Bruno Jeanbart (Opinion Way) tells me, after reading the blog on the drivers of the National Front vote, that it would be interesting to add in the model data on the religion of voters: is Catholicism still a bulwark against the National Front?
Not easy to answer that question. Up to my knowledge, there is no exhaustive data source at a detailed local level on the topic. We could have merged our database with survey data, but would have lost one of the strengths of our model: its exhaustive character.
There is however a through-the-back-door way to answer the question. It works splendidly and it is worth telling the story with some details.
The French census collects data on the jobs of people. This is available at the level of towns, across 29 categories. One of this category is labelled « Clergy ». Figures are available for men and women separately.
Putting directly in the model the percentage of clerics in each town does not measure what we want to measure: the intensity of religious practice. For several reasons. First, there is only one priest for every other town: one given priest is working in several towns. Then, attitudes towards religion most probably spread beyond the limits of a town and should be measured on areas of a more significant size. We have calculated the percentage of clerics at several aggregation levels: the « canton » and the « sous-préfecture ».
The right level is the « sous-préfecture »: the more clerics in a « sous-préfecture », the less people in the towns of that « sous-préfecture » are voting for the National Front. The impact is significant: -1% in the FN vote in the towns belonging to the 25% “sous-préfectures” with the highest density of clerics. It indeed seems to be a specific catholic phenomena. We separated the percentages of men and women clerics: only the percentage for the men has a (very) significant coefficient. The coefficient for the women is not significant.
A last learning: this example is one more illustration of the power of the Open Data when you want to address very specific issues and uncover weak signals. Only of course if you know how to manipulate them and you use the right analyses techniques. Also true for lots of databases. The interested readers can contact us to know more: we can work with you to assess how the Open Data or your customers’ database can help you take the right business decisions.