This paper reports on the development of a demographic model capable of projecting the spatial distribution of population by age and sex for small areas such as census tracts. The proposed modeling framework makes use of two components: the Rogers multiregional population projection model and the aggregate spatial multinomial logit (ASMNL) model. The Rogers model utilizes cohort vital statistics on fertility, mortality, and migration to project the progression of population by age and sex at the regional level, while the ASMNL model extends the capabilities of the Rogers model enabling it to provide population estimates at a finer spatial scale. The model is applied to the Hamilton Census Metropolitan Area in Canada. Our simulation tests highlight the robustness of the proposed modeling approach in utilizing small area characteristics and regional vital statistics to perform small area population projections.