This article has three goals. First, it explores the effects of changes in census definitions and concepts on the measurement of living arrangements. As part of this analysis, the authors develop new estimates of the number of households and group quarters in each census year since 1850. Second, they evaluate the existing aggregate statistical series on family and household composition, with particular attention to problems in the measurement of subfamilies. Finally, they describe data and methods for developing a consistent set of statistics for the period since 1850 and offer recommendations for the coherent measurement of family and household composition.