By Steve Sailer
12/19/2004
Right after my VDARE.COM article "Democrats Recoil From GOP’s Electoral Secret: Marriage Plus Children" came out last Sunday, VDARE.COM contributor Randall Burns pointed out to me an amazing online resource for anyone interested in differences between Red (Republican) and Blue (Democratic) states: LaboratoryoftheStates.com.
This database contains 377 different measures per state, from Alcohol Consumption per Capita to Welsh per Capita. And it will calculate the correlations among them.
At Randall’s suggestion, the editors of the website added the demographic measures — marriage and fertility — that I had showed correlated closely with Bush’s 2004 performance. This allowed me to see how they stacked up against all the other measures the website had accumulated.
I was enthralled to hear about this new toy, but a little worried. With that many different variables, surely there would be several that correlated with GOP performance better than the factor I had trumpeted in my last article.
As you may recall, I had announced that the average years married between ages 18 through 44 among white women correlated remarkably with Bush’s share of the vote.
When combined in a multiple regression model with the numbers of babies per white woman, the fit with the election results was astonishing.
Well, much to my relief, my Years Married measure came in first out of all 377 as the variable that best correlates with Bush’s performance by state, and by a wide margin.
The Laboratory of the States website came up with an incredible correlation coefficient of r = 0.95, even higher than the r = 0.91 correlation I reported on Sunday.
(The difference is that they took the logarithm of the Years Married, a standard statistical technique.)
Here’s their scatterplot. (You'll find Washington D.C. way down in the lower left corner, but right on the best fit line through the 50 states.)
I know this will make me sound like a total stat geek. (I guess I am, so I won’t try to hide it.) But to find a correlation coefficient of 0.95 between measures as distinctive as Bush’s Share and Years Married is absurdly exciting. Professional social scientists can go their whole careers without coming close to uncovering a correlation of 0.95 between nontrivial variables.
The other demographic factor that I have emphasized, as in my earlier American Conservative article "Baby Gap" — the average number of babies per white woman — came in third out of 377. By taking the log of it, they found it correlated at an r = 0.89. (Here’s their scatterplot.)
Now I should point out that one reason these correlation coefficients are so bogglingly high is that they include Washington D.C. along with the 50 states.
Statistically, D.C. is an extreme outlier (Bush won only 9% of the vote there) because it’s not a state, combining blue cities and red rural expanses, but a pure true-blue city.
Because D.C. is such an anomaly compared to the states, it is weighted very heavily in calculating the correlation coefficients. This leads to some strange findings.
For example, the percentage of immigrants from sub-Saharan Africa delivers the tenth strongest correlation with Bush’s share at r = -0.78. (More African immigrants, the worse Bush did.) That is mostly a product of heavily-weighted D.C. having the nation’s largest African immigrant community. But, even there, they still make up a relatively significant fraction of the population of the nation’s capital. It’s unlikely that the causal connection between having many African immigrants and voting against Bush is truly so strong.
So it’s also worth looking at the correlation rankings among just the 50 states excluding Washington D.C. This makes the percentage of African immigrants fall to a more plausible 84th out of 377 correlations, at -0.51.
In contrast, even without D.C. in the database, my two measures still come in first and third. That shows that my model of voting behavior is robust. Years Married leads by a wide margin, at r = 0.87, while Total Fertility is third at 0.81.
Remember, those are still almost stratospheric correlation coefficients. It’s common to describe correlations of 0.20 as "low," 0.40 as "medium," and 0.60 as "high."
What’s cause and what’s effect? The arrow of causality probably points in both directions.
Being conservative seems to increase Americans' desire for marriage and children. Conversely, being married and having children makes people more likely to be politically conservative, because they have more that’s worth conserving.
From the GOP’s perspective, it’s a virtuous circle.
Which makes it all the odder that the Bush Administration wants to open the borders, which would reduce wages and drive up housing prices, punishing those voters who would like to get married and start families if they can afford it.
Note that, tellingly, in second place as an indicator of GOP predilection, in between Years Married and Total Fertility, is
The negative sign means that the more housing prices have risen, the more Democratic the state was.
For example, housing prices in Massachusetts, the most Democratic state in the 2004 election, rose 516%, the highest home inflation in the country.
In Utah, the most Republican state, house prices were up only 162%.
Expensive housing retards family formation, which helps the Democrats. Rising housing prices transfer wealth from young people to old people.
Importing more foreigners, as the Bush Administration suicidally wants to do, drives up the price of homes by increasing demand.
That makes it harder for young voters to start down the road to homeownership, marriage, babies — and committed Republicanism.
[Steve Sailer is founder of the Human Biodiversity Institute and movie critic for The American Conservative. His website www.iSteve.blogspot.com features his daily blog.]
This is a content archive of VDARE.com, which Letitia James forced off of the Internet using lawfare.