Computational demography involves the use of new data and computational methods to improve measurement and understanding of population processes. This dissertation contains three examples of work in the field. Paper 1 tracks the migration decisions of partisan voters to tease apart the roles ideological and racial neighborhood context on individual mobility decisions. Paper 2 combines consumer trace data with administrative and survey data to measure migration flows at a new level of granularity. Paper 3 uses machine learning to improve methods of inferring neighborhood racial composition from datasets where race is not measured. These papers grapple with common issues that emerge when dealing with data that was not constructed for research purposes, including missing variables and non-representatives. Taken together, the work highlights the potential of this field, the challenges that still need to be overcome, and the many synergies obtained by combining together new and traditional approaches.