Religion and Income

Kevin Drum looks at some Andrew Gellman charts and concludes that while richer states are less religious than poorer ones, "Interestingly, there appears to be no correlation between income and religiosity within states."

But that's not really what this second chart says. Rather, as Gellman puts it "overall we see a positive correlation between income and religiosity in poor states and a negative correlation in rich states." Basically, if you live in a poor state, then the richer you are the more likely you are to go to church, whereas if you live in a rich state it's the reverse. I wonder to what extent that finding might just reflect a U-shaped distribution of church attendance with people in the middle more likely to be observant than those at either extreme. I also wonder how this would look if we used educational attainment instead of income.

Comments

Another explanation might be that religions states and nonreligious states are on different sides of a stark cultural divide. In one culture, there exists a religion that serves as a marker of ethnic/group identity for that culture; therefore, religion is a sign of in-group status, so richer people are more likely to display religiosity to indicate their dominant place within the group. In the second culture, religion is not an important group identifier, and so those most likely to be religious are poor people who seek an alternate source of affirmation.

Posted by Mr. Noah | November 4, 2007 10:22 PM

Isn't it more likely in poor states, the rich voters are more likely to be Republicans, and, hence, they feel a need to lie about their religiosity so that they can play up their stereotype as "value voters?"

Posted by Dan the Man | November 4, 2007 10:37 PM

There are way too many states at the 0.00 line for any relationship to be eyeball-able in this graph, and depending on what the N is, I'd guess that anything between 0.05 and -0.05 is probably not significant.

Posted by Aaron S. Veenstra | November 4, 2007 10:43 PM

I also wonder how this would look if we used educational attainment instead of income.

I wonder what it would look like if anybody bothered to control for cost of living vs. income. I suspect the distinction would mostly disappear, except for the effect that people in more religious states tended to vote for Bush and vice-versa.

max
['Bleh.']

Posted by max | November 4, 2007 10:50 PM

It's Gelman, not Gellman

Posted by spell_check | November 4, 2007 11:00 PM

I wonder if poor in poor states are working on Sunday when they'd rather be (or be seen) in church.

Posted by ed | November 4, 2007 11:29 PM

One should also distinguish between different forms of religiosity. E.g. personal piety, believing in the basic articles of one's faith, praying regularly and having a robust faith in the divine, versus more 'formal' piety (attending worship, voting largely based on the political positions of the church, belonging to church affiliated organizations, etc.)

For example, in Latin America while the poor and working classes are very religious in terms of personal beliefs, formal religiosity has traditionally been more characteristic of the middle classes. (this is in contradistinction to North America). the middle classes were also the bulwark of the centrist Christian Democratic parties while poor and working-class Catholics tended to be more open to voting for the Left (and the elites of course for the far right).

I don't know that working on Sunday is really that much of a problem, most churches I've been to seem to have Saturday evening or weekday services as well.

Posted by Hector Dauphin-Gloire | November 4, 2007 11:49 PM

In places without an adequate social safety net, churches fill in the gap--for their members. So you may be looking at the causal arrow running in the opposite direction; in the red states, church members are less likely to become really poor. I'd also imagine that churches serve an important social networking function in many of these states even for well off people, increasing their income. The metaphor that comes to mind is a protection racket; but I'm not a big fan of organised religion so am probably a bit too jaundiced.

Posted by kali | November 5, 2007 4:43 AM

I would appriaciate some definition and an example for the correlation coefficient, because it is hard to interpret it. What does it mean when the coefficient is -0.1 or +0.1?

One can also wonder how important these stats are. There was an article in NYT Magazine predicting a big drop in correlation between religiosity and voting patterns. The fact is that the marriage of conservative religion with free market dogma is not a natural one, and dare I say, not made in heaven, and the war and Social Security debate alienated many conservative Christians.

It seems that this issue was significant among Catholics before it reached Southern Protestants, which makes sense: Catholic church was never fond of free market dogma and Pope was agaist war, but apparently many fundamentalist pastors and lay elders started to notice that Gospels are not really anti-welfare and anti-tax.

Posted by piotr | November 5, 2007 6:00 AM

It is fairly well documented that people significantly misrepresent their church attendance in polls - if actual attendance tracked reported attendance, churches would be a lot more full than they are on Sunday. Given that fact, viewers should note that this chart really tracks reported attendance at church, not actual attendance.

Posted by MDtoMN | November 5, 2007 6:38 AM

I think there's a real danger when you use correlation coefficients as low as the ones reported to argue just about anything.

Typically the R-squared, or square of the correlation coefficient, is used to measure how much explanatory power one variable has in terms of another. Here, we're talking R-squareds of less than 2%, meaning there are a gazillion other things going on.

The signal-to-noise ratio here is pretty much at the noise end of the spectrum.

Posted by Andrew | November 5, 2007 9:57 AM

I would appriaciate some definition and an example for the correlation coefficient, because it is hard to interpret it. What does it mean when the coefficient is -0.1 or +0.1?

It means how much of the variance in one variable corresponds to variance in the other variable. So 1 means the two variables move exactly together and -1 means they move exactly opposite. (Variance is the sum of the squares of deviation from the mean.) So 0.1 means that if you know nothing about a group of people except their incomes, you'll be able to predict about 10% of the variation in church attendance among them. That is not trivial -- in fact it's a pretty big effect to see for income when you consider how many other factors drive church attendance.

To a social scientist, this looks like a fairly strong relationship.

And Max: controlling for cost of living would just compress the graph along the x-axis, it wouldn't change the relationship at all unless you think there are a lot of states with high incomes and low cost of living, or vice versa.

Posted by lemuel pitkin | November 5, 2007 10:14 AM

or more likely -- rural people go to church more than urban folks, and income is only related through geogrpahy.

Posted by Chris | November 5, 2007 10:16 AM

"or more likely -- rural people go to church more than urban folks, and income is only related through geogrpahy."

It's not supported by the data - Vermont is both a rural state and one of the least correlated states. Kansas, North Dakota and Montana are either less correlated or at the same level as Mass. Both rural and urban states can vary widely.

Also, don't assume, however, that rural (whatever that means, the definitions vary massively) equates to lower levels of income or wealth. That's simply not the case in many areas - and in some places, is actually the converse.

Posted by burritoboy | November 5, 2007 10:37 AM

Is anyone else surprised at how much richer blue states are than red states? I mean, I knew that they were somewhat better off, but not THIS better off. Setting aside the three richest red states (all three trending blue), which would not rank in the top 5 of blue states, and the three poorest blue states, which (aside from Maine) would rank in the top half of red states, there does not appear to be another red state that is more affluent than any other blue state. Maybe Florida and Wisconsin.

Posted by Joe | November 5, 2007 2:30 PM

Is anyone else surprised at how much richer blue states are than red states?

Actually, no, but this may not prove what you I think you want it to. Consider the 'richest' state on the graph, Connecticut: my WAG is that 2 or three hedge fund billionaires in Greenwich would equal (in annual income) about 25,000 people making $150+K per year added to, say, Wyoming's population, which could only boost thinly-populated WY's average. And two or three hedge fund billionaires in Manhattan cancel out thousands of New Yorkers under the poverty line.

Posted by Colin | November 5, 2007 4:07 PM

Is Gelman really using average income, or is he using median income? I can't really tell from his site. If he is using average, then yes, Colin's points about hedge-fund billionaires cancelling out zillions of poor people holds true. If he's using median income, then it doesn't. It's an important distinction, and most social scientists use median, not average. Same with housing prices.

Posted by Adam Villani | November 5, 2007 4:21 PM

I don't think you can make any assumptions based on anything with this data, there are too many variables that are not stated. In other words, this thread is just for shooting the breeze. :)

Posted by Dan the "main" man | November 11, 2007 6:41 AM

I don't think you can make any assumptions based on anything with this data, there are too many variables that are not stated. In other words, this thread is just for shooting the breeze. :)

Posted by Dan the "main" man | November 11, 2007 6:42 AM