In this study, we introduce a new method of downscaling global population distribution, for which purpose conventional approaches have serious limitations in application. Our approach is “eclectic,” as it explores the intersection between an optimization framework and the empirical regularities involved in rank-size distributions. The novelty of our downscaling model is that it allows city-size distributions to interact with socioeconomic variables. Our contribution to the urban studies literature is twofold. One is our challenge to the conventional view that the proportionate growth dynamics underlies empirical rank-size regularities. We first show that the city-size distribution of a region can deviate substantially from a log-normal distribution with cross-regional and time variations, and then demonstrate that such variations can be explained by certain socioeconomic conditions that each region confronts at a particular time point. In addition to expanding academic debates on city-size distributions, our study can pave the way for various academic and professional research projects, which need spatial distribution of global population at fine grid cell levels as key input. Our model is applicable to the entire globe, including regions for which reliable sub-regional population data sets are limitedly available, and can be extended easily to function as a forecasting model.