Sunday, October 22, 2006

Average IQ by State: Honest Numbers at Last

October 22, 2006

Average IQ by State: Honest Numbers at Last

By Steve Sailer

One of the most popular Web postings of November 2004 was a table purporting to show that John Kerry had swept the 16 states with the highest average IQs (such as Connecticut with its 113 mean). George W. Bush, in contrast, had carried the 26 dumbest states (such as Utah at only 87).

The first person to post these data after the election exulted: "Wow, what can I say, in the first 24 hours over 540,000+ people viewed this page!"

I would guesstimate that total viewership of the IQ table ultimately approached ten million fellow Democrats —all consoling themselves with the thought that what they lacked in quantity of voters, they more than made up for in quality of brainpower.

But it was all a complete hoax. And I had already pointed it out the previous May. Then, The Economist magazine fell for an earlier Bush v. Gore 2000 version, which it later retracted.

Still, the question of average state IQ scores is one that millions find fascinating, so I'm glad that the scientific journal Intelligence has in press an article entitled Estimating state IQ: Measurement challenges and preliminary correlates [PDF file] by Michael A. McDaniel, a widely published professor at the Virginia Commonwealth University school of business. (Thanks to Dienekes's blog for the tip. We publish this new state IQ table here. The earlier hoax data is in the right hand column.)

How did McDaniel come up with these estimates? He used the federal government's National Assessment of Educational Progress math and reading test scores. That's a technique that I introduced in 2004 at the suggestion of Ken Hirsch.

In the second column of the new table are Dr. McDaniel's reasonable-sounding estimates of IQ, setting the national average on the NAEP tests at 100 and the standard deviation at 15.

McDaniel has added a refinement to my approach. There was this problem: NAEP scores for 4th and 8th graders are just from public school students. But in states with large black or Hispanic populations, there is substantial white flight from the public schools , lowering NAEP scores. For example, 13.7% of South Carolina's white students and 10.6% of California's are not in public schools. McDaniel adjusts for this, raising slightly the IQ estimates of states with heavy white flight. (State IQ was defined as "the average of mean reading and mean math scores.")

The relationship between state average IQ and partisanship turns out to be quite mixed. (In the new table, I’ve marked states that voted Republican in 2004 as red, while Democratic states are blue.) Comparing McDaniel's estimated IQ scores to Bush's share of the vote in 2004, I see virtually no correlation: just -0.12.

In other words, if you were told the state IQs, you'd be only a little more than one percent (-0.12 squared) of the way toward fully predicting the state-by-state results of the 2004 election. (In contrast, the correlation coefficient for the hoax IQ data with Bush's percentages by state was -0.85, or 72% toward a wholly accurate prediction.)

How do McDaniel's results compare to other good faith state-by-state IQ estimates?

A massively representative national IQ test last happened in 1960, when Sputnik scared the federal government into giving an IQ test to 366,000 high school students as part of Project Talent. Four decades later, McDaniel's NAEP-based IQ estimate still correlated at the high 0.63 level with Project Talent's old results.

In a mid-1980s study of Vietnam veterans, 4,321 took an IQ test, with results tabulated by their state of birth. The findings correlate 0.59 with McDaniel's. (Data from Project Talent and the Vietnam veterans can be found here.)

The Social Quotient website estimated IQs from SAT and ACT college entrance exams scores. The results have a correlation coefficient of 0.71 with McDaniel's numbers.

Finally, for whatever it's worth, the free Internet-based Tickle IQ test has published its averages by state. Tickle correlates 0.53 with McDaniel's numbers.

Examining McDaniel's data, you should first notice that even the largest difference between states’ mean IQ—the 10.2 points between Massachusetts and Mississippi—is actually not at all enormous. And that’s especially true compared to the differences among countries. There, gaps of 30 points or more are not uncommon (e.g. Austria vs. Congo-Zaire ).

The median Massachusetts student would score at the 61% percentile nationwide, and the median Mississippi student at the 35% percentile. This isn't a tremendous difference—but it would be noticeable if you moved from one state to another.

McDaniel notes:

"States with higher estimated state IQ have greater gross state product [per capita], citizens with better health, more effective state governments, and less violent crime."

You'll also observe from the map below (where the brighter states are indicated by the brighter colors) that the highest IQ states are extremely, shall we say, Northern. Six of the top eight (marked in yellow) border on Canada . Overall, there's a clear correlation between latitude and NAEP scores.



No comments: