Analysing two opposing parties' votes at the neighbourhood level through political theories, geographical context, and voting preferences is underexplored. This is important in global cities such as Istanbul (T & uuml;rkiye), which is undergoing economic and democratic development. This paper analyses the clusters and hot spots formed by the vote shares of the right-wing Justice and Development Party (AK Party) and the left-wing Republican People's Party (CHP) using Local Spatial Autocorrelation (LISA) and Optimised Hot Spot Analysis (OHSA). Additionally, the paper examines the relationship between the vote rates of the two rival parties and nine socioeconomic, socio-environmental and demographic variables through local spatial modelling. Using Multiscale Geographically Weighted Regression (MGWR), various findings have been reached that these parties, which have been competing against each other since 2002, confirm the polarisation theory in certain social profiles and spaces. The AK Party vote share correlates negatively with the socioeconomic index, but it shows no relation with the built environment or mean household size. Also, there is a positive relationship between CHP vote share and socioeconomic index and built environment, there is a negative relationship between mean household size. This study provides a multi-scale analytical framework to explore the local differentiation of electoral outcomes and political theories. In future studies, comparison with various variables, cities, and periods are recommended.