Accuracy and Coverage Evaluation (A.C.E.) Program.
A coverage evaluation program conducted by the Census Bureau following the 2000 census; it produces estimates of undercount and overcount in the census and forms the basis for statistical adjustment of census counts through dual-systems estimation. In the A.C.E., a sample survey is conducted in a sample of census blocks after the nonresponse follow-up phase of the census is complete. The resulting sample of individuals found by the survey in the selected blocks—called the P-sample—is matched to the set of census enumerations from the sample blocks (the E-sample).
Address Control File (ACF).
The 1990 census analogue to the Master Address File used in 2000. The ACF was the residential address list used to label questionnaires, control the mail response check-in operation, and determine the nonresponse follow-up workload.
Address List Improvement Act of 1994 (P.L. 103–430).
The law that enabled two innovations in the construction of the Master Address File for the 2000 census: the Local Update of Census Addresses Program (allowing local governments to receive and review the address list) and address list updates from the U.S. Postal Service’s Delivery Sequence File.
Records that are collected as part of the operation of federal, state, and local programs, typically fund allocation and tax programs, such as Internal Revenue Service and Food Stamp Program records.
American Community Survey.
A new continuous survey program currently being developed by the Census Bureau to collect the detailed socioeconomic and other data currently included on the census long form, based on monthly surveys of respondents and released annually. It is hoped that implementation of the American Community Survey will allow the Bureau to switch to a short-form-only census in 2010. The planned sample size is about three million households per year.
Type of error cited by the Executive Steering Committee for A.C.E. Policy in its recommendation not to adjust census counts for congressional redistricting. Balancing error occurs when cases in the P-sample and E-sample are not treated identically (e.g., when the search area used to identify P-sample matches and E-sample correct enumerations is defined differently).
A program introduced in the 2000 census that made census questionnaires available in public places, so that residents who believed that they had been missed in the regular census enumeration could file a questionnaire.
Descriptive term used to differentiate basic types of addresses. Areas of residences with mainly city-style addresses (number and street) are said to be inside the blue line, and areas of residences with mainly non-city-style addresses (such as rural route and post office boxes) are said to be outside the blue line. The term derives from the color used on initial sets of maps generated prior to the Local Update of Census Addresses (LUCA) Program; the 1998 implementation of LUCA targeted addresses inside the blue line and the 1999 implementation targeted those outside. Mailout/mailback enumeration was used inside the blue line and update/leave enumeration was used outside the blue line.
Group of one or more census blocks expected to contain about 30 housing units, defined for use in the Accuracy and Coverage Evaluation Program.
A program in which postal workers determine addresses for which they did not receive a questionnaire and notify the Census Bureau.
The smallest entity for which the Census Bureau collects and tabulates decennial census information; bounded on all sides by visible and nonvisible features shown on Census Bureau maps. Occasionally, especially in rural areas, drainage ditches or power lines may be used to define blocks. Because most blocks have small population and housing unit counts, only 100 percent data, or short-form data, are tabulated for them.
The target date of a decennial census. Census Day is the date for which census respondents are supposed to describe their household population, and for which the results of a decennial census are supposed to be an accurate representation of the nation’s population. Since 1930, Census Day has been April 1 of years ending in zero.
A census-defined geographic area of roughly 2,500 households. Census tracts are aggregations of census blocks (roughly 150 blocks, dependent on the population of the area). Tracts are intended to be relatively stable entities over time, though their definitions do shift with each census.
Census 2000 Supplementary Survey.
Pilot program for the American Community Survey; a survey to collect data items from the census long form that was conducted in monthly samples totaling 700,000 households in 2000.
The last stage of nonresponse follow-up when enumerators are instructed to make a last attempt to obtain at least minimal information, from a proxy if necessary. Imputation is used to fill in any missing information.
Coefficient of variation.
An assessment of the variability of an estimate as a percentage of the size of the quantity being measured.
Computer-aided personal interview (CAPI).
The use of a computer to assist an interviewer in carrying out an interview. Advantages include avoiding errors in skip patterns, providing immediate edit checks, and expediting electronic data capture.
Computer-assisted interviewing (CAI).
A group of methods for using computers to assist with data collection. CAI surveys can be either interviewer-administered (conducted in person using a laptop computer or by telephone using a shared computer) or self-administered (conducted by surveys disseminated to respondents by telephone, by the Internet, or on a computer disk).
A (technical) bias in dual-systems estimation by which the estimated counts would be, on the average, either too low or too high, caused by heterogeneity in enumeration probabilities for both the census and the postenumeration survey. The heterogeneities of the probabilities for these two attempted enumerations are typically positively related, which causes the estimated counts to be on the average too low.
Coverage correction factor (CCF).
A term related to dual-systems estimation. The CCF is defined as the dual-systems estimate for a post-stratum divided by the census count (including whole person imputations and late additions); hence, it is the multiplier that can be applied to the population count for a post-stratum in a particular area to generate an adjusted count.
Statistical studies conducted to evaluate the level and sources of coverage error in censuses and surveys.
Coverage improvement follow-up (CIFU).
The second-stage follow-up operation used in the 2000 census (performed between June and August, 2000), verifying findings from the initial nonresponse follow-up.
Coverage improvement programs.
Often (but not always) nationally applied methods and programs that attempt to collect information from individuals and households that might be missed using mailout/mailback or nonresponse follow-up. Before the 2000 census cycle this term referred to such programs as the parolee and probationer program (used in 1990), in which lists of these individuals were checked to see whether they were enumerated, and the non-household sources program, in which several administrative record lists were matched to census records to try to identify people missed in the census for purposes of field follow-up (used in 1980). For the 2000 census, coverage improvement refers more to efforts to complete the address list, use of multiple response modes, and service-based enumeration.
The practice by which a census enumerator fabricates a questionnaire for a residence without actually visiting it.
Current Population Survey (CPS).
Monthly sample survey of the U.S. population that provides employment and unemployment figures as well as current data about other social and economic characteristics of the population. The CPS is collected for the Bureau of Labor Statistics by the Census Bureau. The sample size for the CPS is about 50,000 households per month.
The process by which survey responses are transferred from written questionnaires to an electronic format for tabulation. In the 2000 census, data capture was done by optical scanning; from 1890 to 1950, punch cards were used for data capture, and the FOSDIC process was used from 1960 to 1990.
Delivery Sequence File (DSF).
The master list of deliverable mail addresses maintained by the U.S. Postal Service, organized by carrier route. The Delivery Sequence File was first used as a source of updates to the Master Address File for the 2000 census, following enactment of the Address List Improvement Act of 1994.
A method that uses various administrative records (especially birth and death records, information on immigration and emigration, and Medicare records) and information from previous censuses to estimate the total number of people in various demographic groups resident in the United States on a specific date, and therefore their census undercoverage.
The largest census test, typically 2 years before the decennial census, in which the methods and procedures of the upcoming decennial census are given their final test to identify any operational problems.
An estimation methodology that uses two independent attempts to collect information from households to estimate the total population, including the number of people missed by both attempts.
The set of census enumerations for a sample of census blocks; part of the Accuracy and Coverage Evaluation.
A census field operations employee who collects information from respondents through interviews.
The inclusion of someone in the census in error. Such inclusions may be people born after Census Day or deceased before Census Day, people in the United States temporarily, and people in the wrong location. They also include people counted more than once, i.e., duplicates.
The difference between an estimate and the true value.
Executive Steering Committee for A.C.E. Policy (ESCAP).
The committee of senior Census Bureau staff charged with analyzing information from the 2000 census and Accuracy and Coverage Evaluation in order to decide whether census counts should be adjusted for net undercount. The ESCAP reports to the director of the Census Bureau, who in turn submits a formal recommendation to the U.S. Secretary of Commerce.
A secondary census or survey operation, predominately in data collection, carried out to successfully complete an initial operation. It is most often a telephone or personal visit interview to obtain missing data or clarify original responses.
FOSDIC (Film Optical Sensing Device for Input to Computers).
From 1960 to 1990, census questionnaires were microfilmed. The answers were read from the microfilmed questionnaires using FOSDIC and converted to electronic codes on computer tape.
The sum of erroneous enumerations and omissions in the census. See also Erroneous enumeration; Omission; Overcount; Undercount.
A place where people live that is not a housing unit. There are two types of group quarters: institutional (for example, nursing homes, mental hospitals, and correctional institutions) and noninstitutional (for example, college dormitories, ships, hotels, group homes, and shelters).
The technique used by the Census Bureau to impute missing responses on census questionnaires. Imputations are made based on a continually updated distribution of responses from other, filled-in questionnaires that match characteristics that are known from an incomplete questionnaire.
All the persons who occupy a housing unit.
A house, an apartment, etc., that is occupied (or, if vacant, is intended for occupancy) as separate living quarters, which are those in which the occupants live and eat separately from any other persons in the building. See also household.
Census respondents whose records contain insufficient information for matching, such as would be necessary to obtain adjusted counts through dual-systems estimation. For the 2000 census in particular, the set of IIs contained both persons with substantially incomplete questionnaires and people who were reinstated in the census count at a late stage of processing but excluded from the Accuracy and Coverage Evaluation.
A method for filling in missing information. Sequential hot-deck imputation fills in information from a previously processed respondent (and therefore geographically close) with other similar characteristics.
Integrated coverage measurement (ICM).
The use of a postenumeration survey and some type of estimation method, e.g., dual-systems estimation, to produce adjusted census counts in time for apportionment and therefore all uses of census data. ICM was a key part of initial Census Bureau plans for the 2000 census, but was abandoned after the Supreme Court’s 1999 decision ruling out the use of sampling in generating apportionment counts.
Term used in the 1990 census to describe the collection of data from neighbors, apartment managers, post office employees, etc., when a response from a resident could not be obtained.
A method of enumeration in which enumerators canvass a geographic area, list each residential address, and collect a questionnaire from or enumerate a household.
A method of enumeration in which the enumerators list each residential address and at the same time deliver the census form for return by mail.
Census Bureau program in the 1980 and 1990 censuses in which local officials were given the opportunity to review housing unit counts in census blocks.
Local Update of Census Addresses (LUCA).
A Census Bureau program in which local officials were given the opportunity to review individual addresses on the Master Address File and make corrections, additions, and deletions to that list, and to make corrections to census maps to match any changes that may be needed. The LUCA98 Program covered only local governments in mailout/mailback enumeration areas; LUCA99 covered governments in update/leave enumeration areas.
The census questionnaire that is delivered to a (roughly) one-sixth sample of households, which includes the short-form questions and additional questions about income, commuting patterns, etc. See also short form.
Mail Response Rate.
Measure of respondent cooperation in the census, defined as the number of households returning a questionnaire by mail divided by the total number of questionnaires sent out in mailback areas. See also mail return rate.
Mail Return Rate.
Measure of respondent cooperation in the census, defined as the number of households returning a questionnaire by mail divided by the total number of occupied households that were sent questionnaires in mailback areas (excluding vacant households and nonresidential units). The mail return rate is considered a more refined measure of cooperation than the earlier-available mail response rate.
A method of census enumeration used primarily in urban areas in which questionnaires are mailed to each address and the residents are asked to mail back the completed questionnaires.
Master Address File (MAP).
The list of addresses on which the 2000 census enumeration was based. It is derived from the 1990 census address list or a prelisting by census field staff, and is updated using a variety of sources, including information from the U.S. Postal Service and local officials. See also Topologically Integrated Geographic Encoding and Referencing (TIGER) System.
Master Trace Sample.
A sample of census records (possibly by selecting all records in a sample of decennial census blocks) for which all information relevant to census data collection is retained to assist in analyzing and comparing methodologies suggested for use in the subsequent census.
The process through which it is determined how many persons are included in both the postenumeration survey and the census (in PES blocks) and how many persons are only included in one or the other attempted enumeration.
Multiple response modes.
Generally speaking, methods for being enumerated, not including mailout/mailback and enumeration as part of usual nonresponse follow-up. In 2000 these methods included obtaining and returning questionnaires available in public places (“Be Counted” forms), the use of the telephone and the Internet to obtain or provide census information, and the enumeration of persons at places that offer services to the homeless.
The failure to obtain all or part of the information requested on a questionnaire.
Nonresponse follow-up (NRFU).
The field operation whereby census enumerators attempt to obtain completed questionnaires from interviewing members of households for which no questionnaire was returned in the mail. For the 2000 census, NRFU was performed between April and June, 2000. NRFU was conducted on a 100 percent basis in accordance with the Supreme Court’s decision on sampling for apportionment; the Census Bureau’s initial plans for the 2000 census called for sampling in this follow-up phase, which was then sampling for nonresponse follow-up (SNRFU).
A person missed in the census. See also Erroneous enumeration; Overcount; Undercount.
The total number of people counted more than once or otherwise enumerated erroneously in the census. See also Erroneous enumeration; Omission; Undercount.
Sample survey of respondents in a sample of block clusters conducted after (and independent of) the census enumeration; part of the Accuracy and Coverage Evaluation.
The public law that requires the Census Bureau to provide the decennial census data required for congressional redistricting to the states by April of the year following the year of the census enumeration.
Post-Enumeration Survey (PES).
The 1990 census analogue of the Accuracy and Coverage Evaluation in 2000.
Separating a data set collected through use of sampling into strata on the basis of information gathered during data collection, and then treating each stratum separately in estimation.
A collection (of individuals in the census context) that share some characteristics (e.g., race, age, sex, region, owner/renter) obtained during data collection and that are separately treated in estimation.
Primary selection algorithm (PSA).
Algorithm developed by the Census Bureau to consolidate multiple responses from the same household into a single return; given concerns about opening a loophole for duplicates, the details of the PSA have not been made public.
Enumeration of typically homeless people at food kitchens and shelters.
The census questionnaire that is mailed to about five-sixths of all households. The short form concentrates on basic demographic information. See also long form.
A place where people live or stay that is different from the usual private house, apartment, or mobile home and that requires different decennial census procedures. Examples are hospitals, prisons, hotels, motels, orphanages, nursing homes, dormitories, marinas, military installations, and large rooming or boarding houses. See also group quarters.
Type of error cited by the Executive Steering Committee for A.C.E. Policy in its recommendation not to adjust census counts for congressional redistricting. The population post-strata used to assign coverage correction (adjustment) factors are supposed to be homogeneous in that members of a post-stratum are supposed to be equally likely to be counted in the census and the A.C.E.; synthetic error is produced in adjusted counts when this homogeneity assumption is not satisfied.
The section of the U.S. Code under which the Census Bureau operates. It also protects the confidentiality of census information and establishes penalties for disclosing this information.
Topologically Integrated Geographic Encoding and Referencing (TIGER) System.
The framework for identifying the exact geographic location of residential addresses (as well as other physical features).
A nonspecific term representing either the rate or number of individuals missed (erroneously included) in the decennial census. More specifically, gross undercoverage and gross undercount are the rate or number of those missed for a demographic group or geographic area (similarly for gross overcoverage and gross overcount); net undercoverage and net undercount are the difference between the rate or number of those missed for a demographic group or geographic area and the rate or number of those erroneously included; differential (net) undercoverage and differential (net) undercount are the difference between the rate or number of net undercoverage between two demographic groups or between two geographic areas.
The process by which individuals reported on more than one census questionnaire are identified and counted once at only one geographic location.
(also known as update/leave/mailback). A method of census enumeration used in areas lacking city-style addresses in which the census questionnaire is delivered to an address by a census enumerator. The Master Address File is corrected at the time of delivery (if necessary). Residents at the address are asked to fill out the questionnaire and mail it back.