Sunday, April 26, 2009

For the love of God

A while back there were blog entry by me about a report released by the Iona Institute & written by Patricia Casey regarding the association between well-being and religiousity. I was quite critical since I think the literature surveyed (which tends to find a positive association if any) was essentially a correlation but appeared to be interpreted as causal i.e that religion made one better off. To paraphrase the great theologian, Fr Jack, "That would be an empirical matter". In the absence of some form of exogenous assignment of religion I don't see how these epidemiological analyses can tell us anything causal.
In the spirit of methodological ecumenism & in a moment of idleness I thought to consider the question of whether particular denominations are associated with greater happiness. Using the ESS (rounds 1 to 3) I regress happiness (on a 0 to 10 scale) on denomination, a measure of religiousity and a bunch of controls (which all behave sensibly, details on request).
Relative to the non-religious, Protestants are the big winners, in fact the only winners. Catholics and Jews are no better off. Orthodox Christians & Muslims are worse off. However there is a positive gradient to (self-reported) religiousity. Frequency of prayer doesn't kick-in & is not included.
The results are robust to choice of estimator: using ordered probit or a Tobit doesn't change things.

(1) happy (0 - 10) (robust t statistic in parentheses)
Roman Catholic 0.0190 (1.08)
Protestant 0.0902*** (5.19)
Orthodox -0.195** (3.26)
Other Christian -0.0492 (1.13)
Jewish -0.288 (1.41)
Muslim -0.349*** (5.48)
Eastern -0.171 (1.64)
Other non Christian -0.118 (0.96)
Religiousity 0.0603*** (23.62)

controls:age, living alone, health, disability, sex, education, whether foreign, country dummies.


Liam Delaney said...

This is probably not possible with the ESS data but identifying religiosity could be done in a couple of ways - regional factors related to religion but not well-being (assumptions here but theoretically possible). Also, potentially things like the Random Roommate literature. Again, not without problems but technically possible.


Kevin Denny said...

Yeah interesting idea. There is also information on lagged religion i.e. " did you believe?" but this might not be excludable even 'though it will be correlated with current religion. There is a certain asymmetry: most people I guess don't choose to be religious, they inherit it, but some subsequently chose not to be.

Liam Delaney said...

The living in Ireland survey, for example, has religion all the way through. When I have a sufficiently idle moment, I will do the fixed effects regression. I have done this before with a lot of controls (including some fancy dynamic ones) and religion still does fine. Although, again there is a lot of things going on there but they are getting closer to a good test.

Martin Ryan said...

As well as lagged religion, there can also be lagged agnosticism.

For example, "Lapsed Agnostic" tells the story of one man's journey from belief to un-belief and back again.

Kevin Denny said...

Since religion is almost as fixed as one's sex, in a fixed effects model identification would be coming from the switchers into and out of religion. But these are not exogenous changes i.e. they are not representative of the levels: the people who are or are not religious. So I am not sure that an FE model is very informative. The same issue arises with using panel data to look at any discrete variable say union membership in an earnings equation.
Or as JHH might put it, which effect are you identifying?

Liam Delaney said...

Its an interesting question in general - the local effect of "losing your religion" or "finding it" may of course be quite different from the effect of a long-held faith cultivated by birth. The FE critique also applies to the random room assignments. Long-run regional instruments would be more believable to me, particularly given low rates of inter-county migration in Ireland. However, clearly such effects would need to be looked at cautiously particularly in terms of the network effects.

Mark McG said...

Nice to see Martin Luther being vindicated

Kevin Denny said...

Looking at switchers could be interesting: since I am a big REM fan, the idea of writing a paper called "Losing my religion" is very attractive. Thanks Liam.
Of course if I was a certain type of bigot (which I am not) then I would say that that the small extra utility to being Protestant is small compensation for an eternity in hell.

Martin Ryan said...

A new law to obliterate blasphemy may increase the well-being of the religious?

"A new crime of blasphemous libel is to be proposed by the Minister for Justice in an amendment to the Defamation Bill, which will be discussed by the Oireachtas committee on justice today.

Minister for Justice Dermot Ahern proposes to insert a new section into the Defamation Bill, stating: “A person who publishes or utters blasphemous matter shall be guilty of an offence and shall be liable upon conviction on indictment to a fine not exceeding €100,000.”

Kevin Denny said...

Obliterate blasphemy? Yeah but people like that are a miserable bunch anyway.
It all sounds very Life-of-Brian to me.