In a characteristically perspicacious Taki’s Magazine article, Steve Sailer writes:
Among respondents who graduated from college in the 1960s, the average score on the vocabulary test, expressed on an IQ scale, was 112.3, almost a standard deviation above average. For each decade since then, the average vocabulary IQ has dropped steadily down to 100.0 (two points above the mean) for those who graduated in the 2010s.
Among those who didn’t graduate from college, the mean score has dropped from 97.3 for those who left school in the 1960s to 89.3 in the current decade.
You might be somewhat surprised by the falling vocabulary test scores because raw IQ scores in the 20th century tended to go up. But the Flynn effect was seen less on culture-loaded subtests such as vocabulary and more on subtests that resemble programming your smartphone.
– For the cohort having attended college in the 2010s, 100.0 may be more than 2 points above the population mean. The total sample is based on an assumption of a population average IQ of 98, but the total population is whiter than the under-30 population is. At less than 60% (non-Hispanic) white, nearly 25% Hispanic, a bit under 15% black, and 5% Asian, the average IQ of the current college-age American population is probably closer to 96 than 98.
– Average Wordsum scores for the total population have remained remarkably consistent over time (6.11 in 1978 to 6.13 in 2016 with a mean over the life of the survey of 6.10 and a standard deviation of 2.01–in other words the population average has moved an imperceptible .01 standard deviations!). Wordsum scores for whites have very slightly increased from the 1970s to the present (from 6.26 in 1978 to 6.34 today), in line with the modest increase the Flynn effect has had on vocabulary scores.
That average scores for both those who attend college and those who don’t have fallen over the same time range without the total population average changing is a result of an increasing percentage of the total population going to college today compared to six decades ago.
As college attendance effectively becomes less restrictive, the average IQ at each level of educational attainment will decline without any necessary change to the total population average IQ. Educational romanticists would have us believe that education increases intelligence, but as the Wordsum–among other measures–shows, that’s not the case.
For simplicity, assume that the total population IQ ranges from 50 on the low end to 150 on the high end, a point for each percentile increase. In the 1960s, the top 10% of the general population (mean IQ 145) goes to college while the bottom 90% (mean IQ 95) does not. By the 2010s, the top 40% of the general population (mean IQ 130) goes to college while the bottom 60% (mean IQ 80) does not. In both the 1960s and the 2010s, the population mean IQ is 100 even though the mean IQ of both those who’ve gone to college and those who have not has declined from the 1960s to the 2010s.
Parenthetically, this results anytime the cream of the crap is taken from the broader crap bucket and dumped into the broader crop bucket–like, say, with immigration of the “talented tenth” from third-world countries to the West, a phenomenon that lowers the average IQ of both the sending third-world countries and the receiving Western countries without (sans differential birthrates!) changing the average global IQ.
– A perceptive commenter also pointed to Google’s Ngram viewer program, which tracks the relative frequency of words in books published, by year. Here are the Wordsum words in books published from 1950 through 2008 (the word “space” is shown at just 5% of its total frequency because if included in the results without an adjustment it dwarfs all the others and renders the graph incomprehensible–my thanks to candide3 for pointing out that Ngrams allows words to be rescaled):
“Pact” spiked in the 1940s–ie, the Nazi-Soviet pact–and was in the process of settling back to its historical norm when we pick it up here. Excepting that and “allusion”, there hasn’t been much change in frequency of the words included in the test over the last several decades.
The commenter’s insinuation is that the more (or less) frequently these words appear in books–and by extension the culture generally–over time, the better (or worse) performance on the Wordsum should be during the corresponding period of time.
That’s intuitive, but it may not actually be helpful in understanding the dynamics of the test.
Jason Malloy reported on correct response rates by word. “Allusion”–the word that, after “space”, has consistently appeared more frequently in books than any of the other words have–has the lowest correct response rate of all. That is, it’s the one test takers have the most trouble with even though it shows up in books more than do other words people struggle less with.
Similarly, though “space” is orders of magnitude more common than any of the other words, test takers miss it more than they miss “broaden” or “edible”.
GSS variables used: WORDSUM, BORN(1), YEAR, RACE(1), EDUC(12)(16-20)