The indirect estimation of migration: methods for dealing with inaccurate and missing data

Andrei Rogers, University of Colorado at Boulder

This paper especially focuses on formal methods for describing, smoothing, imposing, and indirectly inferring migration flows in the absence of adequate, accurate migration data. The term “indirect estimation” is used in demography to describe inferential techniques that produce estimates of a certain variable on the basis of data that may only be indirectly related to its value. We seek a formal model-based approach for using such data to estimate directional (i.e., origin-destination-specific) migration flows. The research reported here should be useful to several three user communities: (1) population researchers faced with the prospective loss of the detailed migration data formerly contained in the to-be-eliminated “long form” questionnaire of past U.S. decennial censuses and its replacement by a significantly smaller continuous monthly sampling survey called the American Community Survey (ACS), (2) historical demographers seeking to identify changing mobility patterns hidden in the increasingly available historical population censuses that lack a migration question, and (3) migration analysts studying migration patterns in data poor less-developed countries, and (4) although the focus is on internal migration, it is believed that the methods can be modified to apply also to international migration. But this is not attempted here.