Tuesday, August 12, 2008

Regional intelligence estimates; Wordsum appears a reliable IQ proxy

If the GSS reported data by individual state, it'd offer another valuable data source for getting at average IQ at the state level in the US.

That isn't the case, but data by region does allow GSS results to be compared with other good-faith IQ estimates. Do they corroborate or contradict GSS WORDSUM scores?

A table follows listing the average IQ by region as arrived at through multiple approaches, with population adjustments made and the national average IQ adjusted to 98 for ease of comparison*. The Wordsum scores are converted using a standard deviation of 15. Additionally, estimates based on '05 eighth grade NAEP science and reading scores, averages from the half-century old Project Talent study of 366,000 9th-12th grade students, averages from 38 year-old Vietnam veterans obtained in '85/'86, and Tickle.com's (which offers a free putative IQ test that isn't timed) state averages are presented:

RegionGSS NAEPProj TalViet VetsTickleAverage
New England101.6100.4101.3103.298.7101.0
Mid Atlantic99.798.699.9100.998.499.5
E-N Central97.799.099.5100.297.898.8
W-N Central99.799.9100.4101.998.3100.1
South Atlantic96.097.194.592.697.795.6
E-S Central92.496.393.490.896.693.9
W-S Central96.096.897.492.897.296.1
Mountain100.197.698.699.598.698.9
Pacific100.395.998.6102.398.699.1

States included in each regional category, with the region's national rank and average IQ as suggested from the five methods presented in the table in parentheses:

New England: Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island, Vermont (1st, 101.0)
Middle Atlantic: New Jersey, New York, Pennsylvania (3rd, 99.5)
East North Central: Illinois, Indiana, Michigan, Ohio, Wisconsin (6th, 98.8)
West North Central: Kansas, Iowa, Minnesota, Missouri, Nebraska, North Dakota, South Dakota (2nd, 100.1)
South Atlantic: Delaware, DC, Florida, Georgia, Maryland, North Carolina, South Carolina, Virginia, West Virginia (8th, 95.6)
East South Central: Alabama, Kentucky, Mississippi, Tennessee (9th, 93.9)
West South Central: Arkansas, Louisiana, Oklahoma, Texas (7th, 96.1)
Mountain: Arizona, Colorado, Idaho, Montana, Nevada, New Mexico, Utah, Wyoming (5th, 98.9)
Pacific: Alaska, California, Hawaii, Oregon, Washington (4th, 99.1)

The nine geographic regions correspond to US Census definitions. It's not exactly how I'd slice up the country, especially the Mountain region, which groups the aptitudinally disparate states of Montana and New Mexico together. I'd prefer a distinct Southwest region covering Nevada, Arizona, New Mexico, and the southern half of California. But I suppose these regions are not comprised with HBD realists in mind. Damn time zone and postal code considerations! Anyhow, click on the graphic to see it more clearly.

Returning to the question posed, they corroborate quite impressively. Regional numbers from the GSS correlate with Project Talent, Vietnam Vets, and Tickle in the .90-.98 range, meaning the rankings by region are ordered pretty similarly to one another. If you had lurking concerns that the simple Wordsum test was too rough a measure of intelligence to be so readily proxied for IQ by GSS miners Half Sigma and Inductivist, you should put them to rest**.

The only major exception comes from the NAEP-derived estimates in the Mountain and Pacific regions, which are lower than the other sources suggest. The NAEP numbers correlate in the .50-.70 range with the others when these two are included. When they are removed and only the remaining seven regions are considered, the relationships become strong, in the .90-.97 range (the GSS and NAEP do so at .92).

My speculation is that the NAEP numbers, gleaned from the test results of 8th graders just a few years ago augur the future more than they reflect the present. The Southwest (captured by the Mountain and Pacific regions) has undergone extensive racial/ethnic demographic changes in the last couple of decades unparalleled by any other area of the country. The average age of those surveyed in the GSS is 44, the Vietnam vets were middle-aged in the eighties, and the striplings who provided data for Project Talent are now comparing various retirement communities.

Tickle's numbers are based on internet users who decide to take the free test. Pew reports that Hispanics are less likely to use the internet than whites or blacks are. This results entirely from Spanish-only speakers:

78% of Latinos who are English-dominant and 76% of bilingual Latinos use the internet, compared with 32% of Spanish-dominant Hispanic adults.

So the newest arrivals must be significantly underrepresented.

I wonder also if the Tickle numbers don't offer some insight into the relative size of a state's "smart fraction". DC is at the very top despite in all likelihood having a lower average IQ than any state in the country, probably because one of its intelligence distributions is doing most of the testing. Homogenous places like the Dakotas, meanwhile, come in nearer the bottom. If it is primarily from the right side of the bell curve that Tickle test takers come (and if the test has value, that's tautologically true, as the national average is over 112), the northern half of California could be inflating the state's average.

* In the case of Tickle.com's averages, a significant adjustment is clearly necessary. Without any such adjustments, the national average is 112.6. As flattering as that might be, we don't want to give our friends in the Middle East any more reason to think that the US is a Jewish behemoth than they already have! Without adjustments, the Project Talent data suggests a national average 99.8 and the Vietnam veterans numbers suggest an even 100. With more than one-quarter of the population either black or Hispanic, these seem too high, as 100 is supposed to indicate the white mean. Presuming that 100 is the national average, however, is easy--just add 2 points to all the numbers for all the regions in the preceding table.

** Such concerns were the impetus of this post!

4 comments:

John S. Bolton said...

Now with several measures are you in a position to predict the effect of the entire country going through the same population changeover as California has gone through in the last 40-odd years, over the next same interval of years?

Audacious Epigone said...

JSB,

I'm not sure how precise it would be. But the NAEP scores show the Southwest down 2-3 points relative to the rest of the country and also relative to the other good-faith measures. So a generation out, I'd hazard a national average of 95 or 96 from the contemporary Lynn and Vanhanen estimate of 98. I can't find it at the moment, but I recall Steve Sailer writing a VDare column where he predicted an average IQ of 96 in the US by mid-Century, though I may be wrong. Stated crudely, that would make the intellecutal future US as a whole on par with Arizona or Nevada today.

Peter said...

Stated crudely, that would make the intellecutal future US as a whole on par with Arizona or Nevada today.

Which isn't necessarily dire. Intellectual obstacles notwithstanding, Arizona and Nevada have maintained excellent rates of economic growth. I would be more concerned if the national IQ became more like that of the economically stagnant Rust Belt - even if it were a few points higher.

Audacious Epigone said...

Peter,

Right, not dire. That was the conclusion Steve came to. But the irritating underclass trends will be accentuated and popular culture will become less intelligent as we take a few steps towards the world depicted in Idiocracy. Of course, disparities in economic growth rates will remain. If the real value of the dollar kicks into secular depression, Ohio's futures relative to those of Nevada will improve. The consequences of changes in average IQ will be longer lasting than bull and bear markets in various industries.