| Abstract: | This article considers the kinds of data required to increase scientific knowledge about US immigration and, in light of those requirements, assesses the principal currently available data sets and makes recommendations for improving the data environment. Study of US immigration phenomena requires data with 3 properties: 1) that they cover immigrants, sponsors, and nation-states; 2) that they measure and link the migration-relevant attributes of the immigrants, sponsors, and nation-states, including the visa histories of both immigrants and sponsors; and 3) that they capture the migration-relevant attributes in a longitudinal manner. The 2 principal data sources currently available for studying immigration--the US decennial censuses and the administrative records of the Immigration and Naturalization Service--fall far short of the data requirements dictated by current reasoning and models about the migration process. Current data sources do not permit, except under stringent assumptions, the estimation of models of such phenomena as earnings, language acquisition, marital behavior, and fertility behavior. The authors propose both a new data-collection effort and a set of enhancements to existing data sources. Perhaps the most efficient way to learn more about immigration would be to design and implement a new data-collection effort with great success. In the meantime, proposed enhancements to existing data sources could help fill some gaps in present knowledge. |
