Wednesday, September 14, 2005

Religion, Morality and some Freakonomics

Last week at this book club at work (yeah, yeah, I know. I do not read books on corporate whoring usually but this was an exception. They gave us free food and I got to pick the book), we were discussing Freakonomics. After all the Excel whores in the room were done with why we needed more charts and graphs and numbers in the book in question, we moved on to more interesting things. We came up with a list of questions that we would like to see answered and talked about how we would test our hypotheses. For example, one of our questions was whether there is a correlation between the kind of music one listens to and one's criminal tendencies. Is it possible to predict criminal behaviour of a group of individuals given their music preferences? We talked at length about how we would design studies that would control for all other factors - needless to say, quite a few of these "ideas" would be illegal and so not really practical.

Anyway, one of the questions we came up with but which we did not talk about(for obvious reasons) was whether religion has an effect on morality. I was reminded of this when I saw this post by Primary Red; the claim is that one of the reasons why there is so much lawlessness in India now could be because of "loss of faith". Well, first of all, I am not sure whether A) there is more lawlessness than before and B) we as a society have 'less faith' now than before but hey, those questions not the subject of this post.

This is what interests me - if I would like to test correlation between religion and morality how would I do that? First of all, how do I define being moral? Any definition of morality tends to be relative, so what is my measure of morality? If I were to narrow down a group, say convicted murderers, how would I control for other factors(like environment, "evilness" et al)? Are there are natural experiments(remember the Roe v Wade on crime rate stats was right out there!) at all that I can look at? Any ideas? Open for comments.

PS - I am really not interested in what you think is the answer, most people who read this blog will probably say No anyways. I am much more interested in ideas on proving/disproving the hypothesis.


Falstaff said...

"whether there is a correlation between the kind of music one listens to and one's criminal tendencies."

I'm assuming you're talking about crimes like rape, murder, etc. There are those of who consider listening to hip-hop or boy bands, for instance, a crime against sound waves - in which case the question's fairly moot. :-).

Seriously though, I think the first distinction to make is between crime as acts against the law and crimes as immoral acts - the two are very different (though there's some intersection, I suppose). The latter, I suspect, is untestable, since morality is not only subjective and person / faith dependent, it is also, in a sense revealed faith - so that the very definition of what you believe is in the moral precepts you adhere to (the rest is just talk).

For the faith to illegality correlation, I don't see controlling for other factors being that difficult - with a large enough sample you could just include factors like income levels, education, neighbourhood, etc. in the regression. Two things to think about though:

a) Is your hypothesis related to faith in general (i.e. any faith) or do you see the effects of particular faiths being different? I don't know about crime, but there is certainly research to show that religious preferences influence a number of social actions - so Catholics behave differently from Protestants, etc. If you run the analysis at an aggregate level (i.e. any faith to any crime) you may lose some of the detail. For instance, do particular religions display a higher tendency towards crimes against women?

b) The other thing is - how are you measuring degree of 'faith'. Do you measure it based on self-reports (which are notoriously unreliable)? Do you measure it based on religious activity? If so, how do you control for the correlation of that with your other independent variables. In other words the problem is not that people with low income / education levels may have higher crime hazard (because that you can control for), the problem is that people from low income / education levels may spend more time in Church without being more religious in any true sense. And since both of those things are on the same side of the equation, you're going to have covariance problems in your regression.

P.S. Hey, you ASKED for this.

Veena said...

Falstaff: Yes, I did indeed ask for this and thanks for your time and thoughts. (In other words, dude, you are also so jobless kya?).

Yes, I agree that the only option is to look at crimes as acts against the law. Or the ones where those and the ones as immoral acts intersect - though I guess that becomes difficult when a crime such as mass murder becomes an acceptable moral value to the person or group of people committing the crime.
Digression - It is interesting though to see people associating "immoral values" such as homosexuality to religion -would love to see some data reg those claims.

And I guess you are right - we might just be able to include most of the demographic factors for the regression. But I am not sure whether looking at faith on an aggregate level will skew the study if we are looking at henious crimes. All religions are probably on the same side on this one and I'd think that if we were to see a particular religion display a higher tendency towards crime against women, there probably are other important factors such as geography, education and culture at play. But yes, that needs to be proved before we can proceed. Also, I think there is this huge intersection between what we call culture and what we call religion and I am not really sure how we can clearly demarcate between the two.

How do we measure faith? I have no idea. This one, I concluded was the toughest question to answer when I pondered over this before. What do people really mean when they say they are believers? So many people I know who I think are extremely spiritual would never indulge in any kind of "religious" activity. How do we account for that? And what does being religious in the true sense really mean? That you believe in heaven and hell and that you need to go to church every Sunday and vote for Bush? I am wondering if there's a way we could find out what kind of religious teachings the people in the study imbimbed. Maybe that's a start. Though I guess how do you separate out religious teachings from plain common sense.

Neela said...


If you want to do this experimentally, Vargas and Petty have done something like this (but I don't think thepaper is published yet).

To assess religiousness (whatever the noun form is), you give subjects a vignette which is ambiguous something like "Susan does not go to church regularly, but gives to charity. Blah blah". Then you ask respondents how religious they believe Susan to be (7 or 9 point scales anchored by "not at all religious" to 'extremely religious") What they mark Susan as is the mirror image of how religious they actually are - a not religious person will mark Susan as "extremely religious" while a person who believes himself or herself ot be religious will mark her as "not at all religious".

Now though I am not in favour of median splits because you can lose information, I think you could use this to split into two conditions: religious vs not religious. My fundaes on research methods are still being developed.

Anyhow, after a filler task, you then give all participants a test booklet, consisting of some problems in ascending order of difficulty, say anagrams. They have 15 min or whatever to finish it and the test should be fairly tedious to give them an incentive ot cheat. You tell them that when they finish the test, they should check their answers at the back of the book (which you mention in bold so that even those with poor eyesight will know that the answers are available), and ask them mark themselves and turn the booklet in.

You code people who have cheated (by looking at the answers) - by assessing how many marks they make on the paper (if they don't make any and get the answer right for a difficult question, there is a high probability that they would have cheated. This is what the authors did- am sure you can come up with a more sophisticated measure).

Then you run an ANOVA to figure out whether there is a signifciant difference between groups.

All this is in the Vargas and petty paper - its a paper on attitudes and behavior and the correlation between the two (or rather a survey of previous reserach which shows that often there is little correlation).

Hope this helps!


Falstaff said...

And then people say I'm jobless...