Entropic Spatial Auto-correlation of Voter Uncertainty and Voter Transitions in Parliamentary Elections

Abstract

This paper studies a novel spatial auto-correlation model of voter uncertainty across districts. We use the Moran I index to measure the auto-correlation of Shannon, relative Shannon, Tsallis and relative Tsallis entropies of regional electoral outcomes with respect to geographic adjacency, proximity and sectarian adjacency. Using data from the Lebanese parliamentary elections, we find strong geographic and gravitational adjacency correlations. More importantly, there is a notably strong correlation in sectarian adjacency in both 2018 and 2022 elections, with a very high level of confidence. This result asserts the dominance of the sectarian factor in Lebanese politics. We also introduce the method of maximized general entropy estimation that allows us to determine the Markov transition matrix of voters between consecutive elections.