Doctor Intelligence
[ Part of a sequence of posts on intelligence. ]
Background
The MCAT, LSAT, GRE, SAT, ACT, and AP tests are all curved using a process called "equating". Unfortunately, that term is rather vague, but the intention is to make it so scores are comparable over time even if the difficulty of the questions and the ability of the test takers change.
The sources I've found for the specific tests section are almost all either (a) official but only mention these claims in passing or vaguely or (b) unofficial and discuss these claims in depth. That being said, the claim seems pretty solid for AP tests Different Versions of Exams.
On this webpage, I will accept this equating process as accurate, with one exception: distortions due to the Flynn effect Flynn effect.
Physicians
1945 - 1970
Table 1 in The intellectual caliber of medical students. gives the following results
- Between 1946 and 1958, the average IQ of med school matriculants increased 2 points (SE=1.1) on the same test.
- Between 1960 and 1962, the average IQ increased 0 points (SE=0.8) on the same test
- Between 1961 and 1965, the average IQ increased 1 point (SE=1.4) on the same test.
- Between 1960 and 1967, the average IQ increased 1 point (SE=0.5) on the same test.
A fixed-effects meta-analysis of these four observations suggests an average increase of 0.15 points per year (SE=0.06). This is a tad lower than the 0.3 points per year you'd expect from the Flynn effect Flynn effect, which suggests a mild retardation between the end of WWII and the late 1960s - to the tune of around 1.5 points per decade. The authors don't reach this conclusion, because the Flynn effect wasn't well-known at the time.
They also look at MCAT scores, noting
In two unpublished studies at Oregon, the MCAT has been found to correlate with the WAIS Full Scale IQ about .60 to .75, even in the restricted range of talent represented by a sample of medical students.
it is important to understand that MCAT scores from one year to the next are directly comparable; that is, in scoring the MCATs taken each year by the new applicants, the mean score of the new group by the use of established statistical procedures is made equivalent to that of the original 1951 standardization group. Thus, all scores in Table 2 are based on the same 1951 mean of 500.
They found that between 1951 and 1971 the average subtest score had increased from 500 to 571 - this corresponds to a gain of about 4.3 IQ points. That's 0.22 points per year, a bit below the 0.3 points per year predicted by the Flynn effect Flynn effect, again indicating a mild regression of about 0.8 points per decade.
1970 - 1990
During the early 1980s, Lupo et al gave a recently renormed Wechsler IQ test to 101 students admitted to the Creighton University School of Medicine in the early 1980s. They found a mean of 126 (SE=0.5) Lupo.
1990 - Present
The average MCAT score of matriculants increased from 29.6 to 30.8 (d~0.041 per year) from 2002 to 2009 Table 23. The average increase from 2018 to 2021 was d~0.036 per year A-16. In other words, it looks like MCAT scores of matriculants have been increasing by about d~0.04 per year for the last two decades.
Likewise, Step 1 scores of students also increased, from 200 in 1994 to 232 in 2019. The standard deviation is about 20 points USMLE, so that's an effect size of about d~0.064 per year.
The Flynn effect has essentially been gone since the late 1980s Flynn effect, so these scores presumably reflect a real increase in ability. Step 1 is largely a knowledge test, but the MCAT has significant IQ portions and (at least historically) correlates quite well with IQ tests. All this suggests the average matriculant IQ has risen by roughly 10 points in the last two decades.
I have been unable to find any studies examining the association between MCAT or Step score with IQ, so I think that's the best we'll get here.
Summary
The average IQ of doctors probably shrank about 1 point per decade from the end of WWII through the early 70s, bottoming out in the mid 120s during the early 1980s. It has probably increased ~5 points per decade since the early 90s. This suggests it's probably in the mid 130s now.
It actually makes good historical sense that average IQ of physicians shrank prior to the 1980s, since the 1970s saw a near-doubling in medical school graduatesGordon:
This only realistically makes sense if the they became less selective. It's less obvious why the average IQ has increased since the mid 1980s: both number of medical students and number of 22-year-olds have remained largely flat - though one can't help but suspect that the ~60% increase in college enrollment may play a factor.
Finally, average MCAT score (and therefore, presumably IQ) varies quite a bit by specialty Table B1, ranging from 27.4 (Family Medicine) to 32.5 (Otolaryngology), representing a 1.22SD range.
PhDs
During the mid 20th century, the typical IQ of PhD students was in the upper 120s,
A keynote presentation claims the average IQ of most PhD recipients is around 130 overall, and around 140 for PhD's in physics or those who graduate Phi Beta Kappa (the top few percent of each graduating class) Genius, creativity, and talent: Definitions and manifestations.
Finally, in his book Coming Apart, Charles Murray claims the average IQ of white attainers of "PhD, LLD, MD, DDS" is 124, citing the NLSY-97 sample - a decrease of 2 points from the NLSY-79 sample Murray. When I ran my own analysis including other races on the same dataset, I found a median IQ of 118 for people who obtained a PhD.
So, exempting the non-cited keynote presentation's claim, it looks like most of the numbers suggest the average PhD attainer has an IQ in the low-to-mid 120s and that this number has been fairly consistent over time. The number plausibly reaches the 130s for those who specialize in physics or who graduate Phi Beta Kappa.
One final note, the average IQ of PhD students in Denmark seems noticeably lower: probably the low 110s Akcigit.
JDs
Table 1 from Henderson suggests the average LSAT of law school students slightly decreased between 1992 and 2003. However, this hides heterogeneity: table 3 shows that the average LSAT score slightly increased at the top half of law schools, and increased by 1.7 points at the top 16 schools - the equivalent of around 4 IQ points
I can't find any studies linking the LSAT to IQ, but Mensa and the Triple Nine Society typically accept scores of 167 and 173, respectively - suggesting those scores correlate with IQs of 132 and 146, respectively. The median LSAT score at a top-8 school is 171 2020 Raw Data Law School Rankings, which suggests an average IQ of about 141.
From what I've read, the LSAT is generally accepted as more IQ-test-like than the SAT - with some people claiming it functionally is an IQ test. However, it is also specifically being selected for by Law Schools, which means using it to estimate IQ is probably biased upwards.
In any case, the methodology of Mensa and the Triple Nine Society doesn't appear to include any analysis of the correlation between the LSAT and g, and I suspect the correlation between the LSAT and a IQ test is probably about r~0.85, so I'll reduce that 141 estimate by 15% to 135.
My verdict? The average member of a top-8 law school probably has an IQ in the 130s.
Elite Undergraduates
Alright, these people technically aren't doctors, but I think they're an interesting point of comparison. I've found two relevant studies:
- A 2003 study administered an IQ test to 168 Harvard students and found a mean score of 126 Carson. However, as noted here How well does the SAT correlate with official IQ tests? the IQ test was quite old, which means it was biased by the Flynn effect. The blogger suggests the true average score is 121 - my own calculations put it at about 123.
- A (non-random?) sample of Dartmouth students from 1970 to 1971 had an average IQ of 130 Culver. Again, our friendly blogger corrects this downwards due to the Flynn effect More evidence that Ivy League students average IQ 122. Again, I disagree with his methods. By my math, the Flynn effect adjusts the scores downwards by 3.4 points - down to 127.
Finally, the average ACT score of Harvard students is about 34. Based on the preferred model from last post, this suggests an average IQ of 132. Note, however, that this average ACT score has increased by about 1 point since 2003, which mean the model predicts an average IQ of 130 when the Harvard study above was conducted. The story is similar if we look at SAT scores: the estimated IQ today is 131 and the estimated IQ in ~2003 is about 129.
In all, then, it looks like the average IQ of elite undergraduate students is in the mid-to-upper 120s - dramatically lower than the 145 claimed by Jordan Peterson Strengthen the Individual: Q & A Parts I & II - he doesn't cite this claim, but I suspect he arrived at this number simply by looking by converting SAT scores to IQ scores via a naive percentile table.
The one piece of cited evidence that "mid 120s" is too low is Charles Murray's The Bell Curve, where the author claims (again citing NLSY data) that the average IQ of the "top dozen universities" was 141 in 1990. However, he estimates IQ scores based on the children's SAT scores, which is improper, because those scores are used to choose who attends each college, which makes them statistically biased. Still, after adjusting downwards (assuming r~0.9), it still yields an average IQ of 137.
TODO: repeat Murray's analysis, but use the NLSY's ASVAB or AFQT test scores instead. Also: figure out how on Earth he achieved such a high estimate.
Top Software Engineers
I couldn't find any solid evidence on the average IQ of software engineers at top-paying companies. The closest was Nolan Eoghan claims Eoghan:
...I once did an IQ test in one of the FAANGS, it was organised by the workers (engineers and software folk). Participation of a few hundred.
The average was 130 but it was tightly bunched. Just a few low 150s, as in one or two. I think IQ tests aren’t very useful at higher levels anyway, and useless at estimating creativity or actual genius etc.
Taken literally, this suggests a standard deviation of about 7-8 points.
Everything else I found was either single-person anecdotes or pure speculation.
Overall
It looks like the average MD, physics PhD student, or top-8-school JD student all have IQs in the mid 130s. The average PhD student or elite undergraduate student, meanwhile, is probably in the mid 120s.
There are about 37000 law students each year (2500 at the top-8 law schools). There are about 19000 MDs each year. There are about 55,000 PhDs (about 2000 in physics, 12000 in science and engineering).
Combining the above two paragraphs suggests the three most selective groups number about 23,500 per year and that about half of them are from the top 1% of the population. There are about 4 million 20-year-olds in the US, which suggests about 30% of the top-1%-by-IQ people end up going down one of those three avenues.
I feel it is worth mentioning that the 75th percentile of STEM majors at elite universities probably also have an average IQ of around 135. Ditto for the 86th percentile of non-STEM majors. So, the average IQs of the following groups are probably all around the mid 130s:
- MDs
- Top-8 JDs
- Physics PhDs
- Phi Beta Kappa PhDs
- Top 50% of STEM majors at elite colleges
- Top 30% of non-STEM majors at elite colleges
A quick note on pay:
- The average doctor makes around $300k Wilcox. If we discount for ~8 years of training, that's about $183k. If you account for the fact most doctors spend about 4 years before attending medical school padding their resumes, this shrinks to $143k.
- The median pay of new graduates from Top-8 law schools is around $190k Caron.
If we discount for three years of law school, that's about $145k. - In industry, STEM PHD graduates generally start at ~$127k PhD Salaries and Lifetime Earnings.
If we assume 5 years of training, that's a discounted value of around $92k.