POLIDATA Document Link Page (links multiple pages as one document)

Subject Keywords: CENSUS ADJUSTMENT; REAPPORTIONMENT; FEDERAL AID; REDISTRICTING

Document Title: Census Adjustment-Sampling Basics

Note on format: Some files are provided in .pdf format, created via Adobe (TM) Acrobat software. If you do not have the FREE software required for viewing/printing these documents, it can be downloaded from the Adobe website.

Why a census ? The Census of Population and Housing is a decennial effort, undertaken by the Bureau of the Census, an agency of the U.S. Department of Commerce, to conduct an "actual Enumeration" of the inhabitants of the United States. The Census is a constituional provision of Article I, Section 2, as modified by the 14th Amendment, which reads: "Representatives shall be apportioned among the several States according to their respective numbers, counting the whole number of persons in each State...". Therefore, the Constitutional reason for having a census in the first place is to determine how many seats each state has in the U.S. House of Representatives. The census is also used to distribute federal and state aid funds and to draw political districts for state legislatures and local governments. Of course, the census is integral to the planning and evaluation efforts of the private sector as well.

Sampling as presently used. Sampling is an integral part of numerous programs of the Census Bureau, but for the Census of Population and Housing it is generally applied only to the so-called "long form" questionnaire provided to about 16% of the housing units for the 1990 federal census. The Census is comprised of two types of questionnaires, the "short form", 7 questions on population and 7 questions on housing, and the "long form" which includes the same 14 questions on population and housing but also numerous (approximately 59) additional questions. Approximately 83% of the population were sent short-forms, the remaining 17% were sent the long-forms.

Why the two forms? The basic 14 questions, while providing sufficient information for the important purposes of Reapportionment and legislative redistricting, do not, by themselves, provide the "interesting" data which most users of the census expect. The 100% count population questions include: 1a) name 1b) temporary resident status 2) relationship of each person to householder 3) sex 4) race 5) age and year of birth 6) marital status and 7) Spanish origin. The 100% count housing questions include: 1) temporary resident status 2) units in structure 3) rooms in structure 4) own/rent tenure 5) 10 acre/business status 6) owner-occupied home value or, 7) monthly rent.

All other questions for which we read analyses in the press come from some sort of sampling effort by the Bureau. The "long-form" sample was filled out by approximately 17.7 million housing units in 1990, or about 1-in-6 of the 102.3 million housing units then identified. "The long form results were weighted to provide data approximating those that would have resulted if every household had completed the long form." (Guide to the 1990 Census of Population and Housing, Part A. Text (hereinafter CPH-R-1A), p.38.

Tabulations of 100-Percent and Sample Data. The Bureau's own documentation indicates the general distinction between the 100% count data and the data estimated by sampling.

"100-Percent Data-These data are prepared by tabulating responses to the 14 basic questions that appear on both the short and long forms. Their preparation, therefore, involves the processing of all short-form questionnaires and the 100-percent questions found on the long forms.

Data are needed on a 100-percent basis for several reasons:

A complete count of the population is necessary to determine the apportionment of congressional seats among States and for redistricting within States to assure equal representation.

Accurate statistics for small areas, such as blocks, are possible only from 100-percent data. [Comment: Perhaps this is why the Bureau provides 100% count data at the block level yet 16% sample data at the block-group level. ]

Figures from the 100-percent tabulation serve as control figures when sample data are inflated to represent the total population.

Sample Data- Sample data are prepared by tabulating responses to the questions that appear on the long forms, including both the 100-percent questions and the additional sample questions. This allows responses to the 100-percent questions to be cross tabulated with responses to the sample questions, which appear only on the long form. For example, this permits preparation of tabulations of race (a 100-percent item) by income (a sample item), or age (a 100-percent item) by educational attainment (a sample item).

Sample data prepared from the long forms are statistically weighted or inflated to produce estimates of what a complete enumeration would have produced. The amount of error present between a weighted figure, based upon a sample, and the corresponding figure if the items were collected on a 100-percent basis varies depending upon the size of the sample and the population size of the area. In printed reports presenting sample data, there usually is a headnote under each table title to remind users that the data are based on a sample." (CPH-R-1A, p.38) (Emphasis added.)

Facsimiles of the forms: (in progress) SHORT-FORM | LONG-FORM