Yesterday, I attended an interesting presentation by Matteo Cervellati of the University of Bologna that he gave in the Seminar series at Siena. The title of his paper, co-authored with Marcel Jansen (Universidad Carlos) and Uwe Sunde (Univ. St. Gallen) is "Religious Norms and Long Term Development: Insurance, Human Capital and Technological Change". Here's a section from the abstract of the paper:

Under reasonable assumptions, we show that any implementable norm can either enforce charity or extra effort provision, but not both at the same time. The features of the most desirable norm depend on the relative importance of pure hazard (luck) and returns to effort, suggesting that in early development stages charity norms are more desirable, while norms supporting effort provision and self-responsibility are dominant when returns to human capital are relatively high and safe. This is consistent with the observation that the Catholicism is based on norms of charity, while Protestantism promotes self-responsibility and industriousness.

They basically argue for an endogenous, and basically inevitable, 'reformation' of the church as a consequence of life becoming 'less risky'. This lower risk resulted in guarantees to returns to human capital investment that could not be achieved before (the reformation).  The factors that determine religious adoption had to do with the size of the returns to human capital and exposure to risk.  Prior to the reformation, risk was high and returns to human capital investment (if not investment generally) were quite low.  One of the norms which could allow for insurance in this instance would be one based on 'mutual insurance and charity', which the authors identify as a 'Catholic' norm.  However, once the returns to human capital increase and size of or exposure to risk decreases norms that support increased effort provision are, instead, a better method (i.e. a approximate a first best solution in a general equilibrium framework) of insurance.  This is similar to the 'Protestant', Calvinist, or Puritanical systems.  The paper is a work in progress and they are still look at combinations of the two, i.e. systems in which there is partial mutual insurance and substantial - but not extreme levels of - effort provision. They are still working on the math of this, but it looks promising.  Basically, they should be able to cater to a continuum of religions on a line of 'effort provision' to 'mutual insurance' with signals of what are likely to be specific 'sins' in one religion (norm set) vs. another. 

One of the consequences of this research is that if the division of religions was, in fact, endogenous then the use of religion as an 'exogenous' variable in cross-country regressions as a proxy for culture is deeply problematic.  This obviously has a consequence for using such variables as an instrument for culture and thus, when published, the paper should have a substantial impact on the empirical work using 'religion' as a proxy for underlying, unobservable characteristics in a country.