone comparing the times it took different applicants to answer a question, except maybe in the binary "did they do it under the time cap" sense. And even then, very rarely is software development time capped in the way an interview was so as long as someone's thinking is along the right track I don't think not getting a specific answer in a specific timeframe would disqualify them.
Even for official IQ test taking, they don't recommend taking the test more than once a year, as it invalidates the results of the tests...
Why? The act of doing tests, makes you better at that particular test, and the test ceases to become a good measurement of what it was trying to measure (in this case, IQ).
Honestly I would be surprised if it was possible to create a test that was both impossible to improve with practice and actually measured anything at all.
How far are we from trying to "correct"/normalize SAT scores according to family income quintiles? It seems to make sense if it's actually true that you can't compare scores across quintiles. As someone who did above-average and came from the bottom quintile it would have helped me a lot. As someone whose kids will very likely grow up in the top quintile I'm not sure I'd want that to happen.