Month: January 2015

GRE/MAT and IQ Correlation

LeCarrefour de la Sagesse had an entry a few years back purporting to convert GRE verbal scores to IQ, which I was delighted to find, though perhaps a bit skeptical of from the outset. Granted, I’m no mathematical whiz myself, so my initial reservations were based not so much on methodology as on the feeling that my own score correlated with an IQ that was likely too high . . . much to my chagrin. That’s not some kind of false humility, mind you, but an honest assessment, based on what I know about myself and my past achievements. In fact, I once took the Mensa home test to see how I would do; and from what I was told by admissions, the estimated IQ range given in the results is fairly accurate, which means that indeed the correlation most likely was slightly off, even if by only a few points. To satisfy my own curiosity, however, I decided to plot a linear regression based on some of the numbers I could find, particularly the correlation between the Miller Analogies Test, which is accepted by various high-IQ societies, and the GRE verbal section of 2012-2013, which is unfortunately no longer accepted by any high-IQ societies at all. Assuming that these two tests do match up well enough, and assuming that the Miller Analogies Test is a fairly accurate assessment of IQ, it would seem reasonable that one should be able to take both scores and translate them into a rough IQ estimate that might be more accurate than some of the other attempts that are out there.

If we begin by looking at the MAT scores accepted by various organizations, we find that Mensa accepts those in the 95th percentile and above, which comes out to be a 437, which further translates into a WAIS number of 130. On the higher end, the Triple-Nine Society accepts a scaled MAT score of 472 and a WAIS score of 146, leading us once again to infer a rough correlation between the two as before, while the International Society for Philosophical Inquiry correlates an MAT score of 471 with a 146 on WAIS — pretty much the same as Triple-Nine for all intents and purposes. Intertel sits between Mensa and these other groups, taking a raw score of 74, which comes to a scaled score of 452 and a WAIS IQ of 135.

So far, so good; but we would do well to find some numbers lower down on the scale. This has proven to be rather difficult since I’m not confident enough in much of the data that I’ve found to actually use it for this project, though, surprisingly, librarians appear to be a useful group for this purpose as both their Wonderlic and WORDSUM scores bring them to the higher end of average at 114 which seems intuitively reasonable for someone in that occupation, while we can easily access their average GRE scores through the ETS, thus providing us with the necessary data. At a 157, we can take this number and convert it into the equivalent MAT score of 415 and infer that this 415 loosely coincides with an IQ of 114 finally giving us something that isn’t restricted to the top 2% of the population.

What would be wonderful is finding out the overall average IQ of the GRE test-taking population, though the best we can really do is guess since I don’t think that kind of information exists. Accepting, then, that the average IQ of a college graduate is roughly 105 and assuming that those who attempt the GRE found college tolerable enough that they were willing to potentially pursue more education, we could take a stab and say that the average test-taker has an IQ of about 110, meaning that the 50th percentile on the GRE verbal test would correlate with this number, which once again we can turn into an MAT score, this time 399.

With the help of I was able to construct a linear regression giving a very basic visual of what this all looks like.

Screen shot 2015-01-18 at 12.07.05 AM

On the left are the IQ scores, while on the bottom are the MAT scores, which can be converted into GRE scores using the link that I provided above. (And so as not to forget those who took the test before the most recent revision) For example, then, if you were an English major and scored a 169 on the verbal section, you could look at the chart, find that it coincides with a 449-450, and then consult the graph to see that your IQ falls somewhere within the 132-135 range. Unfortunately there’s not a whole lot of detail here, but given the uncertainties involved, I don’t think I would recommend accepting total precision from these numbers anyway, only the broader generalities. I do, however, think that this regression is more accurate than the one I mentioned earlier, because remember I said that I had taken the Mensa home test? Well, I won’t tell you the estimated IQ range they provided me, but I will tell you that these estimates come much closer.