tag:blogger.com,1999:blog-37421406.post3987926725350700438..comments2023-09-30T08:20:57.971-04:00Comments on Bradley Wright's Blog: Ecological fallacies and studies of religionBrad Wrighthttp://www.blogger.com/profile/07800309833079635465noreply@blogger.comBlogger6125tag:blogger.com,1999:blog-37421406.post-83799831304872622622010-02-11T10:39:56.821-05:002010-02-11T10:39:56.821-05:00Let me see if I have this right. The within-state...Let me see if I have this right. The within-state individual correlation is negative in all states (more religiosity, less theft), but the correlation is less strong in more religious states (MS) than in less religious states (MA). (I don’t think the absolute numbers (fewer in MS) matter because we’re looking at <i>rates</i>. ) So both the Godly and the heathen in MS are more larcenous than their counterparts in MA. That would give you the state-level positive correlation and the individual-level negative correlation. But so would the model I offered in my earlier comment.<br /><br />The differences between states on within-state differences are interesting. Andrew Gelman goes into that in connection with economic inequality and voting. There’s an individual-level correlation (Richer/Republican) that is attentuated or even reversed at the state level – richer states (MA, NY, CA, etc.) vote Democratic. He shows that within-state rich-poor political differences are greater in poor states. That is, in rich states, there’s not so much difference between the voting preferences of the rich and the poor.Jay Livingstonhttps://www.blogger.com/profile/06652075579940313964noreply@blogger.comtag:blogger.com,1999:blog-37421406.post-32635631640600578412010-02-11T10:20:21.572-05:002010-02-11T10:20:21.572-05:00Let's see. Suppose that Mississippi has the h...Let's see. Suppose that Mississippi has the high rates of theft, high rates of religion, and a negative correlation between the two among its residents.<br /><br />That probably means religious people in Mississippi commit more theft, on average, than religious people in other states, but less than the non-religious people in Mississippi.<br /><br />Given that there are fewer religious people in Mississippi, relative to other states, that suggests that the within-state negative correlation between religion and theft might actually be greater than the nationwide, no?Brad Wrighthttps://www.blogger.com/profile/07800309833079635465noreply@blogger.comtag:blogger.com,1999:blog-37421406.post-34501512812767910162010-02-11T09:36:38.817-05:002010-02-11T09:36:38.817-05:00Fantastic post. Thanks for sharing it with us.Fantastic post. Thanks for sharing it with us.K T Cathttps://www.blogger.com/profile/10259428595745509790noreply@blogger.comtag:blogger.com,1999:blog-37421406.post-44705579841087149382010-02-10T14:53:53.337-05:002010-02-10T14:53:53.337-05:00Corey: My reasoning went like this: States with h...Corey: My reasoning went like this: States with higher scores on Religiosity have higher rates of Theft (I’m not checking the data here, let’s just assume it’s true). But individuals who are more religious have lower rates of theft than do nonreligious individuals. A religious state (e.g., Mississippi) had a high rate of Theft relative to other states. It also has a high level of Religiosity. It can’t be the religious Mississippians who are giving it that high rate, so the non-religious Mississippians must be working overtime at their larcenous trade. <br /><br />There's probably a flaw in that logic.<br /><br />Actually, what I really think is that you need to factor in economic variables like poverty and inequality, especially when you’re looking at state-level data. If you can copy the data in the table into a spreadsheet and then copy a state GINI coefficient table (I found one recently on the Web but can’t remember where), you might have something interesting to work with. <br /><br />And what I <i>really,</i> think is that it's time to go outside and play in the snow.Jay Livingstonhttps://www.blogger.com/profile/06652075579940313964noreply@blogger.comtag:blogger.com,1999:blog-37421406.post-87686699504903609812010-02-10T13:03:30.439-05:002010-02-10T13:03:30.439-05:00@Jay... I don't quite understand your question...@Jay... I don't quite understand your question. Could you rephrase? <br /><br />The aggregate data presented here demonstrates that occurrences of theft correlate with religiousness in the population (though, to quibble just a bit, the measurement of religiousness is conceptually weak and points more to a cultural orientation rather than a behavior set; that is, in Mississippi, one could claim - and even believe - that religion is important for life, but behave in ways in congruent with that belief). <br /><br />In the meta-analysis studies that measure religiousness (again, differently and one would hope more stringently than gallup does for its poll) find a negative effect between religious belief and theft. (It's not clear from my cursory glance at the article if we are predicting offending, or victimization. I'm sure Brad can clarify). I don't pretend to understand meta-analysis... that's for smarter people than me.<br /><br />In this case, the ecological fallacy still holds. The thorny issue is to theoretically account for the weird covariance. My hunch is that material concerns trump existential ones.Coreyhttps://www.blogger.com/profile/00119612449610307692noreply@blogger.comtag:blogger.com,1999:blog-37421406.post-35720759515344497282010-02-10T11:29:34.668-05:002010-02-10T11:29:34.668-05:00How do you get a positive correlation (with theft ...How do you get a positive correlation (with theft for example)at the state level but a negative correlation at the individual level? Does this mean that in very religious states the non-religious people are really, really bad -- i.e., much more thieving than nonreligious people in nonreligious states?Jay Livingstonhttps://www.blogger.com/profile/06652075579940313964noreply@blogger.com