After reading the paper linked in my comment more thoroughly I found the computational method:
It's (1 - (normalized) Herfandahl index). (Page 198, 1st paragraph, sentence 3 in the PDF)
A high number indicates a high fractionalization, a low number (near zero) a mono-religious society (e.g. the State of Vatican City with 100% Roman-Catholics has an RFI of 0.0000 (1 - (1²) = 0.0000)).
Depending on whether you subsume or split all Islamic branches you will get different results for the Religious Fractionalization Index.
Assuming the percentages are:
- 80% Sunni 18% Alevi 1.5% Alawi (sums up to 99.5%)
- 0,2% Christian
- 0,04% Jewish
you get two different results:
1 - (0.995²+0.002²+0.0004²) = 0.00995
1 - (0.8²+0.18²+0.015²+0.002²+0.0004²) = 0.327
Alesina probably didn't take into account the different Islamic branches in his Religious Fractionalization Index for some countries and got a result of almost zero for the Turkish society. The same non-distinction was applied to Yemen (~62% Sunni/~38% Shia or 100% Muslim) but probably not to Saudi Arabia (~88% Sunni/12% Shia or 100% Muslim - not considering 1.5 Mio Christian and 0.4 Mio Hindu foreign workers) and an RFI of 0.127 or Afghanistan (~80% Sunni/~20% Shia or 100% Muslim) and an RFI of 0.2717.