I decided to use a linear regression analysis to convert LSAT scores to IQ’s. I found five data points from five high-IQ societies that accept LSAT scores. The International Society for Philosophical Inquiry accepts a 178 LSAT score, which corresponds to a 146 IQ. The Poetic Genius Society accepts a 174 LSAT score, which corresponds to a 139 IQ. Intertel accepts a 172 LSAT Score, which corresponds to a 135 IQ; and Mensa America accepts a LSAT score that is in the 95th percentile, which is historically a 167 LSAT Score corresponding to a 130 IQ. I’m assuming a 15 SD for all IQ scores. The LSAT average is historically between 150-151. I’ll choose 151 to be generous. Inductivist calculated that the average college graduate’s IQ is a 105. So the last data point is a 151 LSAT score corresponding to a 105 IQ.
My slope is 1.4891, which I’ll round to 1.49, and my y-value is -119.7679, which I’ll round to -119.77. The resulting equation is IQ = LSAT*1.49 – 119.77. Here is an incomplete chart; the decimal points don’t convey greater accuracy:
180 146+
179 146+
178 146
177 143.8
176 142.3
175 140.8
174 139
173 137.8
172 135
171 134.8
170 133.4
169 132
168 130.4
167 130
166 127.4
165 126
164 124.4
163 123
162 121.5
161 120
160 118.5
GreySwan
27 March 2009
You conversions are simply wrong! They are at least 5 points to low, and maybe more. From my own data-set — my brother has a Verbal IQ in the mid 150′s and got at 172. Your median is also at ~105 IQ, which is far to low, since the average test taker must have an IQ closer to 115.
Lover of Wisdom
28 March 2009
Hi GreySwan:
I address your worry in Steve Sailer’s post.
Franklin Carroll
7 April 2009
Your analysis is wrong. The average high school graduate has an IQ of 105—the average college graduate has an IQ of 115. I think this throws your whole analysis off somewhat.
Lover of Wisdom
7 April 2009
Franklin:
I doubt it. the average college graduate had an IQ of 115 in the 1960′s, but not any longer. Let me quote Inductivist’s post at length:
“Mean IQ of college grads dropped 9 points from the 60s to the 90s: The General Social Survey has collected data on a simple measure of IQ since 1974. I calculated means for whites across various degrees and decades. (Black college grads appear to have IQs three or four points lower than whites, but sample sizes are very small, so I didn’t include them). Most of the action is observed with a 4-year degree, so I’ll focus on it:
Mean IQ–white college grads
1960s 113.72
1970s 110.59
1980s 108.04
1990s 104.42
2000s 105.12
The trend is clear with a slight reversal in this decade. I looked at people who graduated during the 1950s and before, but I didn’t list the scores since they are similar to the 60s cohort.
America is obsessed with educating everyone, regardless of ability, so admission standards have been lowered. A bachelors degree has not held its value over the years.
Fortunately, the mean score for those with advanced degrees has not fallen, but the mindless compulsion to educate more and more people will likely fix that. In my experience, the only thing keeping IQs up is that colleges cannot convince people below a certain IQ threshold to stay in school.”
I placed the link above, but I’ll post it here too: http://inductivist.blogspot.com/2008/06/mean-iq-of-college-grads-dropped-9.html
prospero
11 April 2009
A simple linear regression based on correlations reported from other agencies will not yield meaningful results. How did Mensa, ISPI, PGS and Intertel achieve their equivalences, and what is the confidence intervals for their analyses? You report no correlation coefficient for your regression line. Another defect in this analysis is the populations you use. The other agencies will arrive at their equivalences with or without correcting for a likely strong selection bias – people who enter their societies self-select, and are not of the general population of LSAT takers. Your data set is from LSAT takers, but if the LSAT and Stanford Binet or Wexhsler were administered to a random sample of people, not taking it to go to law school, then you would be able to fit a curve, probably, to the data and be able to predict, within a certain range of error, IQ given LSAT scores. As it is, the LSAT takers self-select and cannot be considered the same population as those who know their own IQ.
If you look around at the equivalences online, for example the GRE or SAT, while there may be a general correlation, belief that one has a specific IQ given a specific test score is not useful because of the inherent statistical variation within each test, multiplied by comparing two tests, and bias in the populations that take each test. IQ comes from the general population. LSAT from a subset of that population. It is an interesting question, and it would certainly be doable, to administer multiple tests to the population at large, but the sample size would have to be huge, since what you are looking for is the equivalence of outliers – those above 95% in the population – 1 in 20 or less. To get meaningful data, you would have to administer several day long tests to a random sample of the population of at least 600 people. Good luck with that, also given that these are expensive commercial tests. Maybe a quarter million dollars to do it well? It would be interesting though. What you have done here really is statistically meaningless.
david
19 April 2009
“I found five data points… Inductivist calculated that the average college graduate’s IQ is a 105. So the last data point is a 151 LSAT score corresponding to a 105 IQ.”
“IQ comes from the general population. LSAT from a subset of that population. It is an interesting question…”
http://foreigndispatches.typepad.com/dispatches/2004/10/an_interesting__2.html
LSAT Score = (0.048)*(SAT Score) + 100
SAT1600 = LSAT177
http://www.geocities.com/mobiusnu/satlsat.html
LSAT Score = (0.0333)*(SAT Score) + 121.01
SAT1600 = LSAT174
http://www.top-law-schools.com/forums/viewtopic.php?f=2&t=276
Lover of Wisdom
19 April 2009
Thanks for those links, David. How did you find them?
Anonymous
25 April 2009
Your choice of 105 for the mean LSAT score is somewhat questionable. Even if 105 is a good estimate of the mean college grad’s IQ, it’s extremely likely that the test taking population is drawn predominately from the top 1/2 or 1/3 of college graduates. Aside form being a matter of common sense, this is evidenced by LSAC score reports, which report mean GPAs of test takers from each institution. Unsurprisingly, mean GPAs tend to be (substantially) higher than the actual mean GPA at the institution. For example, at my school, the mean GPA of those who took the LSAT was 3.3 while the mean GPA at the institution as a whole was ~3.0.
That said, the significance of this error decreases as you go up the scale and perhaps doesn’t even matter once you get to mid to high 170s .
Dooder
25 April 2009
Your choice of 105 for the mean LSAT score is somewhat questionable. Even if 105 is a good estimate of the mean college grad’s IQ, it’s extremely likely that the test taking population is drawn predominately from the top 1/2 or 1/3 of college graduates. Aside form being a matter of common sense, this is evidenced by LSAC score reports, which report mean GPAs of test takers from each institution. Unsurprisingly, mean GPAs tend to be (substantially) higher than the actual mean GPA at the institution. For example, at my school, the mean GPA of those who took the LSAT was 3.3 while the mean GPA at the institution as a whole was ~3.0.
That said, the significance of this error decreases as you go up the scale and perhaps doesn’t even matter once you get to mid to high 170s .
Larry Raymond
2 May 2009
I agree that your calculation is in error, although I also think that I have found the problem. The top LSAT score of 180 on your chart corresponds to a 146 IQ, just as it does on other comparison charts that I have seen. However, your mistake lies at the bottom of your chart, in your assumption that the average LSAT taker (score: 151), corresponds to the average college graduate (IQ: 105). The LSAT, like any graduate or professional program entrance requirement, has a self-limiting factor. People who take the LSAT are skewed to being more intelligent than the general population, or even the average college graduate. The IQ at the bottom of your chart is too low, and therefore your downward slope is too steep. I haven’t bothered to look for a source re: the average LSAT taker, but I am instinctively certain that anyone who bothers to look it up will find that I am correct.
GreySwan
16 May 2009
I still think your numbers are off, across the spectrum. I believe that ~ 20-30 people get a 180 score every year, out of 120,000 test takers. That is ~1/5000. If this were a random population that would put you at about a 154 IQ with S.D. of 15. However, its not random, so I guess your pushing closer to 160 IQ for a 180. So
I also doubt that the curve is linear.
Rufus
27 June 2009
In the past 4 decades, I have taken an IQ test (LONG ago), the SATs, two sets of GREs, and LSATs. The IQ tests and converted SATs and the GREs all put my IQ within a 5 point range. BUT this LSAT to IQ conversion sets my IQ at 20 to 25 points lower than the other four tests did! I would be inclined to remove this from the internet as the logic has so many flaws and the data points are so few.
Ben
2 July 2009
I don’t know what methodology The International Society for Philosophical Inquiry used to conclude that a score of 178 corresponds to the 3 sigma level in the general population; 178 is already at the 3 sigma level in the population of LSAT takers, so I’d wager that ISPI’s IQ estimate is a pretty significant underestimation if they were just trying to do an equipercentile correspondence between LSAT scores and IQs. However, it might be that ISPI was trying to account for regression to the mean in their estimate, who knows.
Commodore
21 July 2009
I also agree that this conversion chart badly underestimates IQ. If memory serves me correctly, the average LSAT score for Harvard graduates is ~161. Assuming that the typical Harvard LSAT-taker is as at least as bright as the average Harvard graduate, this chart would suggest that the average Harvard graduate has an IQ of 120, barely more than one SD above average. This cannot possibly be correct given the average SAT score of Harvard matriculants.
Lover of Wisdom
21 July 2009
HLS entering class averages around 171-172, which is about 135 IQ.
notthatsmart
10 September 2009
This is absolute hogwash
I have taken online IQ tests that put me no higher than a 125 and I earned a 177 on October 2008 LSAT.
Jamie
29 May 2010
I read that the average LSAT for MIT students was 164. And, that means that some scored lower than that.
By the way, I got this information off of MIT’s website about 1 1/2 ago.
Jamie
19 November 2010
Actually, the average LSAT for MIT students last year was 162.4
Lewis Percher
29 November 2010
Mensa requires a 163 for admission. With my 165 I was well within the top 2 percentile range. I believe the data you’re operating with is incorrect.
Lover of Wisdom
29 November 2010
Do you mean LSAT score? Their website (at the top) requires a 95th percentile (of the testing population) or better on the LSAT, which is a 166-167 LSAT score. A 163 LSAT score is clearly below the 95th percentile, unless they very recently re-normed the test, which I haven’t heard of.
Lewis Percher
30 December 2010
Lewis Percher:
Mensa requires a 163 for admission. With my 165 I was well within the top 2 percentile range. I believe the data you’re operating with is incorrect.
Lover of Wisdom:
Do you mean LSAT score? Their website (at the top) requires a 95th percentile (of the testing population) or better on the LSAT, which is a 166-167 LSAT score. A 163 LSAT score is clearly below the 95th percentile, unless they very recently re-normed the test, which I haven’t heard of.
They accepted me in with my 163, so I suppose test must have been re-normed after all.
In any case, the LSAT is bunk. I took it three times and scored 68th percentile, 57th, then 92nd. No reasonable test of intelligence should produce such a ridiculously wide range of results for the same person.
Lewis Percher
30 December 2010
I should mention that I’m in Canada. Maybe the whole system works differently here.
Lover of Wisdom
30 December 2010
Lewis:
I just checked Canadian Mensa. They do indeed accept a 163 LSAT score, but that is different from admission requirements for American Mensa, who accept scores at the 95th percentile, which is always normed as a 166-167 LSAT score. Perhaps they just do things differently in Canada (even though the admission scores on the other tests, e.g., GRE, where spot on, I think).
Will
21 February 2011
Your analysis is simply wrong. You can’t start with the average college graduate. The reason you cannot begin with the average college graduate assumption is that the LSAT is not a test taken by every college graduate such as the SAT is for high school students. The LSAT is only for students considering law school, and given the general rigors of getting to the point of sitting for the test, most of the those taking the test are sincere in their desire to go onto law school. What does this mean? It means that the average LSAT taker is already well beyond the average college graduate. The average LSAT taker has done well in college and has likely gone to a more difficult college. Your starting assumption is so flawed that I’m curious why you would even bother? Simply put, you are trying to take an elite level test and bend it to average standards. There is your fault and it takes little academic training or proficiency to grasp. Disclaimer: I got a 163 on the LSAT, 1390 on the SAT (old style) and my IQ was tested at 134.
Ken Huston
10 March 2011
My LSAT – 172 — 98.7 percentile
IQ – 151
Your conversion is too low!
Nietzsche
21 March 2011
The operative word is “average” guys, just because you are a special outlier doesn’t mean anything to the overall data and conclusion.
For an IQ of 120+ you all seem to lack basic knowledge of statistics.
israeliteknight
15 April 2012
Major SAT Math Score 2006 US Statistical Abstract, Table 674, Money Income of Households, 2003
Math 607 $62,000
Asian Men 575 $65,555
Physics 574 $50,000
Engineering 553 $49,800
White Males 542 $53,705
Asian Women 541 $45,843
White Females 507 $37,556
Average 505 $43,318
Hispanic Males 489 $38,836
Biology 480 $36,300
Computer & Info Tech 479 $46,300
Humanities/writers 478 $44,000
Economics 465 $60,000
Sociology 464 $35,000
Hispanic Females 449 $27,158
Education 446 $37,900
Black Males 431 $34,901
Black Females 416 $24,406