Reconstructing History: Using Language to Estimate Religious Spread

We introduce a data-driven approach to use language to reconstruct history, and apply the methodology to estimate the geographic origins of religious spread. To validate the approach, we use language data to estimate origins of Islam and Buddhism to within 500km of their true (and uncontested) origins. We then apply the methodology to the more complex (and contested) cases of Christianity, Judaism, and Hinduism. We show that language-based estimates, in these cases, are significantly more aligned with the origin of scripture than with the origin of the religion.