SAT Scores and Family Income: The least surprising correlation of all time | Hacker News Reader

SAT Scores and Family Income: The least surprising correlation of all time

economix.blogs.nytimes.com

17 pointscwan27 comments

It's too bad these data aren't reported each year by the College Board as a scatterplot, to show the large variance in SAT scores at all self-reported income levels.

(which appears to be based on genuine data from the College Board, the one year it released such data) for another view of what might be going on.

After edit: See the other thread just opened here on HN also:

http://news.ycombinator.com/item?id=791381

What surprises me most about these graphs is how relatively flat they are!

Let's pretend for a moment that the distribution of SAT scores across income strata would be identical if all kids were forced to take the test. Given that a much greater percentage of wealthy kids take the SAT and the probability one takes it likely increases along with his expected score, you would think the correlation would be the inverse of the NYT graph. If a wealthy child is an idiot, his parents will encourage him to take the test in hope that some college somewhere will accept him. But, a dumb poor kid probably dropped out of school before he had a chance to even take the SAT. So, the positive correlation in the NYT article suggests a much greater disparity of test taking ability (and all that it might imply) than just the slope of the graph naively implies.

When I taught SAT at Kaplan I participated in a program called SAT Advantage where we offered SAT prep for free at high schools in low-income areas. I had several students score in the 95th percentile (Kaplan hires teachers who score 90th percentile or higher), and one got a perfect score in math.

It is possible with a little effort to beat the odds. The wealthy kids I usually taught had parents who could afford $1000 for an SAT class or $3000 for private tutoring. An idea I have for an educational technology startup would level this playing field using the web.

My impression was that it's pretty rare to move more than half a standard deviation by using test prep. And that median is skewed by the fact that scores fluctuate from one session to another, but the lower scores are discarded.

I went from a 1340 to a 1520 after a Kaplan course when I was in high school. On the average, massive score improvements are rare, but there are definitely some test prep teachers who can consistently get better score improvements from students. The secret is tailoring your approach to different students' needs. I always looked at the types of questions my students got wrong to identify patterns. If you can fix a bad habit you can get bigger score gains with less effort. My best success story was a student who went from a 980 to a 1420 (out of 1600). That's about 50th percentile to 95th percentile. Another student went from hoping to meet the NCAA minimum for a baseball scholarship to getting an academic full ride. The last student I taught almost beat my score...I took the new SAT in 2005 along with him and he got a 2180 out of 2400 (I got a 2280...99th percentile).

> It is possible with a little effort to beat the odds. The wealthy kids I usually taught had parents who could afford $1000 for an SAT class or $3000 for private tutoring. An idea I have for an educational technology startup would level this playing field using the web.

I once took an SAT prep class; it wasn't one of the 3k private tutoring classes, but I remember thinking, as I finished the last class, 'I learned and did nothing here that wasn't already in this mammoth SAT prep book'.

Libraries around here have tons of these SAT prep books, even for the newer SATs. I suspect that libraries (school and otherwise) even in your low-income areas have at least 1 or 2 decent SAT prep books.

What could your startup do for the kids in these low-income areas that they can't already do for themselves with the prep books available to them? If it is a lack of motivation or awareness, how would your startup motivate them to actually do the studying or get them more aware of the startup's resources than of the books?

Live instruction is more valuable than self-instruction to a majority of children of any age.

The point can be made by taking your question 1 step further.

If the information is available to kids via books, then why have school at all?

The obvious answer is that what a class buys is a sunk cost, to force the parent to send the kid to the class, and then the kid out of sheer boredom or osmosis will wind up learning something. If the kid had the motivation to begin with, the book would be quite enough. (I had the motivation simply because I wanted to show up my elder sister; so the class was a waste for me.)

Which directly leads to one of the questions I posed: how could this startup force the target kids to learn, if it's remote? Without a good answer to that, I don't think it's a good idea.

(And on a tangent: school can be a good idea, even for motivated students, simply because some things are very difficult to teach yourself out of a book. It'd be kind of hard to learn how to read and write from just books, for example.)

http://gregmankiw.blogspot.com/2009/08/least-surprising-corr...

wth. If I read that right, he just stated that intelligence is something inherited.

All the existing data says it is largely inherited

Boy is that false. http://bit.ly/2RzLIp

Edit: full URL is http://books.google.com/books?id=YdnK_PLtDVQC&pg=PA33...

First, please don't use shortened URLs.

Second, here's a few counterpoints:

http://www.sciencedaily.com/releases/2009/03/090317142841.ht...

http://www.nytimes.com/1990/10/12/us/study-raises-the-estima...

There is huge difference between saying

1. Smart parents -> Give their kids more exposure -> More Exposure and guidance make the kids smart

2. Smart parents -> kids inherit their smartness.

Greg was talking about genes , or case 2, and not case 1. which is what I think is unsubstantiated.

All the existing data says it is largely inherited

The existing data say "partly" rather than "largely." There is still a substantial influence of environmental factors found in any study of "broad heritability" (which is a characteristic of groups, not of individuals) of IQ. And at the individual level, there is a very large malleability of IQ through the best known interventions. Important data showing the powerful effect of environmental influences on IQ include the trends in raw scores on IQ tests over time in most countries of the world, resulting in the need to renorm IQ tests every half generation or so.

All these data and more can be found cited to primary sources in Intelligence and How to Get It: Why Schools and Cultures Count by Richard E. Nisbett,

http://www.amazon.com/Intelligence-How-Get-Schools-Cultures/...

a recently published book that is well worth a read.

The correlation between parental income and IQ also holds for adopted kids, which makes it hard to blame genetics: http://bit.ly/2RzLIp

Assuming parents adopt at random...?

The way I have always understood it, being smart isn't an extra above and beyond sort of thing(except for the truly gifted). It is avoiding all the possible ways of being dumb: getting good nutrition, getting enough sleep, avoiding diseases, getting new and interesting experiences, etc. Also while IQ does correlate well with income in general After you get to a certain level not so much(think professors and engineers vs upper level management )

It'd be much more interesting if they tried to tackle which specific factors made the most difference. (Based on my own anecdotal experience, rich people are not especially smarter than the poor or middle-class)

Perhaps a better question is which school did you attend? Or, did you take a prep course?

(Based on my own anecdotal experience, rich people are not especially smarter than the poor or middle-class)

IQ correlates better that parental socio-economic standing. But there are lots of factors that make this appear false -- the people you spend time with might be mostly of a certain class, or in a particular IQ stratum.

Based on my experience that's half true: rich kids aren't noticeably smarter than middle-class kids, but middle-class kids are noticeably smarter than the truly poor kids.

This is about what we'd expect, assuming a society which is reasonably meritocratic and intelligence which is at least partly inheiritable.

It's easy to explain that phenomenon without calling upon meritocracy or inherited traits: people with more resources (human and material) have more opportunity to become smarter.

Or maybe poor people value something other than intelligence. This argument works for both heritable and learned attributes.

There's one thing about it that surprises me a little: the comparatively big jump moving between the $180k-$200k and the $200k+ categories. Sure, that last category includes the veryveryvery rich as well as the merely quite rich, but I'd have expected -- wrongly, perhaps -- that both the link between (parents') intelligence and wealth, and the ability of wealth to purchase better education (in a broad sense). would be showing diminishing returns by that point.

Perhaps it's just a statistical artefact: the last category is relatively small. On the other hand, the absolute numbers are pretty big.

The scores are mostly all in the average range. I'd be curious to see the income graph for scores at -2, -1, +1, +2, and +3 standard deviations from the mean.

The graph shows the central tendency of scores (the median score) for each self-reported income range. As I posted in my first reply here, it is regrettable that College Board doesn't simply report a scatter plot of all score levels found in all self-reported income ranges.

And here's an example pointing to how this correlation can be broken: http://www.hcz.org/our-results

Reply on news.ycombinator.com