–2–
Residence Rules: Development and Interpretation

EVERY 10 YEARS, the decennial census data provide the means to assess the size and dynamics of communities and social groups. The data are put to myriad uses every day. However, as complicated as the census is and as varied as its uses are, it is not an exaggeration to say that the census relies fundamentally on the core concept of residence. In the end, each census stands or falls on its ability to gather accurate information on how many people live in particular structures at specific geographic locations; all other questions in the census and the information they elicit are secondary to getting residence information right.

Residence rules form a crucial connective link in the census process, taking the data and attributes of each American resident and producing results that can be tabulated by whatever geographic boundaries may be needed. Residence rules are critical to assigning each person to a “correct” address; a second key linkage—between the address and a specific geographic location—is provided by the Census Bureau’s Master Address File (MAF) and geographic database systems. Properly understood and executed, residence rules provide structure to the highly complex task of census data collection.

The specific residence rules used in the census have changed with time; so, too, has the role of residence rules changed, as the census has come to rely on mail-based, self-administered forms in recent censuses. In this chapter, we discuss the broad context of residence in the census, discussing the nature and scope of residence rules (Sections 2–A and 2–B). We also discuss the



The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
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 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census –2– Residence Rules: Development and Interpretation EVERY 10 YEARS, the decennial census data provide the means to assess the size and dynamics of communities and social groups. The data are put to myriad uses every day. However, as complicated as the census is and as varied as its uses are, it is not an exaggeration to say that the census relies fundamentally on the core concept of residence. In the end, each census stands or falls on its ability to gather accurate information on how many people live in particular structures at specific geographic locations; all other questions in the census and the information they elicit are secondary to getting residence information right. Residence rules form a crucial connective link in the census process, taking the data and attributes of each American resident and producing results that can be tabulated by whatever geographic boundaries may be needed. Residence rules are critical to assigning each person to a “correct” address; a second key linkage—between the address and a specific geographic location—is provided by the Census Bureau’s Master Address File (MAF) and geographic database systems. Properly understood and executed, residence rules provide structure to the highly complex task of census data collection. The specific residence rules used in the census have changed with time; so, too, has the role of residence rules changed, as the census has come to rely on mail-based, self-administered forms in recent censuses. In this chapter, we discuss the broad context of residence in the census, discussing the nature and scope of residence rules (Sections 2–A and 2–B). We also discuss the

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census general problem of why the seemingly simple topic of residence can be so dauntingly complex to both census designers and respondents (Sections 2–C and 2–D). We outline some of the consequences of difficulties in defining residence in Section 2–E, and close in Section 2–F with a description of the Census Bureau’s preliminary residence rules for 2010. 2–A WHY ARE RESIDENCE RULES NEEDED? The explicit constitutional mandate for the decennial census is the generation of counts used to apportion the U.S. House of Representatives. Through this action—assignment of the number of seats in the House and, accordingly, the number of votes in the electoral college for president—representative power is distributed among the states. Hence, the accurate placement of residents by their geographic location (at least to the state level) has always been key to the accuracy of the census. A related use of census data—and arguably as primary a use as apportionment—is the division of states into legislative and voting districts. The use of the census for redistricting greatly heightens the need for accurate links between people and geographic locations: the redrawing of districts demands data at the fine-grained resolution of blocks and tracts. According to McMillen (2000b), in the early days of the country, the division of states into legislative areas (redistricting) was left primarily to the states. In 1842, Congress enacted a law requiring states with more than one legislative district to divide the state into districts with one representative per district. However, nothing was said about district size until the latter half of the 1800s, when geographical compactness and equal population emerged as requirements. In 1929, Congress passed a permanent apportionment act that contained no directives on the geographic size or population of districts, beginning a period of great variability in legislative districts. This period lasted until the U.S. Supreme Court’s ruling in 1962 (Baker v. Carr, 369 U.S. 186) that required legislative districts within states to be drawn to include equal numbers of people (McMillen, 2000b). In the wake of that ruling, and other cases that reinforced the “one person, one vote” standard, legislative redistricting is now based entirely on the most recent decennial census counts, and districts are held to exacting standards of numerical equivalence in population. Since 1965, enforcement of the Voting Rights Act and protection of minority rights have also depended on census counts. Another important use of census counts is in the distribution of funds by the federal government to the states and substate units. In 1998, $185 billion in federal aid was distributed to states and substate areas based in whole or part on census counts (U.S. General Accounting Office, 1999). In all of these uses of census data, the need to count each person once,

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census only once, and in the right place, is crucial. Yet the ideal goal—unequivocally linking each person’s census record to a single specific geographic location—is difficult to achieve in practice. Accordingly, the census relies on a set of residence rules to define what “in the right place” means for various census respondents, so that their census returns can be accurately tabulated. Because they are meant to establish the “correct” location of a person in the census context, residence rules can help alleviate problems of duplication; however, as we will describe, residence rules alone can not solve all problems of census error. 2–B WHAT ARE THE RESIDENCE RULES? As they have developed over time, the residence rules for the decennial census are a formal list of clarifications and interpretations, indicating where people in various residence situations should be counted in the census. In recent censuses, the actual list of residence rules has been an internal Census Bureau document, although a somewhat edited version of the rules was posted on the Census Bureau Web site during the 2000 census.1 The rules were also incorporated in some form into the training materials for census enumerators. The formal residence rule list is used to answer questions, both inside and outside the Census Bureau, on residence questions (U.S. Census Bureau, Population Division, 2004). To understanding residence rules, it is important to remember what the residence rules are not: Most fundamentally, the residence rules are not the specific instructions on the census questionnaire—the guidance to census respondents on who should be included or excluded on the census form, such as those on the 2000 census questionnaire (see Figure 2-1). Confusion on this point may arise because the instructions and questions on the census form are the general public’s primary point of interaction with the census residence rules. However, the instructions are only an extract from the full set of residence rules. Residence rules are not the link between a housing unit and a specific geographic location; rather, they provide the link between an individual person (and data about that person, on the questionnaire) and a specific housing unit. The specific geographic referencing between housing units and geographic locations is done through the Census Bureau’s 1 The presence of these rules online was indicated in a press release detailing the mass mailing of census questionnaires: “a complete set of residency rules telling where students, nursing home residents, military personnel, ‘snowbirds’ and others are counted can be found on the Census Bureau’s Internet site at http://www.census.gov/population/www/censusdata/resid_rules.html” (the link was still functional as of 6/1/06); see U.S. Department of Commerce (2000).

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Figure 2-1 Basic residence question (Question 1), 2000 census questionnaire MAF and geographic reference database, which we discuss in greater detail in Section 8–A. Despite their name, the residence rules are not rules in any legal or regulatory sense. Residence rules are not written into census law; indeed, as we discuss below, not even the term “usual residence” (much less its definition) is written into active census law. Rather, the residence rules are guidelines, internal to the Census Bureau, on how certain living situations should be handled in terms of defining “usual residence.” 2–B.1 Historical Development While the U.S. Constitution specifies that a census be conducted every 10 years for the purpose of reapportioning the U.S. House of Representatives, it offers no further guidance on exactly how the count is to be performed.2 The first U.S. Congress faced the practical problem of performing a count through the Act of March 1, 1790. That act authorized marshals to carry out 2 Clemence (1987:16–17) summarizes the evolution of the Constitution’s census clause from its first draft (apportioning by “the quotas of contribution or the number of free inhabitants, [as] may seem best in different cases”) through several revisions, including one, and only one, instance where “citizens” were distinguished from inhabitants or residents.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census a count such that “every person whose usual place of abode shall be in any family on [August 1, 1790,] shall be returned as of such family.” In addition, “every person, who shall be an inhabitant of any district, but without a settled place of residence, [shall be counted] in that division where he or she shall be on” August 1, and “every person occasionally absent at the time of the enumeration [shall be counted] as belonging to that place in which he usually resides in the United States” (1 Stat. 101, §5).3 The legislative text of 1790 outlined the rules for defining residence for census purposes, with the goal of counting each resident of the United States once and only once and in the correct location; the first census residence rules, like the underlying census goal, have guided every subsequent census. For most people—in 1790, as in 2006—the meaning of “usual residence” is clear.4 These people are affiliated with only one address or household, and have no difficulty identifying it. Indeed, the inaugural census of 1790 presumed that the concept was sufficiently self-evident that the marshals charged with obtaining the counts through personal contact with residents were provided with no written rules or instructions. Yet even in the earliest days of the census, conceptual problems with “usual residence” were evident; as Clemence (1987:18) comments, “the [first] census law contained about 1,600 words and not a single definition.” Moreover, ambiguous residential situations were as plentiful in those days as in modern times. Notably, Clemence (1987:12–14) cites the cases of George Washington and Thomas Jefferson, who served as president and secretary of state during the official period of conduct for the 1790 census (August 2, 1790–April 1791). In those 36 weeks, Clemence observes that Washington “was on the road for 16 weeks, visiting every State in the Union from Rhode Island to Georgia; 15 weeks at the seat of Government, and only 10 weeks at his home in Mount Vernon, Virginia,” yet he was almost certainly counted as head of family at Mount Vernon. Likewise, Jefferson spent most of that period in Philadelphia: “he, like many others, was following the seat of Government around, which had no settled place of residence for itself.” His name appears twice in 1790 census records—once at his Monticello home in Virginia (where he seems to have been tallied) and once in Philadelphia, where both he and attorney general Edmund Randolph signed a census schedule posted on a tavern wall posted (as the law directed) so that people who believed they were not listed at home could still be registered. More generally, Clemence (1987:19) concludes: 3 Records of floor debate from the first Congress suggest that the census legislation was enacted rather swiftly. Though James Madison’s suggestion that residents be listed by occupation was challenged, “no one on the floor spoke a word about place of abode” (Clemence, 1987:18). 4 The schedule used by enumerators in the 1940 census was the first to include the terminology “usual place of residence”; the 1930 and several preceding census continued to use the phrase “place of abode” (Gauthier, 2002; Mills, 1993).

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census In 1790, there were at least a few Americans abroad, including Thomas Jefferson until his summer voyage home from France. There were people in institutions, people with two homes, riverboat captains with no settled place of residence, militia on post duty in the territories outside of any State, and college students attending, for example, Yale, Harvard, Princeton, and William and Mary. The first few decennial censuses did not include definitions or residence rules; indeed, even the exact layout of the schedule used by marshals to conduct the count varied by state. The first set of detailed instructions that touched on residence concepts accompanied the 1850 census.5 The 1850 census instructions were the first to specify residence rules (albeit not formally listed or labeled as such) in order to adapt to changing social conditions. Among the emergent residence rules developed for the 1850 census was a determination on where to count college students (Clemence, 1987:21): Students in colleges, academies, or schools, when absent from the families to which they belong, are to be enumerated only as members of the families6 in which they usually boarded and lodged on [Census Day]. “Because no uniform rule was adopted for the whole United States” regarding the counting of crews of marine vessels, the 1850 instructions continued, “errors necessarily occurred in the last census in enumerating those employed in navigation”; accordingly, the instructions laid out new counting rules that “assistant marshals are required to be particular in following” (Gauthier, 2002:10). Later censuses continued to add rules and instructions for the counting of different groups, typically in response to questions and ambiguities experienced in the field. Though specific residence rules have shifted over the years, as has the date of the census (see Box 2-1), the “usual residence” benchmark inherited from 1790 has generally prevailed as the underlying residence concept for the census. Even though the concept is enshrined in census practice, there exists no definition of “usual residence” in current census law (Title 13 of the U.S. Code). Moreover, Title 13 does not directly specify what residence standard—a de jure enumeration based on usual residence or a de facto count based on current residence or where a person is found on Census Day—should apply to the decennial census. 5 Charged with conduct of the 1820 census during his service as secretary of state, future president John Quincy Adams did provide detailed instructions—and, for the first time, a printed list of questions to be asked by the census—but did not expand on the definition of residence. However, his instructions did ask for information about each person’s settled place of residence and family members temporarily absent; “all of the questions refer to the day when the enumeration is to commence,” and enumerators were cautioned to include family members who had died and exclude babies born after Census Day (Clemence, 1987:21). 6 As discussed in Section 2–C.1, censuses of this period interpreted “family” as any collective of people in a “dwelling house,” without regard to kinship. Hence, the wording of this rule is tantamount to counting students at their college locations.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 2-1 Why Is April 1 “Census Day”? As Anderson (1988:44) notes, “census takers always knew that the count could be affected by the month of the year it was taken.” Early American censuses had to balance the difficulty of making personal contact with residents (slowly, with transportation by foot or horseback) with the prospects of duplicate counting or omissions that would follow from allowing the enumeration to run too long. The first four decennial censuses all used an early August date as the reference, since the summer and fall months were judged to be the best time to find people in what was still an agricultural society. For 1830, “Congress also moved the date of the census ahead two months, to June 1 instead of August [7] as in 1820, on President Adams’s suggestion that this would permit a longer stretch of good weather for the house-to-house enumeration” (Cohen, 2000:121). June 1 remained the census date for the 1840 through 1900 censuses. As the census became more routinized and professional, officials worked to shorten the time of the count and shift the census date earlier in the year. The increasing urbanization of the United States prompted the shift from June 1 to April 15 in the 1910 census; June was “unsatisfactory…because some city dwellers were already out of town for summer vacations, and farmers did not remember enough about the previous year’s crop for the agricultural census” (Anderson, 1988:44). The Census Bureau tried moving Census Day back further still in the 1920 census, to January 1, “at the request of the Department of Agriculture, and also because it was contended that more people would be found at their usual place of abode in January than in April” (Steuart, 1921:571). However, this change “ran into trouble because the winter weather impeded the enumeration and rural leaders complained that many people were working in the city community and hence were not properly counted” (Anderson, 1988:45). After the 1920 experience, census officials were “convinced…that a more nearly perfect and a more rapid count of the people can be made in April than in January.” However—presaging the continuing problem of counting seasonal residents—they acknowledged that this represented a tradeoff (Steuart, 1921:572): It is true that during April and June, when the enumeration has heretofore been in progress, large numbers have been at summer resorts. But at [the January 1920] enumeration it was found that surprisingly high numbers were at winter resorts. Thousands who have their usual places of residence in the northern states spend the winter months in California, Florida, and other southern states. Some of them live in the south several months of each year, and it was difficult to determine their usual places of abode. In this respect the change complicated the work; certainly it did not simplify it. Hence, the 1930 census set April 1 as Census Day, and that date has since been written into Title 13 of the U.S. Code for subsequent censuses. One exception in the 2000 and other recent censuses is the enumeration of remote villages in Alaska, which are rendered unreachable by weather conditions in March and April. In 1930, the count there only began in October; recent censuses have tallied those areas in January or February. 2–B.2 The Changing Role of Residence Rules: From Enumerator Interviews to Self-Response The earliest decennial censuses were conducted by marshals on horseback; though the federal agency charged with conducting the census varied,

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census the method of collecting census information by face-to-face interviewing remained the norm well into the 20th century. The fact that the census was administered in person meant that the field enumerators, ultimately, were responsible for explaining and deciding who should be counted on a “usual residence” standard. This role of enumerator as residence adjudicator was emphasized in the 1880 instructions to enumerators:7 The census law furnishes no definition of the phrase, ‘usual place of abode,’ and it is difficult, under the American system of a protracted enumeration, to afford administrative directions which will wholly obviate the danger that some persons will be reported in two places and others not reported at all. Much must be left to the judgment of the enumerator, who can, if he will take the pains, in the great majority of instances satisfy himself as to the propriety of including or not including doubtful cases in his enumeration of any given family. Arguably, the most significant change in residence rules and their role in the decennial census was brought about by a major paradigm shift in census operations: the switch from an enumerator-conducted census to a mailed-questionnaire, self-administered response model of census data collection. The 1960 census was the first to move significantly toward this model;8 in that year, households were mailed an “Advance Census Report,” which they were asked to fill out but not return by mail. Instead, enumerators visited the household to collect the forms and transcribe the information onto forms more conducive to the optical film reader then used to process census data. If a household did not complete the advance form, the residents were interviewed directly by the enumerator. Subsequently, legislation passed in 1964 (P.L. 88-530) eliminated the requirement that decennial census enumerators personally visit every dwelling place, enabling broader change in census methodology. 7 The enumerator instructions and forms for censuses dating back to 1850 are very helpfully archived as part of the documentation of the Integrated Public Use Microdata Series in the “Counting the Past” section of http://www.ipums.umn.edu/usa/doc.html [8/1/06]. Gauthier (2002) also comprehensively lists enumerator instructions and census schedules from 1790 to 2000. 8 However, the 1960 census was not the first to use the mail in census data collection. The Census Bureau’s procedural history of the 1970 census (U.S. Bureau of the Census, 1970) indicates that mail was used in specialized operations as early as 1890, when questionnaires concerning residential finance were mailed to households with a request for mail return (the same was repeated in 1920, and a similar mail-based program on income and finance was used in 1950). Supplemental information on the blind and deaf was requested by mail in 1910, 1920, and 1930, and an “Absent Family Schedule” was used for some follow-up in 1910, 1930, and 1940. Presaging the 1960 approach, an “Advance Schedule of Population” was delivered to households in 1910; farm households also received an advance copy of an agriculture questionnaire administered as part of that census. Prior to implementation of large-scale mailout/mailback in 1970, experiments and tests of the method were conducted in 1948, 1950 (as a census experiment), 1957, 1958, 1959, 1960 (some of the Advance Census Report responses were requested by mail), 1964, and 1965.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census The 1970 census took the mailout-census model a step forward: census questionnaires were mailed to all households in major urbanized areas (which were thought to include about 60 percent of all housing units), and respondents were asked to mail them back to their local census office on Census Day. This mailout/mailback methodology has been expanded and used in the subsequent censuses; by 2000, 82 percent of the population was covered by mail. The balance of the country was enumerated through a mix of approaches including enumerator visits; Box 2-2 lists the nine “type of enumeration areas” used in 2000. This major change in census data collection techniques dramatically changed the role of residence rules in the census. Prior to the mail-based model, residence rules had been incorporated in the instructions to census enumerators. The Census Bureau’s task in administering residence rules was relatively small: only enumerators had to be trained in the rules, and enumerators could answer questions from respondents, to the best of their ability, during the census interview. But mail-based methods changed the nature of the exercise: the Census Bureau could develop its own interpretation of “usual residence,” but it now had to try to get every census respondent to grasp the Bureau’s definition based only on information included in the census questionnaire.9 Because only so much information can be included on a single piece of paper, the application of residence rules became trickier—trying to find the right combination of words and cues to induce respondents to make their concept of “usual residence” square with the Bureau’s. (We discuss the evolution of the mail-based census instruments in Chapter 6, illustrating the approaches used in the 1960–1990 censuses in Figures 6-1–6-4.) 2–B.3 Assessment of the 2000 Census Residence Rules By 2000, the Census Bureau’s internal list of residence rules grew to 31 specific rules, plus a related statement on the meaning of time cycles (e.g., daily, weekly, or monthly) in determining usual residence. This set of rules is reprinted in Appendix A. The first impression that comes from reviewing the 2000 census residence rules is that they are not organized in a way that a general reader can follow; in large part, this is attributable to the internal (not for public consumption) nature of the full residence rules document. Designed for a more general audience, the residence rules might be structured by major group type (e.g., “students” or “military personnel”) or by the approximate population of the group (so that a living situation in which a person is likely to find oneself comes earlier in the list than rare groups). Instead, the organizational structure of the 9 Of course, the temporary enumerator hired to do field follow-up would also need to develop a solid understanding of the “usual residence” concept; how well training materials achieved that is an open question.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 2-2 Types of Enumeration Areas (TEAs), 2000 Census Mailout/mailback In areas with predominantly city-style addresses, U.S. Postal Service carriers delivered an address-labeled advance letter to every housing unit on the MAF the week of March 6. In mid-March the carriers delivered address-labeled questionnaires, followed 2 weeks later by a reminder postcard. Households were instructed to fill out the questionnaire and mail it back. Update/leave In areas with predominantly rural route and post office box addresses that could not be tied to a specific location, census enumerators dropped off address-labeled questionnaires to housing units in their assignment areas; respondents were instructed to return the completed form by mail. While delivering questionnaires, enumerators updated address entries to include new units not on the list, noting for each its location on a map (map spot), so that follow-up enumerators could find units that did not mail back a questionnaire. List/enumerate In remote, sparsely populated, and hard-to-visit areas, census enumerators combined address listing and enumeration. There was no MAF for these areas created in advance. The enumerators searched for housing units, listed each unit in an address register (also its map spot), and enumerated the household at the same time. Remote Alaska The enumeration procedure in remote areas of Alaska was similar to list/enumerate. It was conducted in February, before ice break-up and snow melt. Rural update/enumerate The Census Bureau determined that some blocks originally planned to be enumerated by update/leave would be better handled by a procedure in which address list updating and enumeration were conducted concurrently. Areas covered by the operation included resort areas (believed to have high concentrations of seasonally vacant housing units), some American Indian reservations, and colonias on the U.S.-Mexico border. “Rural” refers to the source of the address list, which were operations focused on areas with mainly non-city-style addresses. Military Mailout/mailback procedures were used for all residential blocks on military bases (excluding group quarters). Such blocks in type 2 enumeration areas (but not those in type 1 enumeration areas) were assigned an enumeration area code of 6 because there was no need to update the address list or provide map spots. “Urban” update/leave It was determined that some blocks originally planned to have questionnaire delivery by the Postal Service would be better handled by having census enumerators follow an update/leave procedure. Such blocks contained older apartment buildings that lacked clear apartment unit designators, or they had many residents, despite having city-style addresses, who received their mail at post office boxes. “Urban” update/enumerate Some American Indian reservations contained blocks in more than one TEA. In these instances, all blocks in the reservation were enumerated using update/enumerate methods (type 5). However, those blocks for which the mailing list was developed using “urban” procedures and for which no map spotting was required were made type 8. Mailout/mailback conversion to update/leave Late reexamination of planned TEA 1 areas, conducted in 1999, suggested that some blocks contained a significant number of non-city-style addresses. These were converted to update/leave but treated as a separate type. SOURCE: Adapted from National Research Council (2004c:Box C.2).

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census 2000 census residence rules is both weak and operational in nature. Its major headings are obscure—“household population,” “group quarters population, [usual home elsewhere (UHE)] allowed,” “group quarters population, UHE not allowed,” “overseas population,” and “do not list population”—and use jargon (e.g., “group quarters” and “do not list”) and acronyms (UHE) that are opaque to a lay audience. Further, major population groups are not addressed coherently. The rules for counting college students are dispersed into rules 5, 6, and 25 (although that group of rules does, helpfully, provide explicit directions to look at the other related rules); newborn babies are mentioned almost as an afterthought in rule 3 and lost in technical detail of counting hospital patients; military personnel are divided across rules 4, 13, 26, and 27. As we discuss further in Chapter 6, census residence rules serve several different purposes, and would thus be better handled by crafting different products to fill those various purposes. The 2000 census residence rules document takes an omnibus approach and tries to satisfy all the needs. By straining too hard to be an operational blueprint (e.g., categorizing by “household” versus “group quarters” operations, complete with specific group quarter code numbers in rules 13–19), the document becomes less effective in its primary purpose of clarifying the meaning of “usual residence.” Finding 2.1: As developed and used in the 2000 census, the residence rules for the decennial census were too complicated and difficult to communicate. The set of 31 formal residence rules was not organized for ease in comprehension, and instead seemed to be a loose amalgamation of previously encountered problematic residence situations. The sheer number and redundancy of the rules detract from their effectiveness in training temporary census enumerators. 2–C WHY IS MEASURING RESIDENCE DIFFICULT FOR THE CENSUS BUREAU? In this and the next section we discuss some of the difficulties associated with measuring residence, and “usual residence” in particular, from two basic viewpoints. In this section we focus on the challenges faced by the Census Bureau in specifying what it means by residence; in Section 2–D, we consider the difficulties faced by respondents in answering residence questions. 2–C.1 Definitional Challenges An inherent problem with the Census Bureau’s “usual residence” approach—particularly when the primary mode of data collection is self-response by individual persons—is that it requires respondents to interpret

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 2-6 Undercount and Overcount in the 2000 Census Prior to the April 2001 deadline for delivery of redistricting data, the Census Bureau’s Executive Steering Committee for A.C.E. Policy (ESCAP) weighed a set of highly discrepant results. Relative to the census count of 281.4 million, estimates derived using the A.C.E. suggested a 1.18 percent net undercount at the national level (an improvement from the estimated 1.61 percent net undercount in the 1990 census). Moreover, the A.C.E. estimates suggested a reduction in net undercount among the black and Hispanic populations, from 4.57 percent and 4.99 percent in 1990 to 2.17 and 2.85 percent in 2000, respectively. However, results from demographic analysis (DA)—essentially, adding estimates of births and immigrants to the previous census count and subtracting estimated deaths and emigrants—told a different story. The hard-to-estimate size of the illegal immigrant population is important to a DA count; depending on which estimate was used for this subgroup, DA suggested either a much smaller national net undercount (0.32 percent) or an unexpected net overcount (−0.65 percent). In light of these discrepancies, ESCAP concluded that there might be a significant flaw in the A.C.E. methodology; it directed a program of further research and recommended that the census data not be adjusted, a recommendation later approved by the commerce secretary. Further refinement of the A.C.E. methodology was conducted during summer 2001, in advance of an October decision on whether adjusted census data should be used as the basis for census totals used for such purposes as fund allocation. The new studies suggested that use of the A.C.E. to adjust census results would overstate the population because the census itself had more errors of erroneous enumeration (overcount) than originally indicated. A careful evalution follow-up study, combined with methods for detecting duplicates by matching by name and date of birth (discussed further in Box 8-1), found an estimated 2.9 million erroneous enumerations that were not discovered in the original A.C.E. Bureau analysts and a National Research Council (2004c:180) panel conjectured that problems in defining census residence might explain many of these duplicates: cases seemed to include college students double-counted at their parental homes and at school, children in joint custody counted at the homes of both parents, and seasonal residents with more than one house. The Bureau once again recommended against statistical adjustment of the 2000 census data for fund allocation and other purposes. For a third and final time, the Bureau recommended against adjustment in March 2003; this time, the decision involved whether adjusted census data would be used as the basis for postcensal population estimates. Between October 2001 and March 2003, Census Bureau staff engaged in a further major set of evaluation studies, producing a final set of results dubbed A.C.E. Revision II; these results indicated a 0.5 percent net overcount of the population. Final DA results suggested a very small, 0.1 percent net undercount. SOURCE: National Research Council (2004c:Tables 5.1, 5.2). appear to exist between persons counted in both a group quarters and a regular housing unit, sharp duplication rate peaks occur in the college-age population and for people starting around age 70, suggesting the importance of college students and hospital and nursing home patients as potential sources of duplication. Analyses of the 1990 and 2000 censuses, as well as their predecessors, suggest that census omissions—the undercount—are heavily concen-

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census trated among young black males, children under 10, people living in homes where they are unrelated to others, movers, and those who rent rather than own their homes (National Research Council, 2004c). Census residence rules that are clearly conveyed are a key part of a strategy to combat person duplication; given the prominence of groups like college students and nursing home patients as potential census duplicates, a more effective way of ensuring that these groups are counted at one place is certainly needed. But residence rules alone can not solve the entire problem of census omission and unduplication. As was also learned in 2000, even census programs designed to counter census errors can sometimes serve to exacerbate them. For example, the coverage edit follow-up (CEFU) operation both added and subtracted people from the census. This telephone follow-up operation was intended to collect data on people in large households (those that indicated more than 6 people living there, even though the census form only allowed for collection of data on 6 residents) and to resolve count discrepancies between the reported household population count and the actual number of data-defined18 people recorded on the census form. The operation was applied to all mail returns, including both census short and long forms, as well as certain Be Counted19 forms and Internet data collection responses processed before June 8, 2000. About half of all CEFU cases were not completed for various reasons, including the lack of a telephone number, but some decision had to be made on the household size for all of them. Since 6.5 percent of the cases that were completed were found to be duplicates, the failure to complete CEFU for all cases (sticking with the original determined count) undoubtedly added some people to the count who were duplicates (Sheppard, 2003). As our predecessor census panels have noted (National Research Council, 2004c:Finding 1.10), the Census Bureau’s coverage evaluation research based on the 2000 A.C.E. is insightful and of high quality, and the subsequent work that has been done on the complete matching of census records against themselves is similarly commendable. Although the work has yielded some better glimpse at the nature of duplication in the census, it does not speak as strongly to the nature of census omission. As the next census draws near and becomes the Census Bureau’s overriding operational goal, it can be difficult to devote resources to continuing to mine the data from the previous census. However, we believe that a continuous process of developing hypotheses from those data and using those lessons for future census planning is absolutely essential. 18 A household is data-defined if at least one member has reported values for at least two complete-count items, including name. A data-defined person has at least two complete-count items, including name. 19 Be Counted forms were unaddressed census short-form questionnaires that could be picked up in local post offices or other locations, and were intended to be returned by people who felt that they had otherwise been missed in the census. The 2000 “Be Counted” program was the successor to the similar “Were You Counted?” program in 1990.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census In that spirit, we suggest that evaluation of the 2000 census not be considered to be a completed process. Finding 2.2: Although the research done to date does provide some information on the nature of omissions and duplicates in the 2000 census, the analyses are not sufficient to fully sort out important effects, and the data that have been collected need further analysis. We note that the person-matching routines face a significant data limitation, which is the lack of coded information on “any residence elsewhere” reported by census respondents. Combining name and date of birth is certainly better than matching on name alone, but very common names (e.g., “Bob Smith”) will still create matching difficulties. Information on “any residence elsewhere” could augment search capabilities by refining the geographic scope. We discuss this point further in Section 8–B. 2–E.2 Group Quarters Enumeration The group quarters population is a particularly challenging one for census residence rules. Difficult as the concept may be to work with, though, it is important to keep the general nature of group quarters data collection in mind when thinking of the consequence of definitional and operational aspects of census residence rules. It is important for census residence concepts to deal with group quarters as accurately as possible because the decennial census has historically been the only comprehensive data source on characteristics of the group quarters population. Regardless of its flaws, the decennial census serves as the best—and sometimes the only—window on this small (2.7 percent) but significant part of the population. The Census Bureau’s fiscal 2006 budget request included funds that would add group quarters into data collection for the American Community Survey (ACS). The quality of group quarters data collection is uncertain—whether improved group quarters definitions and more highly skilled ACS interviewers will be able to offset the disturbingly high missing data rate (and, correspondingly, the rate with which those missing data had to be imputed) for many census long-form data items in 2000 (National Research Council, 2004c). Other federal and private surveys probe parts of the group quarters problem, but the decennial census remains unique in its comprehensive nature. As in previous years, the 2010 census will be examined as a source of benchmark data on the size, growth, and nature of the population living in prisons, college dormitories, nursing homes, and other group quarters.20 20 That said, the ability to measure change from the previous census may be affected by changes to the definitions of group quarters; see Chapter 7.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census 2–F PLANS FOR 2010 2–F.1 One Rule: Proposed Residence Rules Revision In preparation for the 2010 census, the Census Bureau established an internal working group to refine residence rules and concepts, with the objective of testing revisions in the 2006 census test and the 2008 dress rehearsal, and, ultimately, using them in the 2010 census. This panel’s work is intended in part to review this initial work and provide guidance. One of the panel’s regular meetings in December 2004 was devoted almost exclusively to a comprehensive walk-through of the residence rules of the 2000 census. Census Bureau staff briefly reviewed historical highlights of each of the 31 formal rules; “straw man” suggestions for changes to the rules were advanced, and the ensuing discussion provided both constructive criticism of the existing rules as well as possible directions for improvement. In particular, the public discussion at that meeting yielded in raw form the basic argument that we underscore and elaborate on in this report—namely, that there is a need for a clearly articulated set of residence principles, and that it is from these principles that other products (including analogues of the current census residence rules) should be developed. Based on that feedback, the Census Bureau residence rules staff continued work in advance of a March 2005 meeting with the panel. At that meeting, the Census Bureau presented a draft recommendation that replaced the 31 formal rules of 2000 with a single residence rule for 2010; a supporting document, similar to the 2000 census residence rules list, described how this single residence rule should be applied in a variety of living situations. This “one rule” approach and the major proposed differences are summarized in Box 2-7. To its credit, the Census Bureau has also conducted parallel work on redefining group quarters, with the objectives of creating an integrated address file (rather than a separate MAF and group quarters roster) and of testing revised definitions in 2006 and 2008. We discuss the group quarters redefinition efforts in Section 7–B and Box 7-1. 2–F.2 Assessment The Census Bureau’s progress to date in revising the census residence rules has been highly commendable. With its draft revision, the Bureau has shown a willingness to make broad changes; we encourage this reconsideration of core census concepts and urge that old conventions be substantially revised, if not fully for 2010 then for future censuses. In this chapter, we have reviewed the nature of census residence rules and the general issues of residence in the context of the decennial census. The chapters that follow expand our comments and present our recommendations

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 2-7 Census Bureau’s Proposed 2010 Census Residence Rule The 2010 Census Residence Rule The 2010 Census residence rule is used for determining where people should be counted (which means tabulated) in the census. Residence Rule: Count people at their usual residence, which is the place where they live and sleep most of the time. People in certain types of group quarters (GQs) on Census Day should be counted at the GQ. These GQ types are listed in the box below. People who do not have a usual residence or cannot determine a usual residence, and who are not in one of the GQs types listed below, should be counted where they are on Census Day. (As of the Census Bureau’s March 15, 2005, draft, the specific “Group Quarters Where People Are Counted at the Group Quarters” had not yet been determined.) Proposed Changes to Residence Situation Applications The Census Bureau’s proposed “one rule” is accompanied by a listing of how the rule applies to a variety of residence situations. Most of these are adapted from the 2000 census residence rules, with one change and several formal additions: Boarding school students: The Census Bureau proposes changing the interpretation of “usual residence” for boarding school students (below the college level) to the “residence at the boarding school where they live and sleep most of the time”; previously, it had been the parental home. Births and deaths on Census Day: The Census Bureau proposes adding three interpretive statements. “Babies born on or before 11:59:59 p.m. on Census Day” are to be counted at the “residence where they will live and sleep most of the time.” Both “babies born after 11:59:59 p.m. on Census Day” and “people who die before 12:00:00 a.m. on Census Day” are not counted in the census. Movers on Census Day: The Census Bureau proposes that “People who move into a residence on Census Day who have not been listed on a questionnaire for any residence” should be counted at the residence they move into on Census Day; “People who move out of a residence on Census Day and have not moved into a new residence on Census Day who have not been listed on a questionnaire for any residence” should be counted at the residence they move out of on Census Day; and “People who move out of a residence or move into a residence on Census Day who have been listed on a questionnaire for any residence” are not to be counted again.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census on specific issues, but we first end this chapter with our assessment at the macro level. First is a basic but surprising fact arising from examining the legal mandates of the census. Finding 2.3: Though the concept inherits from a long tradition of practice dating to the 1790 census act, active census law and regulation do not define the residence standard for the decennial census (de jure or de facto), nor do they define what constitutes “usual residence.” To be clear, the panel believes that it is very much for the best that Title 13 is not highly prescriptive of the exact mechanics of the decennial census; the open-endedness gives the Census Bureau much-needed latitude to develop and continue to refine its craft. Nor do we relish or advise a full congressional review and reopening of Title 13. What we do find is that, in contrast with the enabling laws for international censuses (see Appendix B), it is somewhat unusual for core census residence concepts in the United States to be as “hidden” as they have been in the past, deriving from the spirit of a centuries-old statute and kept in full form as a purely internal Census Bureau document. Accordingly, we suggest that census residence rules and concepts be made more transparent. Approaches to promote the transparency of census residence concepts to the public and decision makers could include posting notice of, and inviting comment on, residence standards through promulgation in the Federal Register, as the Bureau does with other basic operational plans. Information on how the Bureau defines residence should also continue to be posted to the agency’s Web site. The second point of the assessment follows from the first: the base residence standard of the decennial census—whether it is a de jure type or a de facto type—is not explicitly written into law. As a result, an evaluation of census residence rules necessarily prompts some consideration of this fundamental question. Since its inception, the U.S. census has followed a de jure-type model with its “usual residence” standard; we believe that it is important to consider whether a change in the residential standard—to more of a de facto orientation—is warranted. Moreover, it is very appropriate that the choice of residence standard continue to be periodically revisited in ensuing decades. The choice of a de jure or de facto approach is particularly salient for the U.S. census because of the adoption of the ACS as a replacement for the census long form. While the decennial census follows a “usual residence” standard that can be considered a de jure style, the ACS relies on “current residence”—defined by a “two-month rule”—as its benchmark, establishing it as a de facto type. Having just begun full-scale data collection in 2005, the ACS is still 2 years away from producing a steady stream of estimates for many geographic

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census areas. When that production begins, however, the Census Bureau will be in the position of having two related flagship products—the decennial census and the ACS—follow two very discrepant residence models. How this discrepancy will affect estimates of population characteristics, and the extent of the gap that may arise between ACS and census estimates in decennial census years, is unknown. However, the coexistence of the census and the ACS makes the choice of residence standard as an open and important issue; we discuss the ACS and the census residence standards in detail in Chapter 8. Both the de jure and de facto ideals have strengths and weaknesses. One is not clearly superior to the other; rather, the choice between which standard to follow depends critically on the uses of the resulting data. The case can be made that a pure de facto count is the least ambiguous concept for respondents to understand, and it may be the simplest to describe: “Who is here right now?” As a result, “current residence” models based on a de facto approach are generally considered easier to explain and implement; they have become the standard for many surveys and polls, as their point-in-time snapshot orientation makes them well suited for measuring the characteristics of a population (rather than an absolute count). However, for purposes of reallocating political representation based on population counts, the emphasis in a pure de facto model on counting people exactly where they are found would emphasize people’s stays in situations that are temporary or transitory: away on trips, in hotels or motels or staying as houseguests, or in transit at the time of the count. As a result, a de facto approach would arguably be less appropriate than a de jure-type rule that counts people where they belong, in some legal or other sense. Some nations strive for the de jure ideal by maintaining population registers or using counts derived from administrative records in lieu of a census. Appendix B summarizes residence concepts used in several foreign nations that—like the United States—conduct regular population censuses rather than compile records-based counts. Of these countries, most adhere to a de jure-type concept: Austria, Canada, Finland, Israel, Italy, Portugal, Switzerland, and the United Kingdom (the United Kingdom switched to a de jure count in 2001 after using a de facto standard since 1801). However, that choice is not universal. Australia, Estonia, Ireland, New Zealand, and South Africa use a strict de facto standard; Japan uses a 3-month hybrid rule akin to the ACS rule. Several nations, among them Japan and New Zealand, ask both usual residence and current residence questions; perhaps more interesting are cases like Austria and Finland, which ask their respondents to identify their “main” or “permanent” residence but also ask them for a “secondary” or “temporary” residence if one exists. Likewise, de facto-type rules are common in general surveys and polls, but some major federal household surveys vary and follow a usual residence rule: prominent among these are the Census Bureauconducted Current Population Survey and the Survey of Income and Program

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Table 2-1 Residence Rules for the Current Population Survey Living Situation Include in Household? A. PERSONS STAYING IN SAMPLE UNIT AT TIMtE OF INTERVIEW Person is member of family, lodger, servant, visitor, etc. Ordinarily stays here all the time (sleeps here) Yes Here temporarily—no living quarters held for person elsewhere Yes Here temporarily—living quarters held for person elsewhere No Person is in Armed Forces Stationed in this locality, usually sleeps here Yes Temporarily here on leave—stationed elsewhere No Person is a student—Here temporarily attending school—living quarters held for person elsewhere Not married or not living with immediate family No Married and living with immediate family Yes Student nurse living at school Yes B. ABSENT PERSON WHO USUALLY LIVES HERE IN SAMPLE UNIT Person is inmate of institutional special place—Absent because inmate in a specified institution regardless of whether or not living quarters held for person here No Person is temporarily absent on vacation, in general hospital, etc. [including veterans’ facilities]—Living quarters held here for person Yes Person is absent in connection with job Living quarters held here for person—temporarily absent while “on the road” in connection with job (e.g., traveling salesperson, railroad conductor, bus driver) Yes Living quarters held here and elsewhere for person but comes here infrequently (e.g., construction engineer) No Living quarters held here at home for unmarried college student working away from home during summer school vacation Yes Person is in Armed Forces—was member of this household at time of induction but currently stationed elsewhere No Person is a student in school—away temporarily attending school—living quarters held for person here Not married or not living with immediate family Yes Married and living with immediate family No Attending school overseas No Student nurse living at school No C. EXCEPTIONS AND DOUBTFUL CASES Person with two concurrent residences—determine length of time person has maintained two concurrent residences Has slept greater part of that time in another locality No Has slept greater part of that time in sample unit Yes Citizen of foreign country temporarily in the U.S. Living on premises of an embassy, ministry, legation, chancellery, or consulate No Not living on premises of an embassy, ministry, etc. Living here and no usual place of residence elsewhere in the U.S. Yes Visiting or traveling in the U.S. No SOURCE: U.S. Census Bureau (2002:Figure 7-5).

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Participation (see Table 2-1 for a summary of the Current Population Survey’s residence rules). As discussed above, data from the U.S. decennial census have several principal uses, and it is relative to those uses that the adequacy of either a de jure or a de facto approach must be assessed. If the principal use of the decennial census data was the allocation of federal and state funds, a de facto-type approach emphasizing current physical presence (and hence demand on resources) in an area would likely be preferable as a base for postcensal population estimates. However, the co-primary uses of census data—the reallocation of political power through reapportionment and redistricting—argue for a de jure approach that puts less weight on short-term or transitory residence situations. In short, we know of no pressing reason for a major change in the residence standard for the decennial census from the de jure-based “usual residence” concept to a de facto model. The balance of this report offers recommendations predicated on the assumption that a de jure approach will continue to be the standard. However, we reiterate that this basic question is one that should be revisited from time to time, particularly following further experience with the ACS’s “current residence” standard and the completion of research on respondent difficulty in interpreting the two types of residence standards; we discuss both of these items in more detail in Chapter 8. Third, we offer a recommendation that underlies the balance of the report. Recommendation 2.1: The residence rules for the 2010 and future censuses should be substantially rewritten (relative to those used in 2000), and the Census Bureau should make a concerted effort in 2010 to improve the communication of residence rules. Core concepts should be expressed as a small number of concise residence principles. These residence principles should then be used to develop other products, such as any instructions or cues to respondents on the census questionnaire, training materials for enumerators, census processing and editing routines, and a “frequently asked questions” list for enumerator and respondent reference and posting on the Internet. In offering this recommendation, we reiterate that the Census Bureau has made some good first strides in this direction with its proposed revisions for the 2006 census test and 2008 dress rehearsal. We applaud and endorse the Bureau’s continued effort, and urge that further work be given high priority commensurate with the importance of residence concepts to the accuracy of the census. We also wish to acknowledge and emphasize, in this context, a tension that will recur throughout our detailed discussion in later chapters. Some principles, and strict application of principles, will necessarily run counter to

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census the most optimal or cost-effective census operations, and some may be extremely difficult to implement in practice. Striking a balance between pure principle and effective operations may not be easy, but the accuracy of the census will be improved by the development of basic residence principles and their use in developing implementation plans.

OCR for page 23
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census This page intentionally left blank.