Last week, the Pew Research Center recently released its report on marriage in America. Based on data from the U.S. Census American Community Survey for 2008, it provides a wealth of data on marriage and divorce across the 50 states. Check out the map here. Catherine Rampell provides a nice summary over at Economix.
The thing that jumped out at me was the "Larry King" statistic - the number of people who have been married three or more times.
About one-in-twenty Americans who ever have been married said they had been married three or more times. That comes to 4 million men and 4.5 million women.
States varied a lot on this. Arkansas had the highest percentage of "serial marrieds," 10 percent. This was five times more than New York, New Jersey, and Massachusetts with just two percent. The study found that multiple marriages were less likely in states with high concentrations of college-educated people, and more likely in states with lower incomes and smaller college-educated populations.
Over the weekend, I enlisted my number-crunching colleague Charlotta Mellander to look at what other factors might be related to such serial marriage. We looked at unemployment, the class composition of the workforce, immigration, gay population, religion, and levels of psychological well-being. Our analysis points to associations and not causal relationships. It shows that a relationship exists, but not that one causes the other.
Class: Serial marriage was less likely in states with high creative class concentrations (a correlation coefficient of -.59). Conversely, it was was much more likely in working class states (.63). The effect of class was about the same as for income (-.58) and human capital (-.65). When we controlled for income, the association between class and marriage remained significant (-.33 for the creative class and .39 for the working class). Class appears to have a relationship to multiple marriage which is distinct from income.
Immigrants, Gays, and Bohemians: Multiple marriage was significantly less likely in states with high immigrant concentrations (-.38). Multiple marriage was also less likely in states with high bohemian concentrations (-.49). So much for the libertine bohemian lifestyle - at least when it comes to multiple marriage that is. There was no correlation between multiple marriage and the share of the gay population.
Well-Being: Multiple marriage was less likely in states with high levels of psychological well-being (-.37).
Richard Florida
I may not be very well informed in statistics for social science, but can you actually fit lines through those data points? Especially with an R^2 of 0.176? I feel like for many sciences that wouldn't be a very good fit.
RE: InF, the r^2 doesn't really matter but we do need standard errors for those coefficients. They mean nothing without them.
We don't need "standard errors." The question is: are multiple marriages correlated with religion? The answer is: yes. There are other factors, obviously, which is why it's more likely in Oregon than in South Carolina; religion explains only a very small part of why people have multiple marriages. But if you chose two states at random, the one reporting higher religiosity is likely to report more multiple marriages.
If the purpose with this would be to predict multiple marriages, then religion will explain very little obviously. However, as stated above, the purpose is just to show the relation between the two variables, and then the low R2 value shouldn't be a problem. It (and the related correlation coefficient) will simply tell us there is a positive and signficant relation between the two, and that religion will predict very little.
One factor you haven't looked at is "percent Catholic". The states with very high multiple-marriage rates all have small Catholic populations; California has a lower multiple-marriage rate its reputation suggests, possibly because it has a large Catholic population.
"Multiple marriage was also less likely in states with high bohemian concentrations (-.49). So much for the libertine bohemian lifestyle - at least when it comes to multiple marriage that is."
I'd imagine that's because bohemians don't marry often, if they marry at all. The data doesn't say anything about the frequency of serial serious relationships.
If I had to guess, I would expect that a strong predictor would be how common marriage is locally. If "everyone is married", then there would be stronger pressure to marry again than in a place where marriage is less common. Of course, I've simply restated the result as a cause but I think the point is clear (and is consistent with the religion, immigrant and, perhaps most importantly, "bohemian" correlations).