The process of conducting the 1990 decennial census involved a complex set of operations.1 The major components of the 1990 census included the following:
address register/control file;
sampling (long form);
coverage improvement (including promotion, publicity, outreach);
data capture, editing, and processing;
efficient staffing and field organization;
geographic tools and materials;
research and experimentation; and
a multitude of products in a variety of formats to meet the needs of a broad range of users.
These activities were carried out over a period of time both concurrently and sequentially; however, the data collection process itself falls into four basic procedures essentially carried out sequentially, as follows:
developing a mailing list, that is, a list of addresses of all housing units in the country—the list is in effect a master address control file;
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.
OCR for page 228
Modernizing the U.S. Census APPENDIX B The Census Process The process of conducting the 1990 decennial census involved a complex set of operations.1 The major components of the 1990 census included the following: address register/control file; questionnaire mailout/mailback; self-enumeration; sampling (long form); nonresponse follow-up; coverage improvement (including promotion, publicity, outreach); coverage evaluation; data capture, editing, and processing; efficient staffing and field organization; geographic tools and materials; research and experimentation; and a multitude of products in a variety of formats to meet the needs of a broad range of users. These activities were carried out over a period of time both concurrently and sequentially; however, the data collection process itself falls into four basic procedures essentially carried out sequentially, as follows: developing a mailing list, that is, a list of addresses of all housing units in the country—the list is in effect a master address control file;
OCR for page 228
Modernizing the U.S. Census mailing census questionnaires to each address on the list, which the householders are to fill out and return by mail; following up those addresses that fail to report; and carrying out simultaneously a number of processes designed to assure as complete coverage as possible. This appendix reviews the basic data collection procedures for the 1990 census, beginning with descriptions by year of the census process for developing the mailing list, collecting the data, and field operations (opening of offices). Subsequent sections describe the 1990 census process for follow-up operations, coverage improvement, local review, and the Post-Enumeration Survey. DEVELOPMENT OF MAILING LIST (MASTER ADDRESS CONTROL FILE) 1988 Computerized address lists for nearly 60 million housing units (mostly for metropolitan areas) were purchased from commercial vendors. These addresses were assigned to census geography (down to the block level) by using a combination of the automated geocoding capabilities of the TIGER system and clerical geocoding using other sources. The resulting automated address control file then was used to conduct the advance post office check (APOC) by the U.S. Postal Service (USPS), which identified missing addresses and those that needed corrections or were not recognizable for mail delivery. An address list for another 30 million housing units (mostly in more rural areas) was created by having census enumerators canvass the areas and record the mailing address of each housing unit. This operation was called 1988 prelist. The enumerators also noted the location of each unit on a map (for use in finding the unit later if other visits were needed). These computer-generated maps were one of the first major products of the new TIGER system. The addresses were keyed to form the automated address control file for these areas. That file then was used in 1989 for an advance post office check by the USPS. 1989 For the address control file that was based on the purchased vendor files, the list was updated for APOC changes, and then the addresses were printed out by block (grouped together to form roughly equal-sized workloads, called address register areas [ARAs]). The list for each ARA was assigned to a census enumerator who canvassed each street and building in the ARA looking for (and adding) housing units not on the address list. This operation was called precanvass. Updates to the address list from the precanvass were made, and the resulting
OCR for page 228
Modernizing the U.S. Census file was the basis for labeling the questionnaires to be mailed out for these areas in 1990. For the address control file created by the 1988 prelist, the address list was provided to the USPS for an APOC as well. In some cases, the addresses in these areas used mail delivery systems, such as rural route/box number, for which the address could not be located on the ground without additional information. As a result, most information on missing addresses and undeliverable addresses received from the APOC had to be checked in the field by census enumerators, if only to determine the correct census geographic block code. This operation was called APOC reconciliation. After the updates from APOC and APOC reconciliation were made to the automated address control file for these areas, the resulting file was the basis for labeling the questionnaires to be mailed out in 1990. For other rural areas (covering about 11 million housing units) for which most mailing addresses were of the type that could not be located on the ground without other information, census enumerators conducted the 1989 prelist (similar to the 1988 prelist described above). These addresses were not sent for an APOC, however, because they would not be used for a mailout census. Instead, they were used to label questionnaires that were delivered by census enumerators in 1990 during the update/leave/mailback operation (see below). This approach was used because the Census Bureau had found it very difficult in the past to list mailing addresses that were accurate enough for USPS delivery in some rural areas. 1990 For areas for which the USPS would deliver addressed questionnaires in late March (basically, the areas for which the address list originally was purchased from vendors or compiled from the 1988 prelist), the list of addresses was reviewed again by the USPS in early March. Any new addresses identified from this check were added to the address control file and, if there was time, a mailing piece was prepared and mailed in hopes of obtaining a completed questionnaire by mail. Otherwise, the address was visited during nonresponse follow-up (see below) to obtain an interview. COLLECTING THE DATA There were three major collection methods for the 1990 census. The first data collection method was mailout/mailback, used primarily in urban and suburban areas to enumerate about 84 percent of total housing units. The Census Bureau compiled an address list in advance, and the USPS delivered
OCR for page 228
Modernizing the U.S. Census questionnaires to the housing units on that list. The questionnaires were to be returned by mail when completed. The second data collection method was update/leave/mailback used in more rural areas, mainly in the South and Midwest, to enumerate about 11 percent of total housing units. (This method was also used in some high-rise, low-income urban areas, and a variation was used in urban areas having large numbers of boarded-up buildings.) Again, the Census Bureau compiled an address list in advance, but census enumerators (rather than USPS) delivered questionnaires to the housing units in those areas. The questionnaires were to be returned by mail when completed. The third data collection method was list/enumerate, used in the most sparsely populated areas, mainly in the West and Northeast, to enumerate about 5 percent of total housing units. Here, no precensus address list was compiled, but the USPS delivered an unaddressed short-form questionnaire to each unit. Then census enumerators canvassed door to door, listing all housing units, collecting completed questionnaires or completing them as necessary, including for units that did not receive one. In addition, at a predesignated subset of units, the enumerators collected responses to the sample (long-form) questions. In March 1990, the major data collection efforts for the census began. These included the following: For mailout/mailback areas, the USPS delivered the labeled questionnaires on March 23. Most units received the short form containing only the questions asked of all households, but a predesignated sample of the units received the long form with additional questions. The mailing package included the questionnaire, instructions on how to fill out the form and to mail it back by April 1, and a motivational flyer to encourage response. One week later, the USPS delivered a postcard to each unit that reminded households to fill out the questionnaire and return it as soon as possible. The USPS returned about 5 percent of these questionnaires to the Census Bureau as undeliverable. Census enumerators were able to deliver about half of them by hand; the remainder did not receive a mailing piece, so they were physically enumerated during nonresponse follow-up. For update/leave/mailback areas, census enumerators delivered the labeled questionnaires while they canvassed each block looking for new or missed units. As in mailout areas, most units received a short form, but a predesignated sample received the long form. Besides the questionnaire, each unit also received the instructions and the motivational flyer used in mailout areas. Again, the household was asked to complete the questionnaire and mail it back by April 1. The USPS also delivered the reminder card in these areas on March 30. For list/enumerate areas, beginning in March, the USPS provided an unaddressed questionnaire package to each household. These packages included
OCR for page 228
Modernizing the U.S. Census instructions on how to complete the form and to hold the completed form for pick-up by a census enumerator. In addition to visiting each unit to pick up the questionnaires, the enumerators also canvassed their assigned areas to list each address. For units that did not receive a questionnaire, and for units in which the household had not yet completed the questionnaire (or needed assistance to do so), the enumerators completed it concurrently with the canvassing. In addition, using a predesignated sampling pattern, enumerators obtained answers to the additional questions using the long-form questionnaire. Special place enumeration also took place at this point. Examples of special places include group quarters, such as boarding houses, nursing homes, dormitories, rectories, convents, hospitals, YMCAs, YWCAs, and so forth. Enumerators visited these places to obtain the information from each resident. About 10 days before Census Day, the Census Bureau also conducted street and shelter enumeration to enumerate components of the homeless population. The first phase of the operation focused on physically enumerating persons staying in shelters for the homeless, while the second phase focused on physically enumerating homeless persons living outside of shelters, for example, on the street. In addition to the above, there were two other components to special place enumeration—transient enumeration and military enumeration. During transient enumeration, enumerators visited travel places at which guests were unlikely to have been reported at their usual place of residence or unlikely to have a permanent residence. These places included YMCAs, YWCAs, youth hostels, commercial campgrounds, and so forth. For military enumeration, special procedures were used to conduct the physical enumeration of domestic military and maritime personnel. The military bases and vessels were self-enumerating. The bases appointed a senior commissioned officer to serve as the enumeration project officer. FIELD OPERATIONS (OPENING OF OFFICES) 1988 To manage and process the 1988 prelist work, the first 110 filed district offices were opened, as was the first of 7 processing offices. Minicomputers and other automated equipment and software were installed to handle both operations and administrative matters in these offices. 1989 The remaining 339 field district offices (including nine in Puerto Rico) began opening in 1989 to prepare for their major activities in 1990. Also, 13 regional census centers opened to manage the 449 district offices for the data collection efforts in 1990. Finally, the remaining 6 processing offices were
OCR for page 228
Modernizing the U.S. Census opened to prepare for their major activities in 1990. The opening of all these offices included hiring staff and installing various automated equipment and software, including minicomputers for all the district offices, regional census centers, and processing offices. In addition, the processing offices' electronic equipment was installed that would be used in 1990 to convert questionnaire data to computer-readable form. With the installation of the network of computers across all offices, the Census Bureau significantly increased the full-scale decentralized use of its new automated management information system (MIS). This system allowed detailed monitoring of costs, progress, staffing, and other information for all census activities in the district and processing offices. Depending on the size and critical nature of an operation, reports from each office were transmitted to headquarters as often as daily, but more typically they were weekly reports. Throughout the census, this system offered a major improvement in the ability of management officials to track the status of operations, costs, staffing, and possible problem areas in real time. Thus the Census Bureau could take corrective action ''midstream" for many operations, a capability that did not exist for any previous census. 1990 FIELD OPERATION—FOLLOW-UP Mail Return Once the delivery of questionnaires had been completed, the forms began to arrive by mail in the district or processing office serving each area. Mail returns for Type 1 district offices (which covered hard-to-enumerate areas in central cities) went to the processing offices for check-in.2 For all other district offices, the mail returns went to that office directly, as did questionnaires completed by enumerators during list/enumeration or special place enumeration. Both the processing offices and district offices used automated equipment to check in the forms by scanning a bar code on the return envelope. The associated listing in the address control file then could be coded to show a questionnaire had been received for that unit. At the conclusion of the check-in phase, each listing not so coded represented a case that would have to be visited by an enumerator during nonresponse follow-up. Questionnaires checked in at the processing offices immediately were sent for data capture using electronic equipment to produce computer-readable data. This concurrent processing scheme was new for 1990—in previous censuses, this phase did not occur until all physical enumeration was completed. Using the resulting computer file of data, the processing offices conducted an automated edit, or review, of the data to identify which households had not provided complete data or otherwise would need telephone or personal visit follow-up. For questionnaires checked in at the district offices, a similar edit was done by clerks
OCR for page 228
Modernizing the U.S. Census using the actual paper questionnaires. Clerks in both the processing and district offices then attempted to contact these households by telephone to obtain the additional information needed to complete the questionnaire. When this could not be done, and depending on the extent or type of missing data, some of these units later had to be visited by enumerators during field follow-up (see below). Daily reports on the mail return check-in rates for each district office were transmitted to headquarters through the MIS. This information was used to project the likely workload for nonresponse follow-up that was expected to require over 200,000 temporary enumerators to visit a projected 30 million units over a 2-month period. By the end of April, the Census Bureau had to determine the actual number of persons to hire and to begin preparing lists of addresses that had not returned a questionnaire. The response rate (from all households—occupied and vacant) was 63 percent, significantly lower than the projected 70 percent, and as a result, the Census Bureau needed to hire more enumerators for nonresponse follow-up, which in turn led to a request to Congress (which was approved) for an emergency supplemental appropriation of about $110 million. Nonresponse Follow-Up For the nonresponse follow-up, enumerators were provided with a complete address list for their assignment area. The units for which no questionnaire had been received were coded as such. In multiunit buildings and for most rural areas, the listing also provided the surname (if available) of households that had returned a questionnaire. This information was used to help the enumerator locate the nonresponse case in situations for which addresses/unit designations were not clear-cut. During nonresponse follow-up, enumerators were required to make as many as six attempts to contact a household member and complete the questionnaire. If this was not possible after three personal visits and (if possible) three telephone calls at different times and on different days, the enumerator attempted to obtain at least basic information on the household members from a knowledgeable source, such as a neighbor or building manager. Because the nonresponse follow-up had to be completed in a timely fashion so that other operations could be conducted, each district office was authorized to begin a close-out phase once 95 percent or so of the operation had been completed. During this phase, enumerators made one more visit to each remaining case to obtain as complete an interview as possible. Because of very high workloads and/or staffing problems, some offices took up to 6 weeks longer than planned to complete this operation. To reduce these problems and prevent even longer delays, the Census Bureau raised the pay rate in selected areas to attract additional staff and to motivate many part-time staff to work more hours.
OCR for page 228
Modernizing the U.S. Census Field Follow-Up After the nonresponse follow-up was completed, the next major operation (field follow-up) was conducted for two reasons: For each unit that had been classified as vacant or deleted by a nonresponse follow-up enumerator, a different enumerator was sent to verify this status. Past experience had shown that a significant number of such cases (perhaps up to 10 percent) actually were occupied units. This vacant/delete check was the first of several post-Census Day coverage improvement operations. For mail-return questionnaires that had failed the automated or clerical edit, and for which follow-up by telephone was not possible, an enumerator visited the household to obtain information needed to complete the questionnaires. OTHER COVERAGE IMPROVEMENTS In addition to the vacant/delete check component of field follow-up, a number of other coverage improvement efforts took place after the conclusion of nonresponse follow-up. The "Were you counted?" campaign provided an opportunity for persons who believed they had been missed to report data for their household on a form printed in newspapers, distributed through other mechanisms, or by calling one of the toll-free telephone numbers. These questionnaires then were processed during the search/match operation (see below). The parolee/probationer check was conducted because research had shown this group may have been disproportionately undercounted in previous censuses. Each state and the District of Columbia were asked to participate by distributing questionnaires through parole and probation officers to those under their jurisdiction. The parolees and probationers were asked to provide their Census Day address, and completed forms from this operation then were processed during the search/match operation (see below). During the search/match operation, forms received from various activities were checked against completed questionnaires to see if the persons on the forms needed to be added to the census questionnaire for the reported Census Day address. This had to be done because these types of cases might otherwise result in duplication. For example, a "Were you counted?" questionnaire might be sent in by someone who did not know that another household member had mailed back the original questionnaire for the address. Similarly, parolees or probationers might have been reported as a household member by someone on a regular census questionnaire. This approach also was needed to process individual forms filled out by persons temporarily away from home in hotels; military
OCR for page 228
Modernizing the U.S. Census personnel at U.S. bases or on shipboard; and whole households who were at a second home or temporary address, but reported their usual home was elsewhere. Between late July and early October, the Census Bureau recanvassed over 500,000 blocks containing about 15 million housing units, or about 15 percent of all housing units. This operation, called the housing coverage check, was done for these blocks based on a variety of data sources, most of them internal to the Census Bureau. The sources included: on-going internal count review analyses that were based on comparing 1980 data, new construction and demolition data, and the preliminary 1990 count; results of earlier postal checks and local knowledge in the district offices about where significant new construction had taken place such that the address list creation and updating might need further review; a review of correspondence or media reports that indicated areas or buildings that might have been missed; and a review of the "Were you counted?" calls and forms to look for clusters of unenumerated households. These blocks were systematically canvassed to identify and list missing addresses. The recanvass identified 300,000 housing units as potential adds. Enumerators visited each of these and obtained an interview if the housing unit was in existence on April 1, 1990. LOCAL REVIEW Local review was an attempt to bring to bear local knowledge, records, data, and other types of evidence to improve the accuracy and completeness of the count at small levels of geography. The local review program was designed to provide local officials an opportunity to review aggregate counts and provide supporting evidence while the census field offices were still open and field rechecks were possible. The Census Bureau provided local offices with a booklet explaining the program, a technical guide containing specific guidelines, and detailed census maps showing boundaries for tracts and blocks, so that local officials could compile their own data, especially housing unit estimates, at the block level for comparisons with the counts to be provided by the Bureau. Local offices were asked to provide "hard evidence" of possible discrepancies and enumeration problems at the ED level. Significant discrepancies would then be examined and resolved in a variety of ways including recanvassing. Over the period from August 1987 to September 1988, the Census Bureau offered the first round of training to local officials for the local review program that would take place in 1990. Officials from about 39,000 functioning governmental
OCR for page 228
Modernizing the U.S. Census units in the United States were offered training on how to prepare for and conduct reviews of preliminary housing counts resulting from the 1990 census. In summer 1989, the Census Bureau offered the second round of local review training (similar to the first round described above) to officials from the 39,000 local governments. In fall 1989, the precensus local review program was conducted. This involved about 21,000 local governments within areas covered by the mailout/mailback methodology (the areas where the address list originally was purchased from vendors or compiled from the 1988 prelist). The remainder of the 39,000 local governments were located in areas where update/leave/mailback or list/enumerate would be used and were not eligible for precensus local review because the initial address list was not available. That is, the update/leave/mailback areas used addresses from the 1989 prelist, and this address file still was being processed at the time of precensus local review. For list/enumerate areas, the addresses would not be listed until March when enumerators canvassed those areas for that purpose and to pick up completed questionnaires. For this program, the Census Bureau sent the involved local governments maps and preliminary housing counts developed from the address file so that these could be compared to local information. For those places where the local government identified blocks for which their count differed significantly from the preliminary census count, the Census Bureau sent enumerators to those blocks to conduct a recanvass. In some cases this added new units, but in other cases it only required that the geographic code be corrected (that is, the unit had been assigned to the wrong block in the address file). For units added from this program, questionnaires were labeled for the mailout in late March. In late August 1990, the Census Bureau sent to 39,000 local governments preliminary housing unit and group quarters counts, by block, for the postcensus local review. As with the precensus review, the local governments were asked to report discrepancies between these counts and their local data. The Census Bureau then recanvassed all blocks with significant differences to make sure units had not been missed or geocoded to the wrong block. After unduplicating the list of blocks with the housing coverage check, about 5.5 million units in about 150,000 blocks were recanvassed during this operation. In December, each governmental unit that participated received a listing showing the net change in the housing count for each block identified as a potential problem. POST-ENUMERATION SURVEY At the conclusion of the nonresponse follow-up operation, the Census Bureau also began the Post-Enumeration Survey (PES). This survey was conducted to reenumerate a sample of households so that the results later could be compared to census results to measure undercount and overcount errors in the census. The results of the PES along with data from other research, such as
OCR for page 228
Modernizing the U.S. Census demographic analysis, provided the information needed to measure and evaluate the completeness of coverage of the census and to adjust the census count for the incomplete coverage if so desired. Although officially the census counts were not adjusted, these coverage evaluation measures are used in a number of ways in the postcensal estimates program. NOTES 1 This appendix is largely a rearrangement and extraction of material from Bureau of the Census (1991). See Citro and Cohen (1985: Chap. 3) for a description of the census process for the 1960, 1970, and 1980 censuses. 2 Type 2 district offices covered the balance of mail areas; Type 3 district offices covered the conventional list/enumerate areas. REFERENCES Bureau of the Census 1991 Planning and Conducting the 1990 Decennial Census. June. Bureau of the Census, U.S. Department of Commerce, Washington, D.C. Citro, C.F., and M.L. Cohen, eds. 1985 The Bicentennial Census: New Directions for Methodology in 1990. Panel on Decennial Census Methodology, Committee on National Statistics, National Research Council. Washington, D.C.: National Academy Press.