from incarcerated persons in order to assess the feasibilities of allocating prisoners to a geographic location other than the prison (Recommendation 7.2). It will also be useful to compare and contrast the ARE information with that generated by related operations—the proposed coverage follow-up operation (Box 6-3) and the postenumeration survey used for coverage evaluation—where more detailed banks of coverage and location probes are permissible (Section 6–E.4).
In the nearer term, we suggest throughout this report a number of deficiencies in current research that should be investigated as part of an ongoing research program at the Census Bureau:
assessing the quality, accessibility, and relevant content of facility and administrative records for group quarters and nonhousehold facilities (Section 7–D.1);
studies that can provide quantitative information (and validate hypotheses based on qualitative techniques like ethnography) on the magnitude and trends in complex and ambiguous living situations (Section 5.2);
basic research on living situations as reflected by census operations, such as the tendency for household members to be listed, roughly, in reverse order by age (Section 5–B.3);
mode effects on response, including both the effects of the mode of administration (e.g., paper versus telephone response) and the general structure of roster types (e.g., whether a question-based or instruction-based approach is easier to follow) (Recommendation 6.4);
effects of visual layout and wording on census questionnaires (Recommendation 6.7); and
impact of length of form (number of questions) on survey response (Section 6–G).
It is with problems like those listed as the shorter-term research tasks in mind that we offer additional comment on the shape and direction of the Census Bureau’s general research and testing program.
The most prominent components of the Census Bureau’s research program are the suite of formal experiments and tests that accompany each decennial census, the evaluation reports of various census operations that are produced after the decennial count, and the set of large-scale tests scheduled regularly between census years. These tend to be large and complex activities—indeed, one of the usual census tests is a dress rehearsal that tries to