[marcosdumay]

I'd try an hypothesis that religion gives a huge fixed amount of happiness, at the cost of a small amount of unhappiness that slowly accumulates through the entire life, and even through generations.

Looks like the most logical outcome for me, but I don't know how to test it - or even if there is enough data in the world to test it now. Anyway, the usual correlation is not causality applies here.

[crassus]

Given that the vast majority of all people that have ever lived have been religious, your hypothesis is going to be very hard to test. Now what do we call untestable beliefs?