SOURCE: Compiled by staff from reviewing census questionnaires.

NOTE: The 1980 and 1990 censuses used one long form with a variable sampling rate depending on the estimated population size of the area; the overall long-form sampling rate was about 19 percent of households in 1980 and 17 percent in 1990.

The 1970 census used matrix sampling, with two long forms, one containing items asked of 15 percent of the households and the other containing items asked of 5 percent; some items were common to both forms (producing a 20% sample).

The 1960 census also used matrix sampling, but to a lesser extent and with a complex design that is difficult to summarize. A few long-form housing items were asked only in large cities of 50,000 or more; a few other long-form housing items were asked only outside large cities. In addition, in areas using the conventional, single-stage procedure (see text), enumerators asked long-form questions of 25 percent of households overall, but asked some housing questions of only 20 percent of households and other housing questions of only 5 percent. In areas using the two-stage procedure, 25 percent of households were asked to fill out and mail back either the large city or other area questionnaire with all of the long-form items. However, enumerators subsequently transcribed some of the housing items for only 20 percent of households and other housing items for only 5 percent. Hence, the combined number of long-form population and housing items in the 1960 census, from the perspective of households asked to respond, ranged from 40-42 items in conventional areas to 48 items in two-stage areas outside large cities (compared with 49 items in the table).

^a Totals should not be regarded as exact, as there is some interpretation as to what constitutes a separate item.

^b Several items on vacant units are omitted because enumerators fill out these items, not household members.