sampling reduced the burden of the long form on individual households (the 1960 census also used matrix sampling but to a more limited extent).
The table indicates that the burden of the long form has progressively declined in terms of the sampling rate. Twenty-five percent of households were designated to receive the long form in 1960, 20 percent in 1970, 19 percent on average in 1980, and 16 percent on average in 1990. The 1970 census used a type of matrix sampling for both population and housing items: 15 percent of households received a markedly shorter long form, and 5 percent received a longer long form; some items were common to both forms, producing a 20 percent sample for those items.
The 1960 census also employed matrix sampling but to a more limited extent from the perspective of responding households. Different forms were used in large cities of 50,000 or more population and other areas; each contained a couple of questions unique to the form. In addition, in conventionally enumerated areas in 1960 (see below), 20 percent of households were asked one set of housing items by enumerators, and 5 percent were asked another set; some items were asked in common, producing a 25 percent sample for those items. In the remaining areas, however, which covered about 80 percent of households, every household in the 25 percent sample was asked to fill out and mail back a long form with all of the sample questions for the type of place (large city or other). Enumerators then transcribed the answers to computer-readable forms, transcribing the answers for 25, 20, or 5 percent of households, depending on the item (see Table A.2).
Table A.2 lists the various detailed items on the census. The housing portion of the long-form questionnaire no longer includes such items as possession of a washing machine or other appliances. The population portion of the long-form questionnaire, over time, has included more questions on disability and ancestry (although the question on birthplace of parents was deleted after 1970) and fewer on marriage and schooling. The greatest number of questions are related to employment, occupation, and income. For the income question, each added income source (e.g., pensions) is counted as a separate item because the respondent must consciously disaggregate the preceding year's income into component parts
Table A.3 lists the total number of both vacant and occupied housing units and the number of occupied housing units (households) in 1970, 1980, and 1990. Total housing units have increased over time from 69 million in 1970 to 102 million in 1990; total households have increased from 64 million to 92 million over the same period. Also shown are the number of households receiving the short and long forms and, for each form type, the number receiving and returning