"Higher educational attainment definitely leads to voting Democratic."That appears to have barely been the case in the 2000 Presidential election but wasn't quite true in 2004. The electorate is much more evenly split educationally than a state-level comparison might be presumed to suggest. I took for granted the assumption that looking at a state as an individual entity gives different results than looking at individuals within a state.
That's simply not true. State level analysis always gets this backwards. Why use such a crude proxy that lumps millions of people together, when you have the GSS?
As you step through each additional level of educational attainment the proportion of Republicans increases UNTIL you get to graduate degrees.
Overall Republicans have more years of formal education than Democrats.
Maybe I shouldn't have. It is a point that needs to be known. The blackest states (Mississippi, Louisiana, Alabama, South Carolina, and Georgia comprise the five darkest) consistently vote Republican, even though blacks overwhelmingly vote Democratic*. More obviously, the wealthier a person is, the more likely he is to vote Republican, but the wealthier a state is, the more likely it is to vote Democratic**. A less pronounced but similar phenomenon is seen with educational attainment.
Why these 'paradoxes'? It's personal speculation, but in part because the increased presence of the less educated, less intelligent, and less white (or Asian) in Mississippi means voting for policies that benefit them at your own expense is less of an abstract moral exercise and more of a real world sacrifice than it is in say, Vermont, where leftist quasi-socialism doesn't have anywhere near the same pathological consequences as it would down South.
If the same educational attainment index used in the previous post (taking the percentage of the population that has a bachelor's-plus, subtracting the high school dropout percentage from it, and multiplying by 100) is employed for voters in the 2000 and 2004 Presidential elections, Bush gets a 36.6 in '00 and a 36.3 in '04. Gore gets a 34.2 in '00 and Kerry gets a 38.5 in '04. The educational attainment of the candidate's respective electorates is almost indistinguishable.
Recall also that Colorado topped the list of states with a 19.6. The voting electorate is substantially better educated than the country at large is. While nearly 20% of the public lacks a high school education, only 4% of those who voted in the '04 Presidential election did. While about 25% of the public has a bachelor's degree or better, 42% of those who voted in '04 did. The middling bloc, with a high school diploma but no bachelor's degree, represented 54% of the '04 vote, in line with the 56% of the population it represents.
* Presumably when blacks approach comprising half of a state's population, this tendency will abruptly evaporate (think DC), but the closest blacks come to comprising the majority of any single state is Mississippi, at 37.1%. There is no statistically significant relationship between how black a state is and its propensity to vote for Bush in '04 if all states are included, although the weak correlation is positive, because so many states have a negligible black population (lily-white North Dakota voting heavily for Bush and lily-white Vermont voting strongly for Kerry attenuates the black effect). If only states where blacks comprise 10% or more of the population are included, however, there is a correlation of .42 (p=.07)--the blacker the state, the more likely it is to vote Republican.
This trend was clearly visible in the Democratic Presidential primaries--the fewer blacks a state had, the better Obama performed among white voters. In states like the Carolinas, Georgia, Alabama, and Mississippi, where blacks comprised a majority of voters, he did poorly among whites, but it didn't matter as the black vote alone carried him. It was in states like Tennessee, New Jersey, and Michigan, where the electorate included more than a token black population that nonetheless was clearly in the minority where Obama fared most poorly.
** The GDP per capita by state correlates inversely with Bush's share of the vote at a strong .70 (p=0). That does not include any attempt to adjust for cost of living differences, which would weaken the relationship, but it clearly exists in a real sense.