It has long been recognized that mare basalts on the Moon are preferentially located both on the Earth-facing hemisphere and within large impact basins. A popular model that accounts for this observation assumes that these magmas were denser than the lunar crust, that they accumulated at the crust-mantle interface, and that eruptions occurred only when this magma chamber became overpressurized. In this paper, we re-evaluate this model and argue that it is not consistent with the available data nor with models of dike propagation. As an alternative hypothesis, we propose that magma buoyancy is the predominant factor that determines whether mare basalts erupt at the surface or form crustal intrusions instead. We have computed the densities of mare basaltic magmas and find that some are, in fact, less dense than the Moon's upper anorthositic crust. Based on the widely accepted view that the lunar crust becomes more mafic with depth, we also show that all mare basaltic magmas should be less dense than the lower portion of the crust. Thus, if the upper anorthositic crust was regionally removed by an impact event, then any mare basaltic magma could have risen to the surface there based on buoyancy considerations alone. In support of this model, we note that mare basalts are indeed found wherever geophysical crustal thickness models predict the upper crust to be absent. Furthermore, many of the basalts that erupted within the anorthositic highlands are predicted to be less dense than the underlying crust based on remote sensing data. The high titanium flows within Oceanus Procellarum are somewhat problematical to our model in that an anorthositic crust is predicted to be present beneath them. Using results from recent lunar thermal models, we suggest that these magmas may have overcome their negative buoyancy in the crust by possessing superliquidus temperatures. If magma buoyancy does indeed control whether or not a basaltic eruption will occur, then this implies that the quantity of magma produced beneath the South Pole-Aitken basin was about 10 times less than that of the nearside.