I find this Wall Street Journal piece fascinating. Charles Murray’s thesis is that (A) American education can’t be markedly improved and (B) IQ sets an upper threshold for what students can understand.
The second part is pretty much just plain true. The first part, however, was shocking to me — after all, Murray himself has argued that school vouchers could drastically improve education, but now he’s saying nothing can. And we always hear about how US students placed such-and-such in a ranking of so many countries on this or that skill.
I set out to test the idea, because Murray provides little in the way of data. The first thing I found is that, while the Flynn Effect was making IQ rise, US academic test scores dropped in the two decades starting in the mid-60s. Part of this was due to more kids taking the tests, but the government also found these things made a difference:
“[D]ecreases in the quantity of schooling which students experience, curriculum changes, declines in student motivation, and deterioration of the family system and social environment.”
So, IQ isn’t everything. Better education can improve test scores — after all, if no one teaches you the information that’s on the test, all the IQ in the world won’t pick the answers out of thin air. How well is the US using its IQ assets?
First of all, things have improved since the 20-year problem the ’60s caused. A 2000 report found:
“Standardized achievement tests attained record high levels in the mid- to late 1980â€™s and remain there.
“Scores on the National Assessment of Educational Progress have risen to all-time highs. Gains have been especially dramatic for blacks and Hispanics.
“The proportion of students scoring above 650 on the SAT mathematics section attained record levels around 1995 and has remained at the all-time high. This cannot be accounted for by Asian-American students who are too few in number, constituting some 9% of all SAT test-takers. Of the 75% increase between 1981 and 1995, black, white, Hispanic and Native Americans accounted for 57%.
“The number of students taking Advanced Placement examinations his risen from just over 1000 in 1961 to over 1,000,000 currently.
“American students are second in the world in reading.”
Great news, but just because the US improved since its fall doesn’t mean it’s measuring up. Using a list of countries’ IQs from IQ and the Wealth of Nations (data here and here), along with a list of reading and math PISA scores for OECD countries’ students, I ran some numbers.
The data is on Google Spreadsheets here. I excluded countries for which there was no IQ data.
The first thing that struck me was the correlation between test scores and IQ, r=.73 for reading and .82 for math. Clearly, there’s a link there.
But what I found more fascinating was that the correlation between math test scores and reading test scores, r=.94, was even higher. This is odd, because IQ includes both verbal and spatial components — it should predict the two tests better than they predict each other.
There are a variety of possible reasons for this, including the fact that the reading and math numbers were collected by the same organization, while the IQ numbers are estimated from a wide range of studies. But I think there’s something else going on, education. Education prepares people for both reading and math tests. If a country has a high average IQ, its students will do better. But if they augment that high IQ with good education, the students will truly stand out. The reading and math tests pick up on that effort, while the IQ tests are designed not to (IQ doesn’t change after about age 5, before most of the educating takes place).
So the money question: How does the US measure up? Our IQ is 98 (normed to the UK at 100), our reading score is 495.18 and our math score is 482.89. The countries I looked at had averages of 97.45, 482.97 and 486.59 respectively. The standard deviations are 4.99, 39.98 and 52.08.
So, the US has a slightly above-average IQ for this sample of OECD countries. We’re .11 standard deviations above the mean. We’ve done well with it in reading, where our kids score 12.21 points (.31 standard deviations) above the average. (We’re far from second in the world, though, in the PISA data.) Where we’re not doing so well is in math, where our above-average IQ gets us below average (albeit slightly, .07 standard deviations) results.
Murray himself admits that some schools, particularly in the inner cities, are terrible. Improving these schools, no doubt, could boost performance. Where he’s shockingly right, however, is in that US schools are about where they should be, given the country’s IQ.
Hat tip Steve Sailer.
I reviewed Murray’s latest book here.
UPDATE: I’ve re-uploaded the spreadsheet and changed the link. I added what I call the IQ utilization score — calculated by averaging a country’s math and reading performances in deviation from the mean. Then, I subtracted the country’s IQ advantage, also in standard deviations. Essentially, it’s a way of asking, given a country’s intellectual abilities, how much is the culture and school system getting?
Finland, Canada and Ireland fared quite well. Italy and Uruguay, not so much. The US had a very slightly positive score, meaning we’re outperforming our IQ just a little.
UPDATE II: I found an interesting thing about my IUS measure. While two of the worst-performing countries, Italy and Uruguay, aren’t high-IQ states, there is a correlation of -.344 between IQ and IUS. In theory there should be no tie — by and large, if a country is X standard deviations above the mean, its students should be too.
There are two reasons I can think of. One is that countries simply aren’t getting the mileage they should out of smart kids, even in countries with lots of smart kids.
The other is the Asian phenomenon. As an ethnicity, Asians do very well on math tests, but not so much on language ones. It’s possible the IQ scores pick up on the math, but the performance measure picks up more reading. This would make high-IQ Asian states have low IUS’s.
UPDATE III: I found some slight corroboration to the Asian thesis. I calculated separate math and reading IUDs, and the math one correlated higher with IQ (r=-.37) than the reading one did (-.29). However, the reading one still correlates quite highly with IQ, so that’s not the whole story. For some reason, countries with high IQ don’t get as high of test scores as they should. Some sort of regression-to-the-mean-like phenomenon?
I suppose it would have been ideal to compare the reading scores to the verbal portion of an IQ test, but I don’t have that data.
UPDATE IV: I wanted to figure out where, exactly, the IUS-IQ correlation was coming from, so I divided the countries into the top half and bottom half of IQ. Those in the top half had an r of -.73 (!); the bottom, POSITIVE .31. It seems that the further above average your country’s IQ gets, the less good it does in increasing school performance. Diminishing returns, if you will. Meanwhile, if your country has a below-average IQ, the closer it gets to average, the more good it does.
In other words, increasing average IQ from 98 to 99 does more good than increasing it from 92 from 93 — but it also does more good than increasing it from 105 to 106.