Marriage Gap Located

With regard to yesterday's question about whether the "marriage gap" stands up to more sophisticated statistical scrutiny, Steve Sailer recommends his article "Values Voters" in the current American Conservative which, in turn, cites "Unmarried Women in the 2004 Presidential Election" by Anna Greenberg and Jennifer Berktold which finds:

The marriage gap is a defining dynamic in today’s politics, eclipsing the gender gap, with marital status a significant predictor of the vote, independent of the effects of age, race, income, education or gender. Marital status had a significant effect on the way in which these voters performed, whereas a voter’s gender did not. This was true of all age groups. Younger unmarried women supported Kerry while younger married women supported President Bush. Unmarried 18- 29 year olds gave Kerry a 25 point margin, while younger married women, like their older counterparts, gave President Bush an 11 point margin.

As you can see over on the left, this suggests that the hypothesis that unmarrieds vote Democratic because the unmarried population skews younger and young people support Democrats is wrong. On the contrary, being 18-29 seemed to have a weak independent correlation with voting for Bush while being unmarried had a strong correlation with voting for Kerry. Thus, that hypothesis actually has it backwards and suggests that Kerry had an advantage with the 18-29 crowd in large part because it contains so many single people.

It does seem worth saying, however, that "unmarried" is an awful heterogenous situation. Of course, any binary categorization is going to produce diverse groups of people in both categories. But in particular treating divorced people and never-married people as part of the same category seems like the kind of thing that could easily create misleading results.

Comments

Close-href tags are bwoke.

Posted by Jack Roy | February 21, 2008 1:35 PM

being 18-29 seemed to have a weak independent correlation with voting for Bush

Actually it seems to be as close to uncorrelated as you could possibly be.

Posted by right | February 21, 2008 1:36 PM

Countdown to apoplectic comments censuring Matt for referring to Steve Sailer's article in 3, 2, 1...

Posted by Fred | February 21, 2008 1:36 PM

I'd also like to add that my single status does not necessarily correlate with the fact that I'm incredibly lonely. But it certainly doesn't help.
Thank you and gods bless.

xoxo,
CraigGreg

Posted by Craig Greg | February 21, 2008 1:38 PM

Of course it makes sense that young, married people would go for Bush. Generally, those who are quick to get married younger (e.g. if you are the type who is "waiting until marriage", you best get married young as who wants to wait for so long) are more conservative I would imagine.

Posted by DAS | February 21, 2008 1:38 PM

I think we broke their site, can't access it at the moment. This is pretty surprising to me. Is the effect as pronounced among young singles as middle-aged singles? I would think that the reason for their single-ness would be a pretty big factor in voting patterns.

Imagine two 18 year olds. One wants to get married, but just hasn't found someone yet. The other is opposed to the institution of marriage in general. Seems to me the first one would be more likely to vote for Bush than the second. Then fast forward to 30 years old. The first one ends up in the "married" camp, having found someone. The second one is still unmarried.

Posted by Tel | February 21, 2008 1:41 PM

I don't like the looks of this regression (in my old company, this would be a scorecard). The Party ID factor is huge determinant in that thing, so in a sense, the factors are those factors that are relatively *uncorrelated* with party affiliation on whether one voted for Kerry.

Further, there is this huge negative "constant" term.
One could created an equivalent scorecard, but add 2.588 to the constant term, and subtract the same from all racial groups. This would imply that blacks are big republicans, which is really nonsense.

A better thing would be to look at the difference *within* age groups. Here, it would show that middle-aged fogies are Bushies, whereas the old and the young or Kerryites-- with the proviso that you got this honking big PartyID factor that you still have to disentangle.

It all goes to show you-- math is hard.

Posted by Dean Chung | February 21, 2008 1:42 PM

Thus, that hypothesis actually has it backwards and suggests that Kerry had an advantage with the 18-29 crowd in large part because it contains so many single people.

This is right, but I have one minor statistical criticism (along the lines of right's comment): note that the Age 18-29 factor has no asterisks after it, implying the adjusted p-value was > 0.05; the OR listed is insignificant at the usual Alpha cutoff of 0.05, basically at the noise level.

So it is probably more correct to say that after adjusting for the other factors in that list, being a member of the 18-29 age group did not have a statistically significant effect on one's presidential vote.

Posted by Ben | February 21, 2008 1:42 PM

It may simply come down to regional or urban/rural dynamics. Young urban professionals delay marriage until an above average age, for instance. Religion also plays a role - the Christian, no sex before marriage crowd tends to marry younger than the rest of us.

Posted by TH | February 21, 2008 1:43 PM

There is also the correlation between "unmarried" and homosexuality to consider.

Posted by Jerome | February 21, 2008 1:46 PM

Can someone explain the variables B and Odds factor here? I am having trouble with the units of B. What does it mean that party ID gets a 6?

Posted by mpowell | February 21, 2008 1:47 PM

Actually, I may assumed what these factors mean, and they may not mean what I think they do-- like mpowell, so what does B mean?

Posted by Dean Chung | February 21, 2008 1:49 PM

Posted by Dean Chung:
"A better thing would be to look at the difference *within* age groups. Here, it would show that middle-aged fogies are Bushies, whereas the old and the young or Kerryites-- with the proviso that you got this honking big PartyID factor that you still have to disentangle.

It all goes to show you-- math is hard."

An excellent point. Since marraige should be strongly correlated with age, there's a conflict.

As for the meaning of the coefficients:
http://en.wikipedia.org/wiki/Logistic_regression

Posted by Barry | February 21, 2008 2:01 PM

Dean: the B are almost certainly coefficients for logistic regression - log odds ratios. Similarly, the negative constant term is a product of logistic regression - it is _not_ arbitrarily chosen; rather, it represents the log odds of voting Democratic if all your covariates (party ID, black, etc.) are set to zero; here it appears to be the log odds of voting democrat for white, non-gun-owning, non-churchgoing, non-hispanic (etc.) men age 50-64.

The big problem here is collinearity. Suppose that in real life, being black is associated with being unmarried, and also with voting Democratic, but that within racial groups there's no marriagevoting association. In principle, multiple regression should be able to tease the marriage (non-)association out from the race association. In practice, the model may assign some of the race variation to the marriage category. In this fairly complex model, it's reasonable to think that collinearity might be a problem - a variance inflation factor (VIF) analysis would be helpful.

I agree that throwing party ID in the model mucks everything up. In effect, our model asks "how do unmarried self-identified democrats(/republicans) vote differently from married self-identified democrats(/republicans)?" I'd like to see a similar model with party ID taken out to see how the coefficients change.

Posted by Allan | February 21, 2008 2:10 PM

With my slow typing, I'm sure someone will beat me to this, but in response to Dean:

Further, there is this huge negative "constant" term. One could created an equivalent scorecard, but add 2.588 to the constant term, and subtract the same from all racial groups. This would imply that blacks are big republicans, which is really nonsense.

This is a multiple logistic regression model, which in its simplest approach always contains an intercept (or constant). The constant reflects the odds of voting for Kerry for a group which consists of the opposite of all the groups in the model.

So in this model, the constant reflects the odds of voting for Kerry among (in order): Male, White, non-gun totin', Married, Non-unionized, Non-Hispanic, ill-educated, non-Democrat, non-Church attending, upper income Americans aged 50-64.

But I agree that it might be better to stratify the analysis by party ID factor; that is, two separate tables, one for self-identified Democrats, another for independents/non-Dem, non-GOP voters.

To the other readers:

B is the estimate of each factor's influence over the outcome (voting for Kerry) in nonlinear terms; in technical terms, B is each factor's log odds of voting for Kerry.

The OR is the odds ratio: the odds you will choose Kerry rather than Bush.
* An OR=1.00 implies equal odds.
* An OR>1.00 implies greater odds in Kerry's favor.
* An OR

These are all adjusted for the effect of one another, and then each is assigned a p-value telling us the probability the results we are seeing arose from chance alone (given that the null hypothesis were true). A p-value below 0.05 is generally taken to mean that a factor is statistically significant.

Note the lack of asterisks for Gender, being Hispanic, and a few age groups. This table is telling us that after adjusting for all the factors in the table, only the ones with asterisks should be taken to matter.

Yes, math is hard.

UPDATE: Allan did indeed beat me to it! Good point about the possible collinearity, too.

Posted by Ben | February 21, 2008 2:16 PM

Barry- thanks. The values then are what I thought them to be. I'd junk the PartyID term from the regression for the purposes of understanding which groups skew democratic-- mixing up your result variables with your test variables can really screw thinks up.

Posted by Dean Chung | February 21, 2008 2:17 PM

Ah, I didn't notice that it was a logistic regression. So if you were unmarried you were 60% more likely to vote for Kerry. Or, 38% of unmarrieds vote for Bush and 62% voted for Kerry. That makes a lot more sense.

I guess you have the large constant b/c the inputs have a 0 or 1 value. If the mean for each input were made to be zero, the constant term would be small and represent the popular vote split b/w Kerry and Bush.

Posted by mpowell | February 21, 2008 2:20 PM

As a married person with kids, I find it difficult to believe that marriage has a stronger influence wrt to voting than kids.

The difference between my life pre- and post-marriage was virtually nil. We still lived in the same apartment as we did before we were married. We still both held the same jobs. Still travelled.

After the first baby, mom stopped working, we moved to a house to get more room, had to get a different vehicle. The entire tax structure we lived in changed too.

These kinds of changes (IMO) would likely have a much larger impact on voting patterns.

Still a democrat, tho.

Posted by Strega Nona | February 21, 2008 2:21 PM

I guess you have the large constant b/c the inputs have a 0 or 1 value. If the mean for each input were made to be zero, the constant term would be small and represent the popular vote split b/w Kerry and Bush.

If I had to guess, I would say that the constant is large and significant because Party ID is in the model -- so in essence the constant can be simplified to tell us the likelihood of voting for Kerry in 2004 if you were a Republican voter.

Posted by Ben | February 21, 2008 2:23 PM

Ah crap. Looks like I mangled the constant term-- it just sort of looked weird to have such a big one floating out there.

Posted by Dean Chung | February 21, 2008 2:25 PM

As a married person with kids, I find it difficult to believe that marriage has a stronger influence wrt to voting than kids.

I tend to agree.

Posted by Strega Nona

BTW, my 4-year old has your pasta pot. But there's this one problem....

Posted by Ben | February 21, 2008 2:30 PM

Countdown to apoplectic comments censuring Matt for referring to Steve Sailer's article in 3, 2, 1...

Wow, Fred. You certainly are paranoid. As a White guy, I wonder if that has any affect on your IQ. I sure hope it doesn't drag down the median White IQ.

But you're right, Fred. Steve is a proud racial warrior and has nothing to be ashamed of, nor should Matt be censured, let alone, apoplecticly, for publishing the guy who wrote this:

In contrast to New Orleans, there was only minimal looting after the horrendous 1995 earthquake in Kobe, Japan—because, when you get down to it, Japanese aren't blacks.

And this:

New Orleans' blacks are not even an above-average group of African-Americans, such as you find in Atlanta or Seattle, but more like Miami's or Milwaukee's.

And this:

What you won’t hear, except from me, is that "Let the good times roll" is an especially risky message for African-Americans. The plain fact is that they tend to possess poorer native judgment than members of better-educated groups.

Can we get an article on how voting patterns related to racial IQ, Steve/Fred?

Is it only truly the brilliant, White elite who flock to Ron Paul?

Please enlighten your racial inferiors, Steve.

Posted by SoCalJustice | February 21, 2008 2:35 PM

I. Can't. BELIEVE. You. Linked. To a...STEVE SAILER piece!!!

Posted by razib | February 21, 2008 2:55 PM

SoCalJustice,

I hope that made you feel better.

As for the quotes from Sailer's essay on the aftermath of Katrina, "Racial Reality and the New Orleans Nightmare", even the NY Times's Nick Kristoff noted the difference between the Kobe and Katrina aftermaths, though, unlike Steve Sailer, he feigned perplexity as to the reasons for the difference. If memory serves, Kristoff said the "reasons are complex, and partly cultural". In the sentence immediately following the first one you quoted, Sailer gave a specific and relevant example of the differences, on average, between Japanese and blacks:

"For example, the per capita imprisonment rate for Asian-Americans is about 1/30th that of African-Americans."

The rest of the essay -- with the possible exception of the unfortunate term "native judgment" -- is insightful and accurate, despite your attempts to selectively quote from it out of context. For example, you left off this first part of the second sentence you quoted,

"Judging from their economic and educational statistics"

To give the impression that Sailer was making a baseless invidious comparison between Atlanta and Seattle blacks versus New Orleans blacks. Of course this sort of thing is easier for you to do than to actually refute any of Sailer's points.

Posted by Fred | February 21, 2008 3:24 PM

Here are some simpler numbers from Democratic pollster Stanley Greenberg's article, as summarized in "Value Voters:"

"Bush carried 61 percent of married non-Hispanic white women but merely 44 percent of single white females—a 17-point difference. Among white men, Bush won 53 percent of the single and 66 percent of the married guys—a 13-point difference."

Posted by Steve Sailer | February 21, 2008 3:40 PM

Feel better?

As a member of the race with a high median IQ, Fred, why should I feel bad?

Oh, that has me thinking (something that Steve has told me my racial group is really good at...), perhaps it's not me that needs to be made to feel better, but the insecure White yahoos who make up the core of Steve's audience?

See Fred, Steve's "main goal" (which you were so concerned about yesterday) and his raison d'etre is to make people like you feel better, not me.

It's almost working, even. Almost. But your defensive and paranoid nature is getting in the way of the fulfillment of your IQ, which is your birthright.

You should relax.

Posted by SoCalJustice | February 21, 2008 3:41 PM

The marriage gaps between states have a big impact on the Electoral College results: Bush carried the top 25 states, while Kerry won 16 of the lowest 19 in terms of white women being married during the 18-44 year old child-bearing years.

http://www.amconmag.com/2008/2008_02_11/article.html

Posted by Steve Sailer | February 21, 2008 3:48 PM

Strega Nona writes:

"As a married person with kids, I find it difficult to believe that marriage has a stronger influence wrt to voting than kids."

Yes, I too was surprised to see that on a state-by-state basis, being married among non-Hispanic whites seemed to have more influence on voting than giving birth. I'd noticed the fertility-politics connection back before the 2000 election when comparing Vermont to Utah, but I didn't notice the being married connection until shortly after the 2004 election.

Of course, there's _much_ overlap between the two, but when you put whites' total fertility rates and years married rates together in a multiple regression model you get an even higher r-squared of 88%, so they seem to have some independent effect.

I've noticed that Southern state whites tend to have high years married rates, but only moderate total fertility rates, meaning low fertility per number of years married. In contrast, Great Plains / Great Basin whites tend to be high on both numbers, with Utah being the most extreme example, with a sizable lead on all other states in both years married and total fertility.

In contrast, Arizona, Nevada, and Colorado are fairly low in being married, but high in total fertility. My offhand guess is that family-oriented refugees from California's high land prices are driving up their fertility rates. In other words, their cultures aren't particularly pro-marriage, but they attract young couples from California who specifically want a house with a big yard for the kids they intend to have.

Posted by Steve Sailer | February 21, 2008 4:03 PM

If these strong correlations hold up to scrutiny, the question then becomes: what is the dominant causation? Is it that being married causes people to become conservative, or is it that being conservative causes people to marry earlier? Both are probably true to some extent.

I live in an expensive blue town in a blue state, and it seems to me that the median age of women with small children is OLD. As in, early to mid 40s. These people are generally very liberal, so it supports the idea that liberals tend to marry later and have fewer children.

Posted by Jim W | February 21, 2008 4:07 PM

Isn't linking to Steve Sailer in a way a violation of that new warning / Terms of Use thing below the comment box?

Posted by Reality Man | February 21, 2008 4:09 PM

SoCalJustice,

Still no explanation from you as to what part of Sailer's essay on New Orleans you dispute, and still no attempt by you to refute any of his ideas or argue another position. If and when you decide to argue about ideas, I'll respond. If your next post is another puerile attempt at provocation, I'll ignore it.

Posted by Fred | February 21, 2008 4:12 PM

Here's something Greenberg that answers questions about the effect of age vs. marriage:

"Unmarried 18- 29 year olds gave Kerry a 25 point margin, while younger married women, like their older counterparts, gave President Bush an 11 point margin."

Now, that's not quite satisfactory because race, among other things, plays a big role. But, let's be clear, the importance of the marriage gap on individual voting is not something I discovered. It's been well known among serious voting analysts for quite a few years. It doesn't get all the publicity of the Gender Gap, but people who know their numbers know that the Gender Gap is mostly driven by single women, perhaps because they tend to rely on government more for jobs, Social Security, disability, and welfare.

Posted by Steve Sailer | February 21, 2008 4:12 PM

"It does seem worth saying, however, that "unmarried" is an awful heterogenous situation."

But if married as a category includes more than one heterogenous group (e.g. never-married and divorced), then it would be more remarkable, not less, that there is still a statistically significant correlation.

Kerry lost the gender gap. Another good reason not to nominate military people who will then try to out-bellicose the GOP candidate.

Posted by Colatina | February 21, 2008 4:16 PM

As others have pointed out, putting party ID in the regression makes it hard to interpret. The idea of a "marriage gap" is that married people are more likely to both support republicans and register as republicans. This regression shows only that married DEMOCRATS were more likely to support Bush than single DEMOCRATS, and married REPUBLICANS were more likely to support Bush than single REPUBLICANS. (And even this conclusion is suspect without doing a lot of checking to see how robust this result is to model specification.)

A little statistics can be a dangerous thing, because so few people really understand it.

Posted by ed | February 21, 2008 4:16 PM

Any correlations with children? I.e. are single mothers or married mothers less likey to vote Dem than single or married childless females? Sorted by age.

Posted by MNPundit | February 21, 2008 4:17 PM

Is that a promise, Fred?

I dispute the part about Black people being dumb, the part about them possessing poorer "native judgment" then members of "better educated" groups, and I dispute (I guess that's not the best word, but I'll go with it anyway out of sheer boredom on this topic) the relative ease with which he conflates alleged low IQ and lack of education.

But feel free to ignore me anyway. You should be much more concerned with your high levels of paranoia and defensiveness, honestly.

Posted by SoCalJustice | February 21, 2008 4:24 PM

treating divorced people and never-married people as part of the same category seems like the kind of thing that could easily create misleading results.

right, or people like me who have a 7-yr relationship but don't happen to conform to our state's statutes regarding number of people in the proposed marriage who have dangly parts (conforming answer: 1...ALWAYS 1)

Posted by MattInLA | February 21, 2008 4:53 PM

treating divorced people and never-married people as part of the same category seems like the kind of thing that could easily create misleading results.

right, or people like me who have a 7-yr relationship but don't happen to conform to our state's statutes regarding number of people in the proposed marriage who have dangly parts (conforming answer: 1...ALWAYS 1)

Posted by MattInLA | February 21, 2008 4:55 PM

If your next post is another puerile attempt at provocation

Like your very first post in this thread, Fred?

Are you and Sailer really claiming that a statistic about Asian-Americans meaningfully supports a broad cultural contention about Japanese people who live in Japan? You really aren't worth wasting the time to refute.

Posted by hello | February 21, 2008 5:00 PM

Sorry, but the missing variable (gay/straight) makes this regression pretty much useless in terms of marriage.

Working with a few cautious assumptions (at least 3-4% of the population is gay, that they vote 80% or more for Kerry, and that 99% of them are "unmarried") means that we have basically a subset of unmarried that are completely biasing results.

Posted by dan k | February 21, 2008 5:15 PM

Now the really interesting issue is the correlation among states between both being married and total fertility among non-Hispanic whites and the affordability of forming a family. House costs and house inflation correlate quite well with marriage and fertility among whites, which correlates well with voting Republican:

"Moreover, the Mortgage Gap has been growing. Bush was victorious in the 26 states with the least home-price inflation since 1980, while Kerry triumphed in the 14 states with the most. Home prices rose fastest in Kerry’s Massachusetts (515 percent) and second slowest in Bush’s Texas (89 percent), trailing only nearby Oklahoma. The correlation between low housing inflation and Bush’s share of the vote was strong, with a correlation coefficient, or “r,” of 0.72.

http://www.amconmag.com/2008/2008_02_11/article.html

Posted by Steve Sailer | February 21, 2008 5:16 PM

And, what the Red-Blue gap among states boils down to is inland vs. coastal:

"First is the Dirt Gap: Republican regions simply have more acres of land per person. Even excluding Alaska, counties that voted for Bush are only one-fourth as densely populated on average as Kerry’s counties. Blue State metropolises, such as Boston, Seattle, and Chicago, are mostly located on oceans or Great Lakes, so their suburban expansion is permanently limited to their landward sides. (That’s why Chicago has a West Side but not an East Side.) In contrast, Red State metropolises (such as Atlanta, Phoenix, and San Antonio) are mostly inland. They tend to be surrounded by dirt, not water, allowing their suburbs to spread out over virtually 360 degrees. The supply of suburban land available for development is larger in Red State cities, so the price is lower."

More land means lower land prices, which means people can afford to buy a house with a yard at a younger age, so they get married and start families younger -- as Ben Franklin pointed out in 1751.

"

February 11, 2008 Issue
Copyright © 2007 The American Conservative

Value Voters

The best indicator of whether a state will swing Red or Blue? The cost of buying a home and raising a family.

by Steve Sailer

No matter who wins the 2008 presidential election, pundits will afterwards hypothesize feverishly about why the country is so divided into vast inland expanses of Red (Republican) regions versus thin coastal strips of Blue (Democratic) metropolises. Yet looking at 2000 and 2004, few will stumble upon the engine driving this partisan pattern, even though the statistical correlations are among the highest in the history of the social sciences.

The Republicans lost the popular vote in 2000 while advocating a “humble” foreign policy and won in 2004 while defending a foreign policy that Napoleon might have found bombastic. But all that happened from 2000 to 2004 was that virtually every part of the country moved a few points toward the Republicans. The relative stability of this Red-Blue geographic split suggests that more fundamental forces are at work than just the transient issues of the day.

Neither Jane Austen nor Benjamin Franklin, however, would have found the question of what drives the Red-Blue divide so baffling. Unlike today’s intellectuals, they both thought intensely about the web linking wealth, property, marriage, and children. They would not have been surprised that a state’s voting proclivities are now dominated by the relative presence or absence of affordable family formation.

First-time readers of Pride and Prejudice frequently remark that Austen’s romance novels are, by American standards, not terribly romantic. She possessed a hard-headed understanding of how in traditional English society, wedlock was a luxury that some would never be able to afford, an assumption that often shocks us in our more sentimental 21st century.

Economic historian Gregory Clark’s recent book, A Farewell to Alms, quantified the Malthusian reality under the social structure acerbically depicted in Austen’s books. The English in the 1200-1800 era imposed upon themselves the sexual self-restraint that pioneering economist Thomas Malthus famously (but belatedly) suggested they follow in 1798. By practicing population control, the English largely avoided the cycles of rapid growth followed by cataclysmic famines that plagued China, where women married universally and young. The English postponed marriage and children until a man and woman could afford the accouterments suitable for a respectable married couple of their class.

In the six centuries up through Austen’s lifetime, Clark found, English women didn’t marry on average until age 24 to 26, with poor women often having to wait until their 30s to wed. And 10 to 20 percent never married. Judging from the high fertility of married couples, contraceptive practices appear to have been almost unknown in England in this time, but merely three or four percent of all births were illegitimate, demonstrating that rigid premarital self-discipline was the norm.

Remarkably, a half-century before Malthus’s gloomy and Austen’s witty reflections on life and love in crowded England, Ben Franklin had pointed out that in his lightly populated America, the human condition was more relaxed and happy. In his insightful 1751 essay, “Observations concerning The Increase of Mankind,” Franklin spelled out, with an 18th-century surfeit of capitalization, the first, nonpartisan half of the theory of affordable family formation: “For People increase in Proportion to the Number of Marriages, and that is greater in Proportion to the Ease and Convenience of supporting a Family. When Families can be easily supported, more Persons marry, and earlier in Life.”

He outlined the virtuous cycle con-necting the colonies’ limited population, low land prices, high wages, early marriage, and abundant children: “Europe is generally full settled with Husbandmen, Manufacturers, &c. and therefore cannot now much increase in People. … Land being thus plenty in America, and so cheap as that a labouring Man, that understands Husbandry, can in a short Time save Money enough to purchase a Piece of new Land sufficient for a Plantation, whereon he may subsist a Family; such are not afraid to marry…” Franklin concluded, “Hence Marriages in America are more general, and more generally early, than in Europe.”

The Industrial Revolution broke the tyranny of the Malthusian Trap over food, but the supply of and demand for land never ceased to influence decisions to marry and have children. As America’s coastal regions filled up, affordability of family formation began to differ sharply from state to state (disparities partially masked over the last few years by subprime mortgages and other financial gambits). CNN reported in 2006: “More than 90 percent of homes in [Indianapolis] were affordable to families earning the median income for the area of about $65,100. In Los Angeles, the least affordable big metro area, only 1.9 percent of the homes sold were within the reach of families earning a median income for the city of $56,200.”

When I lived in the Midwest, from age 24 to 34, I attended numerous weddings, but as my social circle matured, the invitations naturally dried up. Yet when I moved back to my native, but now much more expensive, Los Angeles in 2000, I suddenly started being invited to weddings again. Like male characters in a Jane Austen novel, four of my seven closest friends from my high-school class of 1976 got married and bought houses for the first time in their early forties.

Similarly, the cost of childrearing varies more across the country than ever before. A study of census data by the New York Times found that “Manhattan’s 35,000 or so white non-Hispanic toddlers are being raised by parents whose median income was $284,208 a year in 2005.” Second was San Francisco, where the 50th percentile of income for white parents of small children fell at $150,763. That explains a lot about why the city by the bay is last in the country in percentage of residents under 18, below even retirement havens such as Palm Beach.

The culture wars between Red and Blue States are driven in large part by these objective differences in how family-friendly they are, financially speaking. For example, according to ACCRA, a nonprofit organization that measures the cost of living so corporations can adjust the salaries of employees they relocate, the liberal San Francisco-Oakland area is twice as expensive as the conservative Dallas-Fort Worth Metroplex. The BestPlaces.net calculator reports, “To maintain the same standard of living, your salary of $100,000 in San Francisco-Oakland-Fremont, California could decrease to $49,708 in Dallas-Fort Worth-Arlington, Texas.”

Not surprisingly, the San Francisco area is popular with people who don’t need a big backyard for their kids, such as homosexuals and childless couples, while North Texas attracts families from across America. San Francisco is very Democratic, while the Metroplex is quite Republican.

Why? The simplest explanation is that GOP “family values” resound more in states where people can more afford to have families. In parts of the country where “Families can be easily supported, more Persons marry, and earlier in Life.” And where it is economical to buy a house with a yard in a neighborhood with a decent public school, you will generally find more conservatives. It’s a stereotype that marriage, mortgage, and kids make people more conservative, but, like most stereotypes, it’s reasonably true. You’ll find fewer Republicans in places where family formation is expensive. Where fewer people can form families, Republican candidates making speeches about family values just sound irrelevant or irritating.

The arrow of causality points in both directions. Some family-oriented people move to more affordable states in order to marry and have children, while people uninterested in marriage and children move in the opposite direction to enjoy adult lifestyles. This population swapping just makes the electorate more divided by geography rather than tipping the national balance toward one party.

Still, for the many Americans whose innate inclinations fall somewhere in the middle, the cost of forming a family in their current state affects how likely they are to start down the path toward married-with-children conservatism and therefore, cumulatively, which party will eventually prevail nationally.

Imagine a young couple considering marriage who live in the San Francisco Bay Area. He makes $60,000 and she makes $40,000 annually. If he could find a job that pays $50,000 in northern Texas, where costs are only half as high, she could stay home and raise the children. But if they can’t bear to leave California, with its inspiring scenery and lovely weather, she will have to keep working. And if she has to work, are children really such a good idea? And if they aren’t going to have children, why get married at all? And if they aren’t married, are they going to appreciate the nagging of socially conservative politicians?

Four interlocking reasons explain why the affordability of family formation paints the electoral map red or blue.

First is the Dirt Gap: Republican regions simply have more acres of land per person. Even excluding Alaska, counties that voted for Bush are only one-fourth as densely populated on average as Kerry’s counties. Blue State metropolises, such as Boston, Seattle, and Chicago, are mostly located on oceans or Great Lakes, so their suburban expansion is permanently limited to their landward sides. (That’s why Chicago has a West Side but not an East Side.) In contrast, Red State metropolises (such as Atlanta, Phoenix, and San Antonio) are mostly inland. They tend to be surrounded by dirt, not water, allowing their suburbs to spread out over virtually 360 degrees. The supply of suburban land available for development is larger in Red State cities, so the price is lower.

To demonstrate this, consider the 53 percent of the nation’s population who live in the 50 largest metropolitan areas. Among these folks, 73 percent of the Blue Staters live in metropolises bounded by deep water, compared to only 19 percent of the Red Staters.

The second major factor in the Red-Blue divide is the Mortgage Gap. As the law of supply and demand dictates, the limited availability of suburban dirt in most Blue States means housing generally costs more.

This has a striking political corollary. According to ACCRA, Bush carried the 20 states that have the cheapest housing costs, while Kerry won the nine states that are most expensive. The states with the lowest-cost housing are Mississippi (where Bush won an extraordinary 85 percent of the white vote), Arkansas (home state of Bill Clinton but now solidly Republican) and the GOP’s anchor state of Texas.

In recent years, the most expensive state for housing has been California. Although GOP presidential candidates carried California nine out of ten times from 1952 to 1988, they have not come close in the four elections since. Next most expensive are Hawaii and the District of Columbia (where Bush won only 9 percent).

Of course, Blue State cities are also more likely to use environmental and zoning restrictions to limit housing supply artificially. Portland, Oregon, for instance, is an inland city that pretends to be a coastal city by outlawing development of most adjoining land, thereby inflating home costs. This has helped turn Portland, once a blue-collar burgh, into one of America’s most fashionable cities. Indeed, so many young whites have moved to Portland that some are now gentrifying stretches of the inner city’s Martin Luther King Boulevard. (A cynic might suggest that the fact that Portland’s leftist land-use regulations tend to drive out poor blacks and slow the influx of Hispanic illegal immigrants is not an accidental bug but a planned feature.) These development restrictions make children more expensive, as the title of a 2005 New York Times article focusing on Portland made clear: “Vibrant Cities Find One Thing Missing: Children.”

Moreover, the Mortgage Gap has been growing. Bush was victorious in the 26 states with the least home-price inflation since 1980, while Kerry triumphed in the 14 states with the most. Home prices rose fastest in Kerry’s Massachusetts (515 percent) and second slowest in Bush’s Texas (89 percent), trailing only nearby Oklahoma. The correlation between low housing inflation and Bush’s share of the vote was strong, with a correlation coefficient, or “r,” of 0.72.

A rule of thumb in the social sciences is that correlation coefficients of 0.2 are low, 0.4 moderate, and 0.6 high. Thus 0.72 is quite high, especially given the complexity of voting patterns.

To put the influence of housing inflation in perspective, compare its correlation with voting to a more obvious factor influencing who a state votes for: the minority proportion of the state’s electorate. Nationally, Bush carried 58 percent of the white vote compared to only 23 percent of the minority vote. Yet the percentage of minority voters in a state correlated with Bush’s share of the vote only at the moderate -0.37 level.

To further help explain the importance of a correlation coefficient, you should multiply the number by itself. Squaring 0.72 reveals that the amount of variation accounted for by the relationship between housing inflation and 2004 voting was 52 percent of the total. In contrast, squaring the 0.37 correlation for minority share shows it can only account for 13 percent of the variance, just one quarter as much as housing inflation can.

Despite the explanatory power of the Dirt Gap and the Mortgage Gap, these concepts have not been widely discussed. Perhaps they are too objective, too emotionally neutral. What people want to hear instead are justifications for why they are ethically and culturally superior to their enemies."

For an earlier article on Dirt Gap explaining why California shifted from Republican to Democrat and Texas went the other way, see:

http://www.isteve.com/2005_Dirt_Gap.htm

Posted by Steve Sailer | February 21, 2008 5:22 PM

"Countdown to apoplectic comments censuring Matt for referring to Steve Sailer's article in 3, 2, 1..."

Fred may see this as a fit subject for amusement, but I have to admit that I find it a deeply troubling issue. On one hand, I take the idea of the 'reality based community' quite seriously, with the ongoing and shared attempt to ever more closely approach a distant truth being something akin to a moral good. If some individual, no matter how otherwise repugnant, has a genuine contribution to make, - well, it has to be approached with honesty and intellectual integrity, as part of a robust marketplace of ideas (though we accept that even the freest marketplace needs some standards).. Nor can we bluntly crop history according to ideology or even basic decency: life isn't that simple. Louis Agassiz's disgusting works of "scientific" racism in the 1850s and 60s gave great comfort to slaveholders, but he also did important work in 'discovering' the ice ages, and in biology. We can say the same of figures like Madison Grant (Nazi favorite, but also helped save the bison and redwoods) and so on - no doubt, say, Carleton Coon - and perhaps even that evil bastard Eugen Fischer - have genuine insights somewhere in their obsolete and obscene oeuvres. Go far enough down this route, and we're pulling Shakespeare off library shelves because of Shylock, etc.

But. At the same time, linking to VDARE-contributing drooling racists who are in this regard the moral equivalent of a rectal fistula - whatever their other talents - provides them with legitimacy. How many readers had their first encounter with the above-linked individual today solely as a intelligent provider of statistical scholarship? If Madison Grant's hysterical screechings about - for example - how New York was being cunningly taken over by hordes of Polish Jews had merely been, say, occasional letters printed in the Times to general consensus that they were the unbalanced outpourings of an eccentric gent, that would be one thing. Instead Nazis found in his popular work of "Nordic theory" additional (though certainly neither necessary nor sufficient by itself) "scientific" justification to slaughter millions, some of whom might otherwise have been safe in the U.S. had xenophobically strict immigration restrictions not been put in place some years earlier, thanks in part to the tireless efforts of - Madison Grant. For people already in the U.S., Grant's influence included his work towards passing Virginia's 1924 "Racial Integrity Act," the law overturned by Loving v. Virginia 43 years later.

And of course, while Grant's an important figure in this hideous history, I'm somewhat singling him out: for every Grant there were a host of similar types as active, and many even more so in their particular domains.

Now, that's not to say the above-linked individual sinks to such a level - certainly not in influence, and his racism is indeed a less blatant and brutal affair, as befits an age that has yet to entirely forget the consequences of his intellectual heritage.

But still - so what should we do? I don't - can't - insist that Matt mustn't link to him; this would betray that first ideal, no?. But perhaps a warning should be provided - and we have to thank SoCalJustice for giving valuable context here for those who don't know. But what about low-information blogskimmer?

Posted by Dan S. | February 21, 2008 6:53 PM

blogskimmers, that should be.

Posted by Dan S. | February 21, 2008 6:58 PM

Re: the Gender Gap is mostly driven by single women, perhaps because they tend to rely on government more for jobs, Social Security, disability, and welfare.

So single women do not work in the private sector? And married people don't rely on Social Security? Do you live in some alternate reality?

Re: Bush was victorious in the 26 states with the least home-price inflation since 1980, while Kerry triumphed in the 14 states with the most.

Perhaps the folks in the high-housing cost states blame Bush and the GOP for that fact. Which is kind of unfair, but fairly or not, presidents usually do get the blame or economic woes.

Re: The English in the 1200-1800 era imposed upon themselves the sexual self-restraint that pioneering economist Thomas Malthus famously (but belatedly) suggested they follow in 1798.

Not really. Mistresses, bastards and wehores abounded in the era. Probably people practiced primitive contraception (coitus interruptus etc.) to avoid unawnted children.

Re: Judging from the high fertility of married couples, contraceptive practices appear

How does that follow. Married people are far more likely to have and want kids. That's true even today for crying out loud! Hopefully no historian will compare the difference in birth rates to the married and the single among our middle class and conclude that we were ignorant of contraception and must have practiced monkish asceticism before marriage.

Posted by Jonf | February 21, 2008 7:26 PM

Granted I'm in mass comm and not poli sci, but really: standardized betas, please.

Also: If you don't know enough about stats to know that a B of -0.001 isn't "a weak independent correlation," but actually indicative of the absence of a relationship, you ought not be blogging about stats. Seriously, it would be like somebody remarking that a trailing basketball team is getting really undisciplined when they start fouling a lot at the end of the game.

Posted by Aaron S. Veenstra | February 21, 2008 8:53 PM

SoCalJustice,

"I dispute the part about Black people being dumb. I dispute the part about Black people being dumb, the part about them possessing poorer "native judgment" then members of "better educated" groups, and I dispute (I guess that's not the best word, but I'll go with it anyway out of sheer boredom on this topic) the relative ease with which he conflates alleged low IQ and lack of education."

OK, this is at least progress. I haven't been ignoring you, btw. Just had to step away from the interwebs for a few hours. As for your points:

Too my knowledge, Sailer has never called blacks "dumb".

As I mentioned before, I think the phrase "native judgment" was a poor choice of words, but there is a lot of evidence that, on average, African Americans demonstrate poorer judgment than other groups. For example, savings rates for African Americans are dwarfed by those of whites who are at the same income levels. The disparities in illegitimacy and crime rates also support the generalization of poorer judgment. In particular, higher illegitimacy rates and lower savings rates are both symptomatic of a broader failure to plan for the future, which, in itself, is a sign of poor judgment. Similarly, not preparing for an enormous storm when you have several days warning is an example of poor judgment. Most white New Orleans residents evacuated beforehand. With five days notice, you could have walked to safety even if you had no money.

The lower average IQ of African Americans (relative to other groups) is supported by copious data and is not merely "alleged", but I agree that this shouldn't be conflated with education levels (although there some obvious correlation).

Posted by Fred | February 21, 2008 9:06 PM