In the prediction of future household formation, it proves useful to separate movements caused by population changes from those caused by shifts in the occurrence rates of households among a given population. The varying rates of occurrence reflect the differential impacts of economic vairables and social change. This study indicates that until better data are available, forecasts of households will have to rely primarily upon judgment. This can be gained from an analysis of past data plus information carried over from related series which have more frequent observations. No single statistical function can be found to express the manner in which the Census estimates of households have moved. This apparent lack of correlation reflects the need to average several years' data to obtain accuracy. It also stems from relatively few past observations, and the likelihood that significant structural shifts have occurred and have altered past economic relationships.