Multilevel perspectives on modeling census data

Since most census data are released for spatial aggregates, the microscale of people and the macroscale of places are confounded in analyses. Although regrettable, this situation is usually tolerated owing to the other obvious attractions of census data. In this paper, we consider how multilevel statistical procedures offer a solution to this problem. Importantly, we show how they allow places to be considered in terms of several different scales simultaneously. As we demonstrate, this provides important connections with recent moves towards performance review in several areas of public policy. An analysis of data on illiteracy from the 1991 Indian Census provides an illustration of multilevel approach and its usefulness.