Wednesday, January 14, 2009

Cognitive inequality is good for Democrats

Inspired by Steve Sailer's VDare article "Californication--Increasing Inequality--Surprisingly Good for Democrat Pols", one of the seminal posts here described how economic and educational parities are good for Republicans while economic and educational disparities are good for Democrats.

Bruce G. Charlton suggested cognitive parity might also be good for the GOP. From the limited correspondence I've had with him, I've found that when he suggests something, it's worth investigating.

The National Assessment of Educational Progress (NAEP) provides a way of doing so. The Department of Education's NCES site has recently been updated to allow for a wide range of statistical comparisons to be made among states and territories. In addition to providing average state level scores by subject, by year, and by race, the NAEP site can be queried for standard deviations on the same. The wider the standard deviation, the greater the cognitive disparity in the state.

Following is a table ranking the states by cognitive disparity as measured by state-level standard NAEP math and reading scores for 8th graders in '07, the latest year for which there are available data. Both tests are on a 500 point scale. The nationwide standard deviation was 35 points for reading and 36 points for math. The table reports an average standard deviation value from the '07 tests:

Connecticut, Nevada37
Hawaii, Rhode Island36.5
Alabama, Arizona36
DC, Florida, Michigan, North Carolina, Washington35.5
Alaska, Maryland, New Jersey, New York, Oregon,
Utah, Wisconsin
Arkansas, Illinois, Pennsylvania, South Carolina34.5
Colorado, Minnesota, Missouri, Nebraska, West

Georgia, Indiana, Iowa, Massachusetts,
New Mexico, Tennessee

Idaho, Kentucky, Mississippi, Texas, Virginia32.5
Delaware, Oklahoma, Vermont32
Kansas, New Hampshire, Wisconsin31.5
Louisiana, Maine, Montana, South Dakota31
North Dakota28

The size of a state's standard deviation on its NAEP average scores and it's level of support for McCain is an inverse .36 (p=.01). A state's SD and it's level of support for Bush in '04 is a similar .37 (p=.01)*.

The wider a state's intelligence distribution is, the bluer that state tends to be. In addition to inequalities in educational attainment and wealth, inequalities in IQ are also good for the Democratic party. So the party putatively determined to reduce inequality benefits from doing just the opposite, while the party putatively more tolerant of cutthroat competition and the winners and losers it produces benefits not from economic wins and losses but from draws. This seems to provide some justification for the general sense that red states (especially outside the South) tend to be less chaotic and intrusive places to live, but also tend not to be on the cutting edge of technological or cultural trends, either.

The variances in score distributions is not the result of blue states being more intelligent than red states are, as might be anticipated. There is no relationship between a state's average IQ and voting patterns in Presidential elections**.

The 10.5 point range for the states is not huge, but that does not mean it is not important. California's standard deviation is more than one-third larger than North Dakota's is. Presuming for simplicity that both states' average IQ is 100 and that a 15 IQ point standard deviation is the national standard deviation of 35 for reading, a person with an IQ of 112 is in the 84th percentile in North Dakota, while a person must have an IQ of 117 to be in the 84th percentile in California. Conversely, an IQ of 88 would put a person in the 16th percentile in North Dakota, but it would require an IQ of 83 to be similarly placed in California. California would thus have relatively more of both smarties and dummies running around than the more middling North Dakota would (in reality, as North Dakota's average IQ is, presuming a standard intelligence distribution, half a standard deviation higher than California's is, so in reality California probably has relatively more uber-smarties, dummies, and uber-dummies than North Dakota does).

After realizing the above, I presumed the relationships would be stronger when only white students and white voters were considered, but they are nearly the same as when the entire population is considered. In fact, they're a little weaker. The correlation between NAEP SDs by state and support for McCain is .29 (p=.04). For NAEP SDs and support for Bush, it's .37 (p=.01). The potential strength of these relationships appears to be weakened by the fact that white state-level variances in the South and Midwest are similar, but white support for the Republican candidate is firmer in the former than the latter.

Via Swivel, the data are here.

* The state's aggregated standard deviation used in comparison with the '08 election is computed by finding the state SDs for both the '07 reading and '07 math tests and averaging them. The same is done using reading and math tests from '03 for comparison with the level of Bush support in the '04 Presidential election. It was my intention to use writing and science assessments to create a 'deeper' aggregated SD average, but neither assessment receives full participation by all 50 states and DC. But being able to utilize assessments in math and reading allows for reasonable confidence that both verbal and non-verbal aspects of IQ are being measured.

Parenthetically, this is presumably one reason VCU's Professor McDaniels' used NAEP scores for math and reading to comprise his average IQ estimates by state. Previous estimates were made here that correlated almost perfectly with his, although I used math and science scores. Consequently, for a couple of states I had to rely on math scores alone.

** Given the South's support for the GOP, this might come as a surprise. Remember, though, that as overwhelming as the white vote in the South is for the Republican candidate, significant black support for the Democratic candidate means that the strongest levels of total GOP support tend to come not from Southern states but from states in the heartland, like Oklahoma, Wyoming, and Utah (the three most Republican states in '08). Further, several of the most reliably Democratic states are not the sharpest of the blue. Among the top ten blue states in '08--DC (#51), Hawaii (#48), Vermont (#3), Rhode Island (#37), Massachusetts (#1), New York (#30), Maryland (#34), Illinois (#31), Delaware (#23), and California (#46)--only two are in the upper echelons of average state IQ.


The Undiscovered Jew said...

"Among the top ten blue states in '08--DC (#51), Hawaii (#48), Vermont (#3), Rhode Island (#37), Massachusetts (#1), New York (#30), Maryland (#34), Illinois (#31), Delaware (#23), and California (#46)--only two are in the upper echelons of average state IQ."

American politics is still driven to an enormous extent by intra-white political jockeying inspite of the media's obsession with the always predictable minority vote (They vote Democrat, you stupid reporters!).

Understanding white voters by ethnic, educational, and income background and other factors is going to be more relevant to the future of America than the nonwhite vote is for quite a long time because whites are still so culturally and politically dominant not only in the US but nearly all of Latin America.

Perhaps you should think about breaking down white voters by religion and ethnicity? As useful as IQ is, it doesn't tell us the complete story about voting patterns of American whites because white IQ doesn't vary much from state to state as in the same way it does from race to race.

However cultural variation from region to region is quite striking.

California has been losing tens of thousands of white middle and upper middle class Republican voters since the mid 90's. The whites remaining in California are much higher in the income and educational pool because only those types of whites can increasingly afford to stay in California. So not only is California importing low IQ immigrants, they are losing middle and upper middle income whites to more middle class friendly states like Colorado, Nevada, Oregon, etc.

The Northeast and MidAtlantic states have a large population of white ethnic Catholic voters, socially liberal SWPL WASPs, and, obviously, a smaller, but financially very wealthy and influential Jewish population.

Jews of course are strongly Democrat. Although not as Democrat as Jews, and though they claim to be strongly anti-abortion, white ethnic Catholics in the Northeast, and Mid-Atlantic are quite favorably dispossed to the Democrats compared even to other Catholics in the Midwest, to say nothing of white Mormons, Evangelicals and various Protestant groups.

Understanding these internal white voting dynamics is harder, but far more important, than another MSM article recycling "The Historic Latino Vote" article.

Audacious Epigone said...


Right. To the extent that the non-white vote is interesting, it mostly has to with turnout--support will overwhelmingly be given to Democrats. I'd like to create an electoral map as it would look if only white votes were counted, and also if only non-white votes were counted. The former would still be competitive (and probably present a more accurate graphical representation of the blue/red divide than state totals do), but the latter would be completely blue.

Re: breaking white voters down by religion and ethnicity, it's tough to do. The best state exit polling offers is "white Protestant" and "white Catholic". There are no data at all on ethnicity. And the GSS only presents data by Census designated areas, of which there are nine.

Steve Sailer said...


It would be interesting to get more of a sense of what's going on with California's white students. They score below average for whites, but they have the 3rd highest standard deviation of white only test scores behind only Hawaii and Nevada.

So, whites in California's public schools don't do that well, but some do very well. I wonder why that it is?

In LA public schools, white kids at the middle and public school levels tend to be either kids whose parents finagled them into an exclusive program or whose single moms just let them get dumped into the general population, which is conducive to a high standard deviation. It's not uncommon for private schools in LA to have a lot of white kids who aren't smart enough to get into the elite public school academies.

But, I imagine that the bulk of white students live in their own suburban / exurban school districts.

Audacious Epigone said...


My presumption re: the wide white California SD is the disparate nature of northern coastal California, especially the Marin county area on one hand, and southern and interior California on the other hand. But I'm a clueless middle earther without credentials to really speculate on the coasts!

The middling white students in LA are presumably in religiously-oriented schools? Or are they there simply because they have responsible parents who do not them to suffer the LA USD?

The Undiscovered Jew said...


For the year 2008 there was a 3.6% decrease in the absolute number of births in Arizona compared to 2007. The fall in births was strongest during the second half of 2008.

The bulk of the decline appears to have happened in Maricopa county and to a lesser extent Pima county.

This is the first absolute fall year over year in Arizona births since at least 1996:

Neuroskeptic said...

Is the standard deviation in a state's IQ correlated with the state's mean IQ? That seems plausible. In which case could mean IQ explain the Democratic advantage in variable states?

Audacious Epigone said...


Wow, so the trend appears to be accelerating. From January to August, Arizonan births decreased by 2.9%, while the most populous county, Maricopa, showed a 4.3% year-over-year drop (for the first seven months of '08 vs. the first seven months of '07). This is very encouraging.


I'm away now, but just eyeballing the table, I'd guess if there is a relationship, it's pretty weak, and probably trends towards less variability = higher average scores (California, with the highest SD, is one of the NAEP's worst performers, while North Dakota, with the lowest SD, is one of the best).

Audacious Epigone said...


For both math and reading, the relationship between the size of a state's SD and its average score are inversely related. Smarter states have smaller SDs.

In math, the correlation is weak, like I'd presumed it would be (r=.17, p=.22)--too modest to put much confidence in. But the relationship with reading is strong (r=.66, p=0). I'm not sure exactly why this would be. Maybe the functionally illiterate really score in the dumps, while the difference between the functionally literate and the precocious readers isn't as wide. So duller states, having more illiterates, also have relatively large score variations. Any thoughts?