The new algorithm, implemented in the free software R, will allow media, political parties and election researchers to answer new questions to better inform their audience, to design better election marketing strategies or to find answers to apparent enigmas. Jose M. Pavia is a full professor of Quantitative Methods at the Universitat of València and director of the group in electoral Processes and Public Opinion. Rafael Romero is a full professor (retired) of the Universitat Politécnica de Valencia.

Although polls allow the so-called vote transfer matrices to be approximated, they have a number of limitations that make it imposible to use them on many ocassions. On one hand, there are many elections (municipal or historical) where polls are not available. On the other, the results that they offer are very general. The new algorithm allows to answer questions from historical elections, small electoral places and having an estimation of vote transfer matrices from the census tract level. This will allow to study why the behaviour isn’t the same in all the neighbourhoods and to know the variables that explain this different behaviours.

So questions like: where do the Vox votes come from? Who made it possible for Nazis to come to power in the inter-war Germany? what is the voting loyalty of left-wing parties? Where have the voters of the French Socialist Party gone? What do young people vote for? and many other questions can be answered with this new algorithm created by Pavía y Romero.

As the authors point out, the new algorithm will be useful not only to answer questions about electoral behaviour but “also can be used in disciplines such as economy, marketing, epidemiology and law.” Moreover, “although currently there are other algorithms to find solutions to this problem, the advantage of ours is its higher accuracy and its great simplicity, any person without technical knowledge can use them. Simply enter the data and the software will return the results. Furthermore, with the new developments that we are implementing, in which we are incorporating artificial intelligence techniques, we are improving the accuracy of the estimations. We have gone from having a percentage of errors of 10% to having under 7%’.

The article where the new algorithms are described is ‘Improving estimates accuracy of voter transitions. Two new algorithms for ecological inference based on linear programming’. Sociological Methods and Research and the software where the algorithms are implemented is Iphom.  https://cran.r-project.org/web/packages/lphom/index.html.