I think that Charles Murray has hold of the wrong end of the stick in his op-ed in the Wall Street Journal, “The Roots of STEM Excellence”. Much of the op-ed is devoted to identifying young people capable of learning science, technology, and mathematics early by standardized testing and admitting them to top tier institutions of higher learning based on that. Here’s a snippet:
In the 1970s, Johns Hopkins psychologist Julian Stanley established the Study of Mathematically Precocious Youth by administering the SAT to 12- and 13-year-olds. Some 2,000 of the participants have been followed throughout their careers.
Measures of productivity varied substantially within the top percentile, equivalent to an IQ of 135 or higher. Those in the top quartile of the top 1%, equivalent to IQs of 142 and higher, were more than twice as likely to earn a doctorate or be awarded a patent as those in the bottom quartile and more than four times as likely to publish an article on a STEM topic in a refereed journal. There was no plateau. Greater measured cognitive ability was correlated with greater adult accomplishment throughout the range.
These results suggest that we should be thinking in terms of at least the top half of the top percentile of ability when defining the set of people who have the potential to make major contributions in a STEM field. The U.S. has around 130 million people of prime working age: 25 to 54. For any given talent, therefore, about 650,000 are in the top half of the top percentile of ability. That’s a lot of people.
The task is to identify those with STEM talent when they are young. The good news is that standardized tests expressly designed to measure cognitive ability are an efficient way to do so. They are accurate, inexpensive, resistant to coaching and demonstrably unbiased against minorities, women or the poor. Those conclusions about the best cognitive tests are among the most exhaustively examined and replicated findings in all social science.
The bad news is that admissions offices of elite universities ignore this evidence.
I’ll present my reactions in the form of a number of bullet points followed by an anecdote or two.
- IQ testing does measure something real
- There is a correlation between the cognitive ability represented by IQ and the ability to learn science and mathematics
- There used to be a good correlation between SAT scores and IQ. That correlation no longer appears quite as strong.
- There are abilities other than IQ which are more closely correlated with success (however measured) in life than IQ
- There is a weak correlation between income and IQ
- Compensation, like most other prices, is determined based on supply and demand
- The rate at which jobs for physicists is increasing is less than the rate at which the population is, even the population of people with the cognitive abilities to be good physicists, and that has been true for 50 years
- Lawyers and physicians tend to have IQs above normal but by less than three standard deviations
- There aren’t a lot of jobs in mathematics and the sciences that pay enough to lead a middle class lifestyle
The smartest guy I know (he has an IQ four standard deviations above normal) is a brilliant mathematician. He has a doctorate in mathematics and an IQ four standard deviations above normal. After trying for years he finally gave up getting a tenure track position teaching in a university and became a computer programmer. One of the smartest guys I know is a successful creative writer. Before that he held minimum wage jobs. I don’t think these anecdotes are out of the ordinary. Extremely smart people tend not to be lawyers, physicians, or in finance. They find those positions too boring or require interpersonal skills other than those they have.
There are several things we could do to encourage “excellence in STEM” but we aren’t doing them. One of the most important is ensuring that the people capable of excelling at STEM earn enough to make it worth their while.







