Will family formation determine the 2008 election?

Via Matt Yglesias, I found SurveyUSA's tables of Presidential polling in all 50 states (but not DC), as of early March. In each state, 600 voters were surveyed on McCain vs. Obama and McCain vs. Clinton match-ups.

They show McCain losing narrowly in the Electoral College against either Democrat. But that's not what I'm interested in. I want to know whether family formation among non-Hispanic whites will paint the electoral map red or blue again. The answer appears to be: yes, although not quite as much as in 2000 and 2004.

Here are the correlation coefficients (not the r-squareds) for recent two-party races (leaving out the 1992 and 1996 elections that were perturbed by Perot), and leaving out Washington D.C. (an outlier that typically falls beautifully on the best fit line):

Correlation Coefficients	1988 Bush	2000 Bush	2004 Bush	2008 McCain-Obama	2008 McCain-Clinton
Years Married Whites 2000	0.54	0.84	0.86	0.71	0.68
Total Fertility Whites 2002	0.59	0.79	0.80	0.56	0.76

Overall, family formation appears more somewhat more important if Hillary is in the race than if Obama is. I assume that's because Hillary is more of a known quantity, while Obama remains more of a blank slate upon which people are invited to fill in their fantasies.

I would assume that if 2008 elections were held today, the actual 2008 correlations would be higher because they'd be based on the universe of voters, not on samples of 600 per state, which injects random errors into the 2008 numbers, thus lowering correlations. On the other hand, as I've said before, the odds are that the November 2008 correlations will be lower than 2000/2004, both because they were so high in 2000 and 2004 that regression toward the mean will likely kick in; and because those two races featured fairly generic Republican and Democratic candidates, while only Hillary at present looks like a standard representative of her party. Also, the correlations would be higher if SurveyUSA had surveyed Washington D.C. -- it helps drive up the correlations to stratospheric levels because, being, in effect, a city-state, it's an outlier that falls right on the best fit lines).

Both McCain, who considered switching parties early in the decade, and (at least at present, Obama) are more sui generis than Bush and Gore/Kerry. On the other hand, Obama has been running so far as a bipartisan centrist. Eventually, I would assume, people will figure out where he's really coming from, so a McCain-Obama race would likely end up more like 2000/2004 than it looks like now.

It's not exactly clear what, besides decent judicial appointments, the Republicans are doing to merit the support of family-oriented voters and how long the can keep harvesting these votes without doing much in return.

By the way, I get a lot of knee-jerk criticism for correlating demographic statistics just for whites with election results summing all races. But, when you stop and think about it, that makes my findings even more unexpected and interesting. (I offered some explanations for why there's a better fit between voting by state with white family formation rates than with total family formation rates in the American Conservative in 2004.)

These correlations above would be higher if Washington D.C.

And this isn't just post hoc data mining on my part. The 2004 results confirmed a theory I had started to outline even before the 2000 election. I wrote about the connection between Total Fertility Rates and conservatism/liberalism in the case of two mostly white state -- Utah and Vermont -- in VDARE in June of 2000, before the 2000 election. And on 11/22/2000, I pointed out on UPI that Bush had beaten Gore in the 19 states with the highest white total fertility rate.

Methodology: As you'll recall, the second statistic, Total Fertility Rate, is a well-established measurement for estimating the number of babies a woman would have between ages 15 and 44 based on birthrates by age in a particular year. Unfortunately, the Census Bureau hasn't published total fertility rates by ethnicity by state for any year since 2002, so this statistic is starting to get a little dusty.

The first statistics, Years Married 18-44 is one I invented, modeled upon TFR, to denote the average number of years a woman can expect to be married between ages 18 and 44 based on rates of being married in a particular year. I only have it for the 2000 Census, so it's even more out of date.

My published articles are archived at iSteve.com -- Steve Sailer