A Bayesian approach to infer geographical origins of migrants through surnames

Surnames are an easy tool to analyse human genetic structure, mobility and evolution. Few studies use surnames to estimate human migration at different geographical level.

Here we propose the application of a Bayesian method to estimate the probability of geographical origin (pgo) of migrants in a given area using surnames.

This method can be applied with data recordings when they are available for at least two successive periods and in the areas which are the potential sources of emigration. The principle is that the new surnames which are arriving during the second period in the area under investigation can provide information on their geographical origins. The probability of the origin of migrants can easily be estimated iteratively from the frequency of surnames by using the Bayes' theorem.

This method is exemplified using civil birth registers at different geographical scales. The pgo of migrants, estimated between two periods (1891-1915 and 1916-1940), (i) from French departments to Paris (ii), from these departments to Tarbes, and (iii) from counties surrounding Tarbes to Tarbes, are mapped and discussed.