This study uses panel data to demonstrate two dimensions of land ownership: the distribution between households at a given time and changes within a household over time. We note that recognizing the latter dimension is only possible when analyzing rare long-term panel data. We estimate a model for land ownership using a version of the correlated random effects estimator to uniquely identify the determinants of both dimensions amongst Kenyan smallholders. We find life cycle effects are a key determinant of both distributions, and identify important ways in which initial conditions such as inheritance and off-farm income relate to the dynamics of ownership. We find that population density is a key determinant of differences between households, but also that a given household's land ownership is not affected in the short term as population density increases around them. Controlling for population density, households own more land when they are closer to road networks where the economic value of land is higher. We find important vulnerabilities for the land security of widows, but this vulnerability is geographically heterogeneous.