Not so sure of that.. The correlation extends even to the highest reaches of the income scale: there is little sign of diminishing returns! Does that mean that the unfair advantage extends to even the difference between fairly wealthy families and very wealthy ones?
What distribution would have you thinking that the test actually does measure potential? Just trying to get at the underlying model of the statement.

It would be great if you could get a comment from the College Board or ETS on their fairness guidelines, because they have strict rules about not relying on class- and region-based knowledge.

I agree with the reasoning behind the first explanation of the correlation, but not the second. I'm an SAT teacher, and one of the first things we have to get our kids to understand is that the SAT is a reasoning test, and that the students won't need any outside information to answer the questions correctly. They don't need to understand computers if there's a passage about computers, and so on.

I think that you have to at least acknowledge the possibility that the SATs measure "intelligence" in some sense, that whatever you consider to be "intelligence" is at least partially genetic, and is positivitely correlated to income. I get that that it's not consistent with your thesis that income disparity gives an unfair advantage to high income kids, but if you don't even list it as a possible partial explanation? Your thesis appears to leave you unable to see the data.

A lot of this is access to coaching and economic means to take the test again and again. My husband tutored for an expensive prep program where he was often expected to coach kids way out of their talent range to get the highest scores possible. Never mind that he knew full well, having been to a top college, that these kids would get eaten alive if they got into the top schools that their parents expected them to get into.

I totally blew the lid off this curve - income less than $20 K in 1983 and scores above 1500 out of 1600 (two part test back then). So did my husband. We each only took the test twice because we couldn't afford to do it more. Other kids we knew had prep tutors and took it repeatedly until the desired score was attained. That's a big part of the disparity.

Well, aren't we all glad that Dr. Kate had an opportunity to brag about her test score! Good job, Dr. Kate!

The resource-rich schools can have benefits in multiple ways. First, the obvious in terms of smaller class sizes, more educational resources, etc. But also directly in terms of SAT prep. I went to a public high school in an upper-middle class suburb - the high property taxes and resources for the school meant that it was among the top 20 in the state. In addition to most students having the personal/family resources for SAT prep courses, private tutors, and retaking the test as needed, we had SAT prep in the classroom. While this was before No Child Left Behind and the increased emphasis on testing, the school was established, well funded, and not concerned with teaching to the state tests - therefore could spend time in math and English courses teaching SAT strategy.

I've also tutored for one of those expensive SAT prep programs. One of the key things that we repeated to students was that the SAT does not measure how smart you are or how well you'll do in college. The SAT measures one thing: how well you take the SAT. If the SAT really measured intelligence or college preparation/potential, how could access to prep/tutor programs raise someone's score significantly in only a few weeks?

Here in the UK a a recent study found that income had a much lower correlation with exam performance among recent immigrants. Although higher income earners still did well there were many more high achievers in the lower income immigrant groups than among the native population at the same income level.

The hypothesis of the researchers was that cultural capital (focus on books and education in the home, parents who engage their children in conversation, dining together, possible genetic effect on intelligence) in a stable rich society is linked in a self reinforcing way with higher income.

Recent immigrants have not had time for the sorting effect of living in a country where those with greater intellectual inheritance both become higher income earners and pass on that inheritance to the following generation.

There's a certain irony in this result.

It might be tempting for one to conclude that the SAT effectively a means to legitimize an advantage of privilege, in some sense continuing the "old boys networks" of yore. And it seems to some extent it does.

However I suspect on the grand historical scale, the SAT has by far been more of a force toward the opposite. Before the SAT, elite universities could argue on behalf of the distinction of elite preparatory schools and on down the latter creating an institutional network "legitimizing" an advantage of privilege. And if you go many decades into the past, you see exactly this: large fractions of Harvard students being children of Harvard grads, etc.

It appears there is still a distinct advantage along such lines, but over the last several decades an enormous wedge has been placed in it. Today, the vast majority of Harvard students are not children of Harvard grads, indeed by and large Harvard admits a class that is overwhelmingly exceptionally qualified by all standards. And part of this reckoning came from the SAT, which "proved" that there is exceptional talent outside of the elite preparatory system. I'd guess that nowadays the vast majority of Harvard students come from public schools.

is anyone else uncomfortable with what seems like an implied equation of household income with realized potential?

Has nobody read the comments on the New York Times site? It seems the problem lies with statistical interpretation and, particularly, presentation. Here's what one user said:

1) I suspect that the within-group variance is higher than the between-groups variance. But independently of my suspicions, this should have been reported, or at least the predictive power of income on individual test outcomes;

2) The first three charts correctly use a y-axis range starting at 0, while the 4th one doesn’t thus amplifying differences. ["the 4th one" here refers to the chart posted above]

3) Reporting test results vs incomes suggests a strong causal relationship among the two. But this relationship could be weak or even non-existent, after controlling for other variables.

In a related side note, Ben Goldacre's book finally has a release in the US/Canada (for those of you not in the UK). It's a brilliant read, but it also mentions statistical misrepresentation (particularly in newspapers), and has some good directions as to where to read up on statistics (for non-mathematicians).

I grew up in an affluent area, and my public high school had the top PSAT/SAT/ACT/whatever scores in the state. We would have turned our noses up at someone who scored as low as the top figure (Math < 580! and writing or 560; we weren't tested on critical reasoning). A very few kids got extra tutoring or took the test more than once (like 4 out of 200).

I think the test is a chicken in the egg thing. It does measure the potential to do well in our culture as the graph clearly shows. And it makes sense that kids that come from households that have done well in our culture will score high and go on to do well in our culture because they come from the genetic and culture stock that has done well.

Is it fair? Is it fair that a certain group of people have control and money and do well in school and do well in life? I say it depends on what you are measuring. If you are trying to predict how someone will do in the current culture, it is a fair measure.

However, if you are trying to argue that other cultural attitudes and/or genetic stock have the same potential in our current culture and we would see that if only we measured them on a different scale, well, then it's probably not fair.

The reality is that we do live in a culture that is dominated by a certain type, and if you don't fit into that type, you are at a disadvantage if you want to succeed within that culture.

As an analogy, tuna may be great swimmers, but they are never going to beat a tortoise in a land race.

Is that fair for the tuna? Is it morally correct for the tortoise? Is it what the tortoise really values?

Those are all different questions from one that would measure the potential of success for the tuna in a certain situation.

"Second, the test itself may be biased towards wealthier students. These tests tend to be written and evaluated by privileged individuals who may inadvertently include class-based knowledge, not just knowledge, in the exam (asking questions, for example, that rely on background information about golf instead of basketball)."

In 1972, maybe. A massive portion of the question preparation process is testing the questions for bias. The SAT includes an unscored section that they use to test out potential future questions, and they compare the results there with the self-reported demographic data to eliminate any bias. The infamous regatta question is doubly outdated, as it was retired over 30 years ago and was an analogy question, which have been totally eliminated.

The College Board is VERY sensitive to accusations of racial/sex/cultural bias, and take great pains to avoid any legitimate biases. One of the driving forces behind the addition of the writing section was to respond to California's complaints that males scored abnormally high compared to females.

I taught an SAT course at Kaplan for a summer, and then I just couldn't handle it anymore. It was as close to an exchange of money for success as I've ever encountered, and I felt quite dirty for taking their money. It was doubly bothersome because of the nature of the material; it's not, as many people think, a course that focuses on specific mathematical or verbal skills that tend to come up on the SAT. It is largely a collection of tricks designed to eliminate one of the answers entirely. Given the slight deduction for an answered but incorrect question, eliminating one of the questions makes guessing a value-added proposition, which has a significant improvement on your score all by itself.

Just to fully cement their evilness, the Kaplan people recruit high scoring SAT students like Dr. Kate's husband up there, most of whom did NOT use their type of program, to teach these tricks, giving them the appearance of legitimacy. All in all an entirely abhorrent industry, and not one that is easy to combat from any perspective. College Board can't do anything, government can't do anything, students who don't take the courses can't do anything, and parents just want their kids to take the courses. :(

das racist

I'm enjoying this discussion, but am a bit perplexed as to which of the multitude of theories concerning the nature of intelligence is at question. Measurements of intelligence are not very useful unless you know exactly what intelligence is.

I believe college performance is similarly correlated with income, and so if measuring college performance is the goal the SATs do a fine job; it just isn't a fair job.

I think I remember seeing that SAT performance could be correlated to the number of books in a house, which makes sense since it (at least used to) measure vocabulary. Kids spend more time at home than they do at school, and the more big words their parents use or explain to them while reading aloud, the better the chance they've encountered any given word before. I scored 780 on verbal despite being dyslexic and not having any accommodations, simply because I had read so much that there were only one or two words I hadn't encountered before (the questions I missed were all the stupid "what would be the correct title for this piece?" questions, which involved careful deconstruction of their title sentences and more exact comprehension of the readings than I had.)

I'm another SAT teacher/tutor, and I've taught students from wealthy to lower-income and very high-scoring to very low-scoring. I recently heard something about the SAT that rang true to my experience: High scores on the SAT tell you a lot about the test-taker, but lower scores tell you very little.

With high-scorers, I can see the gears turning in their minds while they puzzle out problems. They have a certain gift that I also have (as will any other SAT prep tutors, since it's required by the companies), and I think of it as an ability to "see through" the test--to use a dated reference, it's like being able to see the Matrix. They know what questions are really asking and what path they need to follow to reach the answers.

Average and low scorers are in an entirely different boat--or, more accurately, they're not in any one boat at all. Yes, some don't seem to be all that bright (though with my limited exposure to them I don't really know), but that's inevitable in a large sample of high schoolers and doesn't describe most of these students. Some don't get all the information because they're resentful that their parents made them come, are trying to show off by being disruptive, or are overscheduled and miss class because of other commitments. A large proportion of students can do the work if it's put directly in front of them, without the "trick" answers and misdirection that are part of the test. Some are baffled by the way the SAT presents information because they've never seen problems in this format before, but they grasp the concepts immediately once they're told what the question writers are getting at. Many are motivated, hard workers who will do very well in college but are put off by the artificial time constraints as well as the aforementioned presentation weirdness. Others have test anxiety and just psych themselves out.

The lowest-income students I've taught were a group of black kids from inner-city Philadelphia trucked out to the suburbs for a summer program. They were also the most interesting and diverse in their responses to the lessons. Some students didn't care at all, and I'm sure we could come up with many ideas about that. Some were vocal and inquisitive and picked up new ideas very quickly, but coming in their math skills were third grade at best, which seemed to me to be an obvious indicator of the quality of their educations rather than their intrinsic abilities. One day a student showed up totally unable to focus because he'd just learned a friend had been killed, and he'd come to class so he could see a few other friends there. Another student was always in the front row, raised her hand nonstop, and mostly had the right answers. While her skills were behind those of the wealthier kids I've known, with a different educational background she would have fit right in at my own private prep school. (As it was we had some interesting conversations about historically black colleges, and with her natural intelligence and motivation I have no doubt she went on to be academically successful.)

What I'm saying is that, from what I've seen, the SAT is a very, very crude method of measuring aptitude. In some students it demonstrates a particular ability that might have some real-world applications, but in most it takes a flattened, high-contrast snapshot of a much more complicated situation. I do understand why it's used, but we shouldn't mistake precision for accuracy.

I'm taking the PSAT tomorrow morning. The bookkeeper at my school gave me a booklet. I took it last year and got 94th percentile, and mostly know how to take the test (though I've taken so many tests it gets confusing weather or not I should guess). Interesting to note they allow use of a graphing calculator. For me, this means a TI-83 checked out from my school. It's programmable. Theoretically, I could load it with SAT-like words and look them up there. But I should sleep now.

One thing that is being avoided in this discussion is that domestic abuse is more common among poor people. It's possible that many poor students score lower due to the trauma of being beaten by patriarchal men or seeing their sisters and mothers beaten by patriarchal men.

I think the "cultural knowledge" explanation is actually hard to square with the monotonicity of the results. If that's the principal explanation, we'd expect the test to favor people with the same class background as the test-writers. Do standardized test question writers really make more than $200,000 a year? Maybe they do (the SAT is big business), but it seems unlikely.

If the question-writers are making, say a "mere" $120,000 a year*, how is it that the test nonetheless favors the cultural knowledge of the $200,000+ crowd? The 40-point difference from $120,000 to $200,000 is quite large—according to these data it's a jump of a bit over ten percentile points on each section. Maybe $120,000 test-writers could somehow be writing a test that favors the cultural knowledge of $200,000+ types rather than the cultural knowledge of their own income bracket but some explanation would need to be given.

Of course "cultural knowledge" could be part of the explanation but it seems very unlikely to be the principal explanation.

The "retaking" explanation referenced by some commenters can only be a small part of the explanation since the report is based on the students' most recent scores than their highest scores. Retaking the test produces an aggregate increase of about 40 points (i.e., 13 points per section) and presumably the gains rapidly fall off when the test is taken more than twice. Even if we assume all the $200,000+ students retake the test and none of the < $20,000 students do, that only accounts for 10% of the 120-point difference per section.

The absence of the genetic hypothesis is a bit striking. Given the number of leaps of imagination required to make the "cultural knowledge" hypothesis work it seems at least as plausible (though both seem way less plausible than the obvious "better prepared" hypothesis, which requires very few leaps of imagination).

* I pulled this number out of my ass but it strikes me as a lot more plausible than $200,000.

Keydar -
My intent was to add complexity to the discussion, not to over-simplify. I think that when you hear the word "genetics" it must drown out everything else, to the point that you assume I think those who are poor deserve it. That is not what I believe, and it's not what I'm trying to say.

Since I guess I need to clarify: certainly I agree that background (whether household emphasis on education, ethnicity, income, whatever you'd like to fold into "a skewed perspective of privilege") is a huge influence both on income and SAT scores. I also feel certain that traits are passed down from one generation to the next, but I don't feel competant to declare exactly which of those traits have genetic components, and which are simply passing down a privileged mentality (which, actually, I prefer to think of as passing on a love for books or some such).

But it doesn't seem insane to me that some of the traits that lead to a good score on the SATs might have a genetic component, and might also be contributing factors to landing and keeping a higher income job.

Not everyone takes SAT tests (at least not in Canada) and there appears to be little to no correlation between IQ and income as a rule. Well, at least beyond 120 or so on the Weschler scale, which appears to be the standard.

A couple people have mentioned this (or similar) already, but I'd like to reemphasize a point, especially since I feel like this and soooo many other articles I read about the SAT totally misunderstand what the test is designed to do.

Colleges like to use tests like the SAT because the are designed to PREDICT COLLEGE SUCCESS. Tests designed in this way are based on "predictive validity" (http://en.wikipedia.org/wiki/Predictive_validity), not on how well they test mastery of a certain skillset or how well they test intelligence or anything else. As long as higher family income is a predictor of college success, it will always correlate with higher SAT scores, otherwise the test is not doing its job.

At least in the 1980s, High SAT scores actually do NOT predict good grades in college, so if grades are a signifier of success in college, then the SAT isn't doing that job.

(I have no place to back this up because I got the info from a colleague who does a study of SAT scores and college grades with his statistics students every year. The teacher is not tech savvy so all the data is on photocopied and copied again poor quality handouts.)

It's the same with all standardized tests. Classes alone are expensive. How many lower class or even lower middle class can afford the 2 grand that many of them charge? Let alone what personalized private tutoring would cost (which is far far more effective). Not to mention the cost of study material (which can run in the hundreds even thousands when you consider GREs or LSATs). These tests seem to protect privilege far more than opening doors since they are not tests based purely on some "natural" intellect: they are learned and mastered. 

Or it could be that college success, which the SAT is designed to predict, is strongly linked with family income (http://www2.ed.gov/offices/OPE/AgenProj/report/theme1a.html).  I saw a study recently that found that among children being adopted those adopted by middle-income families ended up with an advantage of 12 IQ points over those adopted by low income families.

I think the policy implications are figuring out what the relevant differences are: access to books, better nutrition, lower stress, explicitly being taught social, meta-cognitive or executive function skills or whatever it turns out to be, and finding a way to bring those advantages to all children and to provide catch-up programs wherever possible.

When we think about what college is, the fact that 'class based' aspects would be on the test makes sense. College is preparation to enter or continue in the managerial, scientific, and clerical class of the society. 

(basketball vs golf is harldy the main kind of class-based issue i'd expect to really matter on the tests. More like Shakespeare vs van halen.

Sailing is to regatta as squash is to 
a. match
b. court
c. ball
d. volley

Poor people generally tend to do worse in all aspects of life.

From the poor man living in the mountains of Tibet, to the white trash of Tennessee.

I do not think that just because people have a higher family income means that they will make a high score on the SAT. I know plenty of people that have a very high family income that have made very low score (even when they had access to the tutoring and everything else you can get with a high income) I also know very many students whose families have a very low income that have made extraordinary scores on their SAT. If anything it might give the low income students a little extra motivation because they want to make a high score, get into a great college, have a great paying job, and then make more money than their family did because they want to give their kids a better life.......I mean that is just what a lot of people have told me that they feel about it. Personally I don't think it really matters what your family's income is. 

This chart is ridiculous. My family falls in the lowest income range but I received solid 700s on each section of the SATs. I am not just one outlier; many people I know that scored high were not necessarily form rich families. 

An interesting graph shown above, most definitely equal to the interest brought upon by the text scribed below. I would like to believe it wasn't so, but my rampant distrust of many things would lead me to believe that in fact these tests are prepared for upper echelon citizens. Er, perhaps my feelings on this are less than definite, as I am no expert by any stretch of the word on SATs and whatnot. I am merely writing a response to the article written above with little to no outside information. I don't know. I think it'd be interesting-terrible, yes, but still interesting-if these tests were built with the wealthy in mind, writing questions that the richer tend to have more background with? But I don't know. I would hope none of this is true. It'd be neat if someone had a collection of SAT questions that appeared to be written in attempt to befuddle poorer students? I can't think of any examples other than making questions about golf. Now my curiosity is piqued. Kinda wanna go sleuthing. Or for someone else to.

Why have you excluded the possibility that highly intelligent parents tend to have highly intelligent offspring? Highly intelligent parents will tend to make more money than parents with low intelligence, and they're also more likely to have offspring with similar traits of high intelligence.

Ignoring that factor and assuming that the children's high achievement is due to money would be like claiming that children of NBA players tend to be taller than average because they're wealthier. It would completely ignore what we know about heredity and the fact that taller parents tend to have taller children.

Could it be that higher IQ individuals score better on the SAT and IQ can be inherited from parents who by having a higher IQ make more money. Or we could just talk about privilege and economic privilege as an explanation.

The correlation is due to one obvious fact: if you are intelligent, you tend to make a higher income. Then given that intelligence is hereditary, you pass this intelligence and, hence higher SAT scores, on to your children.

grades don't correlate with SAT within a college very well, but that is an irrelevant statistic first infamously propounded by a Harvard dean as I recall which was ironic. The kid whom scores 99.9 percentile in both sections but goes to a state school gets straight As or close thereto. The 50% kids don't have a chance against him.

continued...

When all of the kids are brilliant, like at Harvard, grades depend on other things, like how hard you work, whether you parrot the political leanings of your professor, grade inflation in your major, etc.

I know I'm really late to the party on this, but the chart absolutely does not show that SAT scores correlate strongly with income. This chart only allows you to infer the sign of the correlation, (positive), not the strength. Based on this chart, you can't tell whether the correlation is 0.1 or 0.9.

Knowing something about the distribution of income and SAT scores, we can infer that the correlation is moderate at best, because an increase in income of 3-4 sigmas translates to only about a 1.2-sigma increase in mean SAT score.

Given that the correlation is not that strong even before accounting for heredity, the actual causal effect of household income must be quite small.

As a retired USAF fighter pilot I've learned that sometimes it comes down to "that looks about right" despite a panel full of instruments at my disposal. My crude rule of thumb on IQ vs SAT scores is total the math and verbal, divide by 10 and add 5 points for the relatively recent "dumb down" adjustment made to the test's bell curve and voila, IQ. Example: 1200 math and verbal, knock off the zero and add 5 for an IQ of 125.Of course it quits working at a perfect 1600.M/V score.

iQ