SATs and IQ

Can a person’s IQ be estimated based on his/her SAT scores? If so, what’s the conversion formula?

WAG ahead…
I would doubt it. The SATs test reading abilities and basic math. IQ tests measure an ability to think and learn. I don’t think a good performance on one is necessarily indicative of a good performance on the other.

::Educational Statistician enters the room::


Let me elaborate.


While we might say that smarter people score higher on the SAT (probably true) this relationship would be spurious (there may be a correlation between the two). This is mostly due to what the SAT tests. The SAT is the Scholastic Achievement Test while most IQ type tests (e.g., WAIS) is an Aptitude test. This is like measuring height and trying to discern weight from that information. Sure there is a relationship between height and weight, but the relationship is not causal (i.e., being tall does not cause you to weigh more, or being heavier does not cause you to be taller.)

As others have said before me the answer is definitely a no.

Think of it like this:

You have two people who test identically on an IQ test (such as WAIS). Say they both get a score of 130 (above average intelligence).

Now lets say that one person is raised in a poor family in the inner-city and the other is raised in upscale suburbia. Given the generally lower quality of the schools in the inner-city and the other ‘distractions’ those youth have to face (i.e. merely surviving) our inner-city youth may not have as much time to spend on his studies as his suburbia counterpart. As a result when they both take the SAT suburbia kid gets 1500 while inner-city kid gets 1000.

However, their IQ is still the same. Inner-city kid is as capable of learning as the suburbia kid. It’s the opportunity and access to resources to exercise that skill that separates the two.

I’ve wondered how many Einsteins, Mozarts and DaVincis have been born in less than ideal circumstances and lost to the world because the opportunity to manifest their genius was never presented to them.

While we can never know I’d bet dollars to dimes that it has happened…

Quick question, when did it change? I know when I took the SAT it was the Scholastic Aptitude Test. (Which isn’t to say that it measured aptitude, but that’s what they called it.)

My instincts would also say “No - you can’t measure IQ with the SATs”, but realistically, don’t both tests just measure the extent to which you think like the people who make the tests?

This has often been a criticism of the SAT and ACT. Some groups claim that the questions asked are relevant only to suburbia type kids and consititutes some kind of defacto discrimination against inner-city kids. I don’t know if that point was ever settled or debunked.

As for IQ tests they are obviously a bit difficult to nail down. What exactly is intelligence? Do such tests account for artistic ability or athletic ability or social ability? I don’t think so. What about some forms of autism where a person is severly mentally handicapped yet can exhibit nearly super-human intellectual feats in a very narrow application (ala Rain Man)? What would an IQ test have to say about that person?

All of that said they can be of some use anyway. Given at an early enough age (before a great deal of experience or formalized learning) you can test for problem solving ability. The ability to be presented with a totally new and unexperienced situation and be able to noodle your way through it can certainly be described as at least one aspect of intelligence.

So…IQ tests have some validity but take them with a grain of salt. I do not think they are absolute in their indication of a person’s native intellectual capacity in all aspects.

amarinth –

IIRC, the official name change took place about eight years ago, when ETS finally acknowledged that the tests measured acquired knowledge rather than inherent ability.

The SAT was designed as an aptitude test originally, hence the focus on abstract relationships (analogies) and problem-solving rather than any specific body of knowledge. Of course, a student’s ability to solve such problems still has more to do with experience than native capabilities.

FWIW, in my psych 100 class, we were told that there was a decent correlation between SAT performance and IQ. Something like a .6 correlation. Then again, this was pretty much smack dab in the middle of the lecture on causality (why might there be a correlation?), so take this with a grain of salt. Many of 'em, just to be sure.

::Guy with a reasonable knowledge of statistics walks into the room::


Let me elaborate.

Well… sort of. Let me speculate…

As Spritle suggests, there is almost certainly a correlation between the two, enough that you could probably say something like “Those who score between 950 and 1000 on the SAT have an average IQ score of x while those who score between 1001 and 1050 have an average IQ of x+4” and so on. (Just WAGging on the numbers here.) So I suppose you could make a crude - so crude as to be virtually meaningless - estimate of IQ based on SAT.

I disagree with Spritle that there is no causal relationship between the two. While a person’s IQ won’t dictate the SAT score it certainly will contribute to it.

Well, I guess I’m going to disagree with everyone else (on Preview, I see I’m no longer alone). The OP didn’t ask if a person’s IQ can be determined from his/her SAT score, he asked if it can be estimated. The answer to this is “yes”. It doesn’t matter whether the relationship is causal, and it doesn’t matter that IQ and SAT scores are imperfect measures of a person. What matters is that they are correlated.

To estimate IQ bases on SAT, you’d need a set of data with IQ and SAT score data on the same individuals. Given this data, you could generate a conversion formula, with associated error bounds. It wouldn’t surprise me if IQ and SAT test makers have this information. It also wouldn’t surprise me if they consider it proprietary, so you’d never be able to see it.

Aside: I actually had such data a long time ago, and maybe I still do. I took “the world’s hardest IQ Test” or some such thing in OMNI magazine back in the seventies (It was fun, get off my back). With my results came a sheet of data including correlations with other tests, including the SAT (IIRC). I was in high school, so I didn’t have a good understanding of what the data meant, or what i was supposed to do with it. There might be enough data to get a formula (if I still have the data) relating SAT to that IQ test.

I am not a psychometrician but here is a site where someone claims to be able to do the conversion.
This site is for pre-1994 SAT scores only.

Lets try that link again

This is absolutely true; a person’s IQ will contribute to their SAT score. However, this relationship is, correctly, called a spurious relationship. This simply means that there is some relationship between the two, i.e., they covary (sorry to add statistics terms to the mix). It does not convey anything about that relationship. Again, let’s go back to the height/weight relationship. These two variables covary (as one changes the other does) but they are not causally linked. In fact, they are both “caused” by an increase in age. That is, the age-height relationship is causal and the age-weight relationship is causal. Yet, the height-weight relationship is not causal. One does not cause the other, but they are both caused by the same thing.

This, while possible, will allow you to generate some sort of prediction equation for that group and that group alone. To try to apply that equation to a different group of students for whom you have only SAT data will give less than optimal results. This is because the “conversion formula” is a regression equation that is generated by using Least Squares methods. This means that the equation is generated so that error between predicted IQ and actual IQ is minimized for that group. Since it is optimal for that group, it is NOT optimal for another group (due to different variances). Using such a generated equation is quite dangerous. (actually, it’s using the results from such an equation that is dangerous)

puddle, I just checked out that site and it truly scared me. This is an example of why people think statistics are a load of crap - the unabashed charlatanism. This person does not explain how the generation was run, he/she just gives the equation. Note that the equation is linear. How does this person know that the relationship is linear? Given that the SAT has a score ceiling and that IQ does not, I would doubt that the relationship is linear. I would expect something more logrithmic. It is obvious that this person beat the relationship with the “linear wrench” until something came up. Where did the data come from? I doubt a randomized experimental design was used. Anything else would introduce so many threats to validity that any results would be useless.

I agree with Holden that the cite is pretty much worthless because no explanation is given on how the conversion was decided upon.

However, I disagree that IQ tests have no ceiling. Certainly IQ has no upper bound, but any test has a finite number of questions. If someone gets them all right, the testmakers can only conclude that he is smarter than they are.

Furthermore, IQ can not be reliably tested above a certain level (I think around 150, but that is just a WAG). After that, there is no test that can give anything but a qualitative answer about intelligence. It is therefore arguable that IQ has an upper limit of 150 (or whatever it is) just as SAT has an upper limit of 1600. Past that point, nobody can say for sure how smart you are.

IQ tests most definitely do have ceilings. The Stanford Binet LM has the highest ceilings when used with young kids. Once the kid reaches 8, the scores go down (and yeah I know some psychs explain that away by saying that the scores become definite at about the age of 8). The WPPSI-R and WISC III ceiling at about 140. They were designed to track learning disabilities not the extremes of intelligence as measured by an IQ test.

Rumour has it that the Woodcock Johnson which was released last year has a ceiling of 300! I find that very hard to believe as how on earth did they norm it? The highest measured IQ (that is known in the public domain) is Justin Chapman with a score of 298 at the age of 5. The Stanford Binet due out next year is also supposed to have a ceiling of 300.

I know that there is a reasonable correlation between kids testing in the profoundly gifted range and high SAT scores but I definitely don’t think you can use the info to do any extrapolation. If one test measures acquired knowledge and the other test measures IQ how can there be a relationship? A 7 yo with an IQ of 190 who has not been given the opportunity to be radically accelerated probably won’t do astoundingly well on the SAT. A 7 yo with an IQ of 190 who read at 2 and zoomed through school probably will do OK on the SAT.

Primaflora, I hope you don’t mind, but I’ve got a new sig line. Thanks a bunch!

Well, I highly doubt it. The SAT is primarily based on what you’ve learned. Take me for example:

I took the SAT in 7th grade and scored an 890. I took it as a junior and got 1340. Does that mean that my IQ went from 95 to 140 in 4 years? No…I just learned what they were talking about in those four years. IQ is supposed to not really change with age, but almost everyone improves if they take the SAT twice.


I’m in agreement with Jman. According to that web page, my IQ went up ~38 points in 5 years (7th grade to 12 grade). Not bloodly likely. I think I just did more prep work specifically for the test the second time, and I’d had a lot more schooling in the meantime. Also, not to brag, but I’ve always done fairly well on standardized tests. There are people who I think are smarter than I am, who didn’t score as well on the SAT. Maybe I just paid more attention in class? (heh) Or maybe I just don’t get flustered by #2 pencils…

There is a relationship between SAT scores and IQ scores, sure, but I’d say it’s casual*.

[sub]*That’s how I kept reading “causal” and it seems apropos.[/sub]