Part III
Improvements for the Future



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 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Part III Improvements for the Future

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census This page intentionally left blank.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census –6– Residence Principles for the Decennial Census CHANGE DOES NOT COME EASILY to the decennial census; it is too large and intricate an operation for massive overhauls of operations or procedures to be feasible in a short amount of time. The quality of the resultant data is paramount, and so implementing procedures that have not been tested is inadvisable. A discussant at the 1986 Council of Professional Associations on Federal Statistics (COPAFS) conference on residence rules noted similar comments on the nature of change in the census in one of the presented papers and wondered if the statement was mildly euphemistic, whether “this is perhaps another way of saying that we are unlikely to do much to make any needed improvements in time for the 1990 census, since we are beginning to think seriously about the problem in December of 1986” (Sweet, 1987:6). Twenty years later, with the 2010 census looming in the not-too-distant future and even with the benefit of somewhat better lead time, we face something of the same problem. In our case, there are very promising signs of improvements to the collection of residence information in 2010. The advent of the American Community Survey and, with it, the narrowing of focus on a short-form-only decennial census is a considerable simplification and has permitted earlier attention to residence considerations than in the past. As we have noted, residence concepts were a primary focus of the 2005 census test and will be the topic of a further mailout experiment in 2006. Also, as witnessed in our panel’s public meetings, the Census Bureau has made good strides in redrafting and revising both the census residence rules and the definitions of group quarters.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census We conclude that the Census Bureau would be best served by some radical revamping of its basic approaches to collecting residence information. It is unrealistic to expect large-scale changes in the 2010 census. Rather, our intent in this report is to provide a mix of short-term and long-term guidance: short-term proposals that can be implemented in 2010 and long-term research topics (to be informed by tests conducted in 2010) to provide a basis for decisions in 2020 and beyond. In this chapter we discuss the basic nature of residence rules and argue for a principle-based approach to defining residence. A core set of residence principles should be used to develop related products and operations (Section 6–A). The second half of this chapter discusses the implications for the most important such product, the census questionnaire itself and the specific residence instructions in it (Sections 6–C and 6–D). In discussing the general presentation of residence concepts to census respondents, we suggest the need for a major change in the way residence information is gathered, switching from an instruction-based to a question-based approach (6–E). In Section 6–F, we discuss a particular problem in communicating with census respondents, encouraging prompt replies to the census form while still preserving the meaning of Census Day. We close in Section 6–G with recommendations for research on presentation issues. 6–A A CORE SET OF PRINCIPLES As discussed throughout this report, the residence rules for the decennial census must satisfy several needs simultaneously. They have primarily been regarded as an internal Census Bureau reference, but also must be adaptable to the construction and phrasing of items on the census form and must govern the design of related census operations. They must also be a resource for the training of temporary census enumerators and a reference to census staff who must field questions from census respondents. As we noted in Chapter 2 and Recommendation 2.1, the Census Bureau’s current approach of trying to serve all these needs with a single document is seriously flawed. The 2000 census rules were so long and intricate that it was difficult to discern their meaning; with effort, one can intuit some of the logic that guided the construction of the rules, but the task is difficult. The situation needs to be reversed: a concise core set of residence principles should be developed, and all the related products and extracts—question wording and structure, enumerator training materials, and so forth—should be built using the principles as a base. At the 1986 COPAFS residence rules conference, Lowry (1987:30–31) made an early call for the reduction of formal census residence rules to a set of basic principles. He argued that the Bureau would be better served by hav-

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census ing the residence rules (for the 1990 census) “express general principles rather than merely treating the most common examples of residential ambiguity.” Specifically, he proposed six principles: If respondent lived for more than half of the preceding year at an address that he occupies on [Census Day] or expects to return to in the following year, that address is his usual residence. If respondent lived for more than half of the preceding year at an address that he does not expect to return to in the coming year, his usual residence is the address he now occupies or expects to occupy for most of the coming year. If respondent has several homes, no one of which he occupied for more than half of last year, he should select one as his usual residence. If an editor must choose, choose the current residence. If the respondent has a home to which he returns in the intervals between traveling, that home is his usual residence even though he may have been elsewhere for more than half of the preceding year. If the respondent says he has no usual residence or gives no address for his usual residence, he should be enumerated as a transient at the place where he was staying when contacted. If the respondent says his usual residence is elsewhere but gives an incomplete, wrong, or invalid address for that residence, he should be enumerated as a resident of the smallest geographic unit that is unambiguously codable from his response—census block, enumeration district, tract, city, county, or state. Discussants at the workshop found these principles to be generally reasonable; the workshop summary concluded that “a system that relies on well-defined principles, yet can incorporate responses from people in a wide variety of living [situations,] should accommodate appropriate decisions about just where to count each individual as of April 1st” (CEC Associates, 1987:6). The Lowry (1987) principles have attractive features: the third principle subtly suggests an approach of collecting data on multiple residences and putting the onus of determining the usual residence on the Census Bureau, not the respondent. The sixth principle is intriguing, with its bold approach of associating people only with large geographic areas (e.g., cities, counties, or states) if that is all that can be validated from their provided residence information; how these larger-area counts would be distributed in block-level redistricting counts is not specified and would surely be contentious. A strength of the Lowry (1987) principles is their anchoring to a reference period of 6 months, interpreting either a retrospective (first principle) or prospective (second principle) stay of the majority of a year as the basic definition of usual residence. But that specificity is also a limitation, since 6 months is a coarse time interval with respect to some living situations; the third and fourth principles weaken the hard-line 6-months criterion, imparting some ambiguity in the determination of residence.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census We conclude that the census would benefit from the specification of a core set of residence principles and retain some of the desirable features of the Lowry (1987) exemplars, but take a somewhat different approach in our recommendation. A key component of a set of residence principles for a decennial census is an indication of the general residence standard to be pursued. As we state in Section 2–F.2, we know of no pressing reason why the U.S. census should shift away from the de jure/“usual residence” approach as its basic standard (though we also believe that it is useful that this decision be periodically revisited). So, in suggesting a candidate set of residence principles, we support a system that seeks a de jure count; the de facto/current residence location should be used as a tie-breaker in cases where no usual residence can be determined. Assuming a general de jure model, and taking as given the objective of the census as a resident count (rather than a citizen count), we recommend adoption of a set of principles based on the following: Recommendation 6.1: Suggested Statement of Residence Principles: The fundamental purpose of the census is to count all persons whose usual residence is in the United States and its territories on Census Day. All persons living in the United States, including non-U.S. citizens, should be counted at their usual residence. Usual residence is the place where they live or sleep more than any other place. Determination of usual residence should be made at the level of the individual person, and not by virtue of family relationship or type of residence. If a person has strong ties to more than one residence, the Census Bureau should collect that information on the census form and subsequently attempt to resolve what constitutes the “usual residence.” If a usual place of residence cannot be determined, persons should be counted where they are on Census Day. Unlike the Lowry (1987) principles, we do not specify a fixed reference period in our first principle, the basic definition of usual residence. In our deliberations, we found that the varied living situations that must be accounted for in the census do not lend themselves easily to any choice of a time window we could construct. Ultimately, we find that a definition of usual residence must necessarily be somewhat ambiguous in order to be most broadly applicable. In our first principle, we use the wording “more than any other place” rather than other language (like “most of the time”) because we favor use of a plurality rule and because we find it less ambiguous than “most of the time.”

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Even so, ambiguity remains about the relevant time frame against which “more than any other place” is to be assessed; in addition, the issue of whether any element of prospective living arrangements should be considered remains unspoken (e.g., new movers, who intend to live at this location more than any other place but have not been living there for very long). Some of the ongoing research we recommend in the remainder of this chapter should be brought to bear on refining this principle—it should, for instance, consider whether including a specific reference period has a significant impact on response or respondent confusion. In the interim, our definition of usual residence strives to reduce some ambiguity but necessarily leaves some issues open. The second principle asserts that usual residence should be viewed as an individual-level attribute, not one that is tied to family relationships or type of residence. This principle is included to be consistent with changes that we recommend in enumerating what has traditionally been treated as the group quarters population. As we argue in more detail in Chapter 7 (and as follows from descriptions of some group quarters types like health care facilities and jails—Sections 3–C and 3–D.2), the concept of group quarters enumeration requires comprehensive reexamination. The expected length of stay and actual living situations in some group quarters types is such that it is inappropriate to characterize residence status for all facility residents based solely on the facility’s name or type. Including the provision that usual residence does not depend on family relationship speaks to situations such as college students living away or children in joint custody. This is a principle that some respondents may continue to violate, given the strength of enduring ties, but we believe that it is useful to have this concept expressed as a principle rather than something to be inferred from a long list of examples. The third principle foreshadows arguments that we will make later in this chapter. We favor an approach in which the census form asks a sufficient number of questions to get a sense of each person’s residential situation. Significantly, we believe that it is important that the census move toward collection of information on any other residence that a person may be affiliated with. Our ultimate vision is of a census in which residence information is collected without burdening respondents with the problem of deciding who is usually resident at their address; those kinds of determinations should be made by the Census Bureau during processing of the forms, based on the data provided by respondents to a series of residence questions. Finally, the fourth principle is a “tie-breaker” rule. There are some living situations that do not lend themselves to an unequivocal determination of a place where a person lives or sleeps more than any other; examples include dedicated recreational vehicle users and children in joint custody situations who spend equal time with both parents. When a usual residence cannot be determined, the person’s location on Census Day—the de facto residence location—should be used.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Of course, these are not the only principles that could be developed for the census, but we believe them to be an adequate set. Other possibilities include something akin to Lowry’s sixth principle, which raises the question of the level of geographic resolution needed for tabulation; that principle would assign people to the smallest possible level of geography (but not necessarily to a specific geographic coordinate). We revisit the question in Chapter 7 and the counting of the group quarters/nonhousehold population. Another possible principle—perhaps useful for handling some ambiguous residence situations but that may be prohibitively difficult operationally—is to specify that children under a certain age must be counted at the home of their parent or guardian. Though it would provide an alternate solution to the current mismatch between the counting of boarding school and college students and could be more consistent with a “family” interpretation of household, the specification of the age cutoff could make enumeration of college students even more difficult.1 Collectively, our four suggested principles imply the exclusion of American citizens (nonmilitary and nongovernment employee) living overseas, consistent with practice in recent censuses. To satisfy the requirements of current law and court precedent, military personnel and federal civilian personnel stationed overseas, and their dependents, would continue to be assigned to their home states of record for purposes of apportionment only. 6–B PRODUCTS FOR IMPLEMENTATION OF THE PRINCIPLES Articulation of a core set of principles provides a basis from which to work in developing other products, one of which is an explanation of how the principles apply to a variety of living situations. The 2000 census model of trying to craft rules to match all possible living situations is flawed because it lacks a unifying conceptual basis. Yet there is still a definite need for something analogous to the intent of the 31 rules—a list of examples of how ambiguous living situations should be resolved—but one that is grounded in concept and better structured for comprehension. Such a listing is needed for several reasons and audiences. Though field enumerators should be made familiar with the basic residence principles, it is also useful for them to be aware of concrete examples of the application of the principles to living situations that they are likely to encounter. Similarly, staff who deliver questionnaire assistance by telephone should also have common situations and their treatment as a reference. We suggest in Section 2–F.2 that census residence rules be made more transparent to the public and to decisionmakers, and public posting of examples of the application of chosen 1 Such an age-based rule would also misplace minors who have won legal emancipation from their parents.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census residence principles to various living situations is an important part of that effort. A good way to meet this need would be with a “frequently asked questions” (FAQ) list, as has become common on Web sites. Residence situations and their treatment under the residence principles should be grouped thematically for easier use. Box 6-1 illustrates a possible mapping of our recommended principles to residence situations. This structure can certainly be improved—particularly for a version to be posted on the Internet—but it provides a basic starting point for development of an FAQ list and other census documents. The location at which prisoners should be counted has become a major point of contention. Our suggested principles—and our interpretation of them, in Box 6-1—lead us to side with the Census Bureau’s current general procedure of counting them at the location of incarceration. At this time, not enough is known about the exact nature of the alternative of counting prisoners at a place other than the prison, nor about the accuracy and consistency of facility data on inmate residence, to recommend change in counting prisoners in the 2010 census. However, we strongly urge that the 2010 census include a major test on the collection of additional residence information from prisoners and further assessment of the quality of administrative records that could inform future reconsideration of the prisoner counting issue. We discuss these initiatives below and in Chapter 7. The residence principles should also be used to generate residence-based products in addition to the FAQ list of applications. Key among these are any specific instructions or other cues included on the census questionnaire itself; we discuss this in greater detail in the balance of this chapter. The residence principles should be thought of as an integral part of the entire census process, not a small, side component. They should be used as a template for the development of related census operations. Chapters 7 and 8 discuss three major operations—techniques for group quarters/nonhousehold enumeration, programs to update the Master Address File, and routines to unduplicate census records—that should be designed with residence principles as a guiding concept. Other census operations for which residence principles should be kept in the forefront include: development and implementation of unduplication algorithms, including any revisions to the primary selection algorithm (used to screen and combine duplicate census questionnaires) and plans for “real-time” unduplication during the census process (we discuss this briefly in Section 8–B); development of the advance letter that precedes the main questionnaire mailout, including instructions for requesting a foreign language ques-

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 6-1 Illustration of Application of Residence Principles as the Basis for “Frequently Asked Questions” 1. PEOPLE WITH ONLY ONE RESIDENCE Count at the location where they live or sleep more than any other place. Examples include: College students living at their parental home while attending college Unmarried partners Housemates, roommates, or boarders Foster children or foster adults Live-in employees (e.g., caregivers, domestic workers) Military personnel in the United States People who happen to be away Resident staff of embassies, legations, and other diplomatic facilities in the United Statesa 2. PEOPLE WITH MORE THAN ONE RESIDENCE Collect information on “any residence elsewhere.” Count at the location where they live or sleep more than any other place. If time is equally divided between locations, count at the location where they are found on Census Day. Examples include: College students living away from the parental home, in on-campus or off-campus housing Boarding school students People living away most of the time while working People who split time between two or more residences (during the week, month, or year), such as snowbirds, unmarried partners with separate residences, commuter workers, etc. Children in shared custody arrangements 3. PEOPLE WHOSE LIVING SITUATION CHANGES ON CENSUS DAY Count at the place where they live or sleep—or will live or sleep, in the case of babies— more than any other place. Births—babies born on or before midnight of Census Day Deaths—people who were alive at any time on Census Day People who move into a residence on Census Day, where they will live or sleep more than any other place People who move out of a residence on Census Day, but who have not completed the move to a new location on Census Day (the former residence should be considered the place where they live or sleep more than any other place) 4. U.S. CITIZENS LIVING OUTSIDE THE UNITED STATES Count on the basis of U.S. law and court precedent. Currently, this means including the following groups in apportionment counts only; all other U.S. citizens residing overseas are not included in the count. Military personnel stationed abroad Merchant Marines Other U.S. citizens employed by the U.S. government 5. PEOPLE LIVING IN SPECIAL PLACES ON CENSUS DAY Collect information on “any residence elsewhere,” along with information on duration of stay (tailored institutional forms). Based on duration of stay, availability of residence information, and whether census record for that residence lists that person, count at the residence. If no other information is available, count at the facility. Examples include: People in transient quarters, such as hostels, recreational camps, public or commercial campgrounds, racetracks, parks, or carnivals

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census People in federal and state prisons People in local jails or other municipal confinement facilities People in correctional residential facilities People in group homes for juveniles, including Job Corps Centers living quarters People in residential treatment centers People in health care facilities, including nursing facilities and hospitals People in emergency and transitional shelters on Census Day, for people experiencing homelessness People at soup kitchens, regularly scheduled mobile food vans, and targeted nonsheltered outdoor locations People in group homes and other residential treatment centers for adults    aUnder current law and practice, foreign citizens resident at embassies and legations in the United States may voluntarily decline to be included in the census count. tionnaire, and any follow-up mailings (e.g., reminder cards or the proposed second/replacement questionnaire mailout); refinement of the Bureau’s routines for editing census data and imputing for nonresponse; development of experiments to be performed during a decennial census and of formal evaluations of census operations; and design of public outreach programs. 6–C PRESENTATION OF RESIDENCE CONCEPTS TO RESPONDENTS AND ENUMERATORS The Census Bureau took a first step toward self-enumeration—that is, questionnaires filled out by respondents themselves—in the 1960 census. Households were mailed an “Advance Census Report” that enumerators later picked up in person and copied onto computer-readable schedules. On the strength of that experience, mailout and mailback of census questionnaires became the dominant form of census conduct in 1970, when approximately 60 percent of the nation’s housing units were mailed questionnaires. By 2000, approximately 82 percent of housing units were reached by mail. Mail administration of the census has obvious operational efficiencies in comparison with deploying enumerators to interview every household. However, self-enumeration profoundly affects the process of a census, shifting the burden of comprehending and interpreting the meaning of questionnaire concepts from a trained enumerator to an untrained respondent. As a result, the modern census questionnaire has to satisfy several, sometimes conflicting, constraints:

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census indeed counted at home, and limited time before 1990 in which to acquire computer capacity to facilitate such matches…. The costs and complexities of introducing these procedures increase the importance of identifying those groups for which follow-up of UHE reports is most likely to be productive and cost-effective. Computational capacity is a lesser concern now than in 1986, and the coverage evaluation research program for the 2000 census was a showcase for improved searching and matching capability. Ultimately, Bureau staff were able to learn much about the nature of potential duplicate records in the census by performing a complete match of the census returns against themselves based mainly on person name and date of birth (Fay, 2002b, 2004). This methodology is a good base for improvements in 2010 and beyond. The method was sufficiently tractable that the Bureau is considering “real-time unduplication” based on similar matches while census processing is still in progress (and while capacity exists to deploy enumerators for follow-up interviews). These developments suggest that the sheer technical capacity to collect and process additional reports of address information is at hand. However, other logistical concerns—limited field follow-up resources (to resolve discrepancies found during matching or to validate addresses not on the Master Address File) and the time constraint of producing apportionment counts by the end of the census year—carry great weight. Although we conclude that universal provision for ARE reporting is the proper direction for the census to follow, we also recognize the need for testing and evaluation of new procedures before they are applied in the census; our recommendations in Section 6–E.4 reflect these constraints. One of the coverage probes tested in the 2005 National Census Test (Figure 6-8, (b), Question 10) is close to an ARE question, though it asks only for the place (city/state) where the person’s other residence is located. The two Question 10 probes are a starting point for part of an ARE question (or an immediate follow-up to it), asking about the nature of the other residence (e.g., vacation home or school-location address for a college student) or the reason that the person resides there some of the time (e.g., seasonal move or commuter worker). Ideally, the ARE location can be obtained along with some indication of how much time the person spends at the other location so that usual residence can be determined. Our proposed ARE question does not go as far as some researchers of seasonal “snowbird” populations—or demographers, generally, interested in the dynamics of migration—would like. With specific eye toward collecting data on “snowbirds,” Happel and Hogan (2002) suggest a question akin to: “Besides your usual place of residence, have you spent 30 days or more during the past year at another locale? If so, at which location(s) did you stay and for how long (in each location)?” The structure of this question—a full residence profile for each person—also speaks to a limitation of the request for a single

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census ARE per person: such an approach would still not capture the full experience of intensive travelers, migrant workers, and recreational vehicle users with ties to more than two places. However, a universal ARE option at least breaks free from the notion that respondents must specify one, single place of residence and could provide greater accuracy in the many cases—college students, hospital patients, children in joint custody arrangements, and so forth—where the tension in reporting is between two places. Verifying Addresses Current census methodology makes the strong—and sometimes erroneous—assumption that the census questionnaire is properly delivered to the correct housing unit (and to the people who live there). In multiunit structures like apartment buildings, questionnaires may be placed in the wrong unit’s mailbox by mistake. Particularly for large multi-unit structures with a common mail delivery point, mail carriers may also view the census questionnaires as interchangeable (for better or worse, like high-volume advertising or “junk” mail) and put them in mailboxes haphazardly. The sheer volume of the major census mailout in a decennial year can contribute to the perception of the questionnaires as interchangeable, when in fact they are intended to reflect input from a specific designated housing unit. Improper sorting or misplacement prior to giving the mail to letter carriers can lead to questionnaires being delivered to the wrong house. The 2000 census questionnaire lacked a provision for a respondent to indicate that the address printed on the received questionnaire was erroneous, or that the questionnaire was misdelivered. This has not always been the case. The 1970 census questionnaire—the first administered primarily by mail—included a three-line address box; directly underneath, an italicized instruction read: “If the address shown above has the wrong apartment identification, please write the correct apartment number or location here.” One line was provided for response. The 1980 mailing label space shared the same feature and identical copy (save that the entry space was above, not below, the printed label); three lines were provided for entry. Foreign censuses (see Appendix B) generally do not offer guidance on or examples of such a correction question, yet whether self-administered or conducted by an interviewer, they direct a respondent to write in full address information rather than relying only on a printed address label. Providing space for respondents to advise of changes to the address printed on their census form is a relatively simple change, but it would entail some cost. Depending on its exact implementation, respondents’ handwritten corrections would likely be ill suited to automatic optical character recognition; accordingly, the burden of data entry would be shifted to human clerks (working with the raw paper forms or, more likely, scanned images). How-

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census ever, we believe that this cost would be offset by potential gains in the quality of data reporting and assignment of reported persons and households to particular housing units. It would also provide very valuable input for census follow-up, unduplication, and coverage evaluation efforts, and it would also provide useful operational data for evaluation (e.g., quantitative evidence of the magnitude of some mail delivery problems). Recommendation 6.3: The census questionnaire should allow respondents to correct the address printed on the form if it is wrong (e.g., address is listed incorrectly or questionnaire is delivered to wrong unit or apartment number). In addition to mail delivery problems, even questionnaires delivered correctly to the right address may not reach the “right” people. Prominent among these cases are seasonal residences that may be vacant for much of the year or are not consistently occupied by the same people over a year. The occupants of a home in February or March may receive the form—response to which is compelled by law—yet they have no mechanism in the current short form for indicating that their stay at the address in question is only temporary (and that they may correctly be counted elsewhere). People who move residences on or near Census Day may also face the problem of a correctly delivered questionnaire finding them at the “wrong” address. A questionnaire could be carried with the movers from the old housing unit to the new one, or it may be inadvertently forwarded. One approach to handling this problem could be to borrow from the 1990 census example and add a follow-up question to the address correction space—to provide checkboxes to indicate that all residents of the indicated housing units are there temporarily or are in the process of moving residences. The use of such a checkbox flag could be useful in identifying cases that need follow-up from an enumerator. However, we believe a better approach would involve collecting the address information for the other residence location (as did Question 1b on the 1990 census form; see Figure 6-4); this is entirely consistent with the broader change in residence questions that we describe in the next section. Coverage and Housing Type Probes We are encouraged that the 2005 test suggests that the Census Bureau is considering adding separate coverage probe questions, which were absent on the streamlined 2000 census questionnaire. As we discuss in Chapter 8, the Bureau’s interest in probe questions is partly operational, because responses to the probes are intended to trigger eligibility for an enhanced coverage followup operation; see Box 6-3. In addition to addressing these practical concerns, the use of coverage probe questions is beneficial in its own right; the questions

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census Box 6-3 Coverage Follow-Up Plans for the 2010 Census Plans for the 2010 coverage follow-up operation call for an expansion of the coverage improvement follow-up operation from 2000 (Moul, 2003). The revised operation would be triggered to prompt follow-up enumeration if the housing unit is returned as vacant; the household size is large, exceeding the limit of seven family members for whom full information can be reported; the household count reported at the beginning of the questionnaire does not match the number of people for whom data are returned; interview information suggests that people who are reported as living elsewhere should be reported as belonging to this household; interview information suggests that people who are reported in this household may more properly be counted at another residence; probabilistic matching based on name and date of birth suggests that there are potential duplicate entries for members of the household. The problems of implementing this are two-fold: first, the fraction of census returns for whom this operation would be triggered might be too large to be workable (e.g., reinterview of 30 percent rather than 10 percent), and second, it may conflict with the postenumeration survey conducted as part of coverage measurement. provide further opportunity to elicit residence information through questions rather than relying on a preamble of instructions. We note above that some of the probes used in the 2005 National Census Test are useful starts to follow-up questions to accompany a reported “any residence elsewhere” address. The presence of such a question, related to the basic nature of a housing unit, is particularly essential given the short-form-only orientation of the 2010 census; the standard question on housing tenure (own or rent) may be the only question focused on the housing unit itself and the people living in it. In general, we agree with the approach taken in past censuses, which was the basic goal of formulating possibilities for the 2005 test: it is essential that separate probes targeting overcount (potential duplicates) and undercount (possible omissions) be included on the census form. 6–E.3 Mode Effects There is another compelling reason that the Bureau should adopt a question-based framework for data collection, particularly the basic count in Question 1. It is well known from research on surveys that differences in the mode of administration of a census or survey can influence responses. The experience of answering a survey questionnaire from a human interviewer standing in one’s doorway is substantively different than that of an interview

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census conducted in the less personal interface of a phone call. The rapport and trust between interviewer and respondent, the perceived level of “security” in answering private queries, and the impulse to decline to answer some questions (or cut off an interview altogther) differ between the two modes, which in turn are different from the effects of responding through the impersonal media of mailed questionnaires or the Internet. Survey modes also differ in their ability to present instructions and support information. The decennial census has become a mixed-mode survey collection of vast proportion and will likely continue on that path. In 2000, responses were acquired through self-enumeration on paper questionnaires and through computer-assisted telephone interview if respondents called a number for assistance. Response to the census short form was also possible on the Internet though this option was not widely publicized. In 2010, mail and self-response will continue to be the bulk of the census response. However, the Bureau plans to conduct nonresponse follow-up data collection using computer-assisted personal interviewing on a handheld device; in 2000, follow-up interviews were conducted using paper forms. In the initial planning for the 2010 census, data collection by the Internet and by telephone using an automated “interactive voice response” system were expected to be major parts of the census process (National Research Council, 2004b), but this is no longer so. The interactive voice response system appeared to perform badly in the 2003 National Census Test, and the Bureau acknowledged in June 2006 that it planned to scrap Internet data collection.8 There is reason to believe that a question-based mode of collecting residence information may be more robust to mode differences than an instruction-based model, because response to instructions may depend crucially on medium of administration. We emphasize, though, that this is an empirical question, a supposition which requires testing of the form we call for in Chapter 8. Recommendation 6.4: Regardless of the final structure of residence questions chosen for 2010, research must be done on response effects created by mode of administration (mail, phone, Internet, interview with handheld computers). 6–E.4 Testing ARE in 2010 Some elements of a question-based approach to gathering residence information are already in testing, such as the coverage probe questions and the 8 The plans to drop Internet response were discussed at a June 6, 2006 hearing of the U.S. Senate Subcommittee on Federal Financial Management, Government Information, and International Security (a subcommittee of the Homeland Security and Governmental Affairs committee).

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census “worksheet” prototype for Question 1 on the 2005 census test. The most complicated part of the approach—placing an ARE item on the 2010 census questionnaire—is certainly feasible for 2010. However, effective plans for how to work with ARE data in processing census returns and scheduling enumerator follow-up as necessary are vitally important to the resulting census data. In the interests of keeping the 2008 census dress rehearsal as true a rehearsal as possible (and not a major experiment), the Bureau should test a question-based approach as part of the experimentation program of the 2010 census, in order to provide a base of information for further research over the next decade and possible inclusion in 2020. Recommendation 6.5: In the 2010 census, the Census Bureau should conduct a major experiment to test a form that asks a sufficient number of residence questions to determine the residence situation of each person, rather than requiring respondents to follow complicated residence instructions in formulating their answers. The results of this test, and associated research, should guide decision on full implementation of the approach in 2020. As we recommend in Chapter 7, ARE address data should be included on the questionnaires used to enumerate all elements of the group quarters and other nonhousehold population in 2010, and those data should be fully evaluated and used in unduplication screening as appropriate. As has been done in previous censuses, a sample of the household population should receive a question-based form with ARE items. From a research standpoint, it would be useful for evaluations of the collected ARE data to compare the results of this test with those of the census coverage follow-up program, in which more detailed residence probes and rostering questions are possible than on the regular census form. If possible, field reinterviews with some test respondents would be useful, perhaps using a traditional instruction-based form to see if the same results are obtained. It should be acknowledged that a form that implements a fully question-based approach to gathering residence information may also require a departure from recent census norms, in the physical sense. Since the start of self-administered forms in 1970, the census short form has generally been kept to a one-sheet, folded ledger-size booklet (the census long-form questionnaire, of course, included more pages). Accommodating both a modest number of new questions as well as visual features (to aid flow through the questionnaire) and adequate space for optical scanning may require a slightly longer or larger short-form booklet—possibly not an undesirable outcome, as the census questionnaire is now fairly dense and tight.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census 6–F A VIOLATION BY DESIGN: THE CENSUS DAY RESPONSE PROBLEM The language of the census questionnaire and accompanying mailings, together with the basic logistics of the decennial count, combine to create a problem that appears almost trivial at first, but ultimately cuts to the core meaning of the census. The problem is that the census is intended to be a count of the resident population of the United States on Census Day, April 1, of the decennial year. However, the country does not stand still on Census Day; the count is not made solely in that span of 24 hours. Instead, questionnaires are mailed out a few weeks in advance. In order to ensure smooth processing and enable follow-up interviewing to take place in fairly short proximity to Census Day, the Census Bureau’s incentive is to have the respondents answer the questionnaire as quickly as possible. Hence, the problem: what does it mean for a respondent to report the population of a household as of Census Day, if he or she is filling out and returning the questionnaire up to 2 weeks before that date? Since 1960, the questionnaires and their accompanying instructions have offered different guidance on exactly when the form is to be completed and returned: Some households receiving the 1960 form in the mail were asked to return the form by mail, as part of a census experiment; the households that received those instructions were advised to “please mail the completed form within 3 days in the special envelope.” In the most explicit instruction given in this regard, the beginning instruction on the 1970 census questionnaire read, “Please fill it out and mail it back on Census Day, Wednesday, April 1, 1970.” The 1980 questionnaire adapted the 1970 instruction slightly; the initial instruction told the reader to “please fill out this official Census Form and mail it back on Census Day, Tuesday, April 1, 1980.” Later in the instructions, this guidance was softened to mailing it back on Census Day “or as soon afterward as you can.” The 1990 instruction directly contradicted the notion of a resident count as of Census Day by urging responses prior to Census Day. The first page of the form declared, “Remember: Return the completed form by April 1, 1990,” accompanied by the softer instruction to “please answer and return your form promptly.”9 Despite this instruction, the census form still asked respondents to report their household members as of April 1 (as shown in Figure 6-4). 9 The 1990 instructions also added a curious warning, implying that forms should be returned early in order to “Avoid the inconvenience of having a census taker visit your home.”

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census None of the instructions in the 2000 census about mailing back the questionnaire were as explicit as in 1990, but the effective message was much the same. In early March, households received an advance letter: the letter, dated March 6, 2000, advised readers that “about one week from now, you will receive a U.S. Census 2000 form in the mail”; the letter asked the reader to “please fill it out and mail it in promptly.”10 The 2000 census questionnaire itself contained no deadline or guidance on when the form should be completed and returned; a March 13, 2000, press release on the mailout of 98 million census forms noted only that “people are asked to mail the forms back as soon as possible.”11 A week later, households received a follow-up postcard dated March 20 offering “sincere thanks” to those households that already returned the form; “if you have not, please fill out the form and mail it back as soon as possible.”12 Both the letters and the public relations stance suggested by press releases suggested that the 2000 census forms were to be returned as promptly as possible, even if the return was made before the Census Day reference point used in the questionnaire. The effect of possibly failing to fully account for some births, deaths, moves, and displacements in the last week of March may be relatively small. Yet however subtle the effect may be, the practice of encouraging immediate and rapid return of census questionnaires—possibly before the reference date of Census Day—is a basic violation of residence principles and could undermine the credibility of the concept in general. It is an empirical question as to whether being asked to violate one rule at the outset may prompt people to disregard other instructions; that kind of experiment has not been performed. Recommendation 6.6: To be consistent with the principle of the basic residence question on the census form—where did you live on April 1?—the Census Bureau should encourage prompt response but make clear that the form should be completed and returned on Census Day or as soon thereafter as possible. It would also be helpful for the major mailout of census materials to occur as close to Census Day as practicable. Too lengthy a lead time between the arrival of the questionnaire and Census Day encourages the “forecasting” 10 The 2000 census advance letter included a second page in several foreign languages, with checkboxes that allowed the respondents to request that a special-language census form be mailed to them. The advance letter is perhaps best remembered for a processing error; every letter was misaddressed due to the accidental insertion of a prefix digit before every house number (the barcode printed on the mailing was correct, though, and the Postal Service was able to deliver the letters). See http://www.census.gov/dmd/www/pdf/letter_1.pdf [6/1/06]. 11 “Nearly 100 Million Census Questionnaires in the Mail, Census Workers Delivering the Rest.” U.S. Department of Commerce News Release, March 13, 2000. Available: http://www.census.gov/Press-Release/www/2000/cb00cn24.html [6/1/06]. 12 See http://www.census.gov/dmd/www/pdf/letter_2.pdf [6/1/06]. The choice of a reminder postcard was based in part of research reported by Dillman et al. (1995).

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census of Census Day living situations that our recommendation tries to curb. Accounting for variation in mail delivery, the current schedule for delivery in mid-March is quite appropriate, and we do not suggest a delay in mailing out the questionnaires. Indeed, the public relations consequences of erring too far in the other direction—and having major segments of the population not receive questionnaires until after Census Day—would be severe. What we do recommend is that the Bureau not go out of its way (as with the 1990 census form instructions and the pre-Census Day follow-up letter in 2000) to encourage violation of the most basic residence precepts. 6–G RESEARCH NEEDS Issues of questionnaire format and layout can have an effect on response. In recent decades, the Census Bureau has probed some of these issues through its census tests and through stand-alone efforts like the 1993 Living Situation Survey. Recommendation 6.7: The Census Bureau should continue and strengthen its research on the combined effects of visual layout features and specific wording situations in the development and testing of questionnaires and their effectiveness. A particular, vitally necessary improvement to this research is discussed in detail in the next chapter—the need to use more small-scale experiments and field trials, rather than the current course of relying on either a very small number of cases in cognitive tests or on tens or hundreds of thousands of cases in census operational trials. The panel’s review of the evolution of the census forms being developed for the 2005 and 2006 census tests, with an eye toward the final design of the 2010 census questionnaire, revealed that the forms are being developed on the basis of inadequate experimental evidence. For instance, alternative forms being submitted for testing include multiple changes in content and layout (including content and formatting of the residence rules instruction box) so that the cause of improvements cannot be traced to a single feature. There are also discontinuities with previous versions: for example, a modified version of the question used in the 2000 AQE—and not the original—was used in the original plans for the 2005 test. Recommendation 6.8: When designing experimental tests, the Census Bureau should always include a control form—either the questionnaire items used in the preceding census or the exact items used in immediately previous census tests—so that individual modifications can be more effectively assessed.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census To achieve our recommended goals—a small number of core residence principles and a question-based (not instruction-based) approach for gathering residence information—the census short form will have to be longer than it currently is. Determining exactly how long the form should be is a matter for further research, which should include a major experiment in conjunction with the 2010 census. In addition, we encourage the Bureau to work with outside researchers—and conduct its own experiments, as possible—on length-of-form issues, such as respondent unease with questionnaires of various lengths. The Bureau should encourage work on formats that simplify the questions; prior work (Krosnick and Berent, 1993) suggests that respondents may find it easier to answer two simple questions rather than one complicated one and that higher quality data can result.

OCR for page 179
Once, Only Once, and in the Right Place: Residence Rules in the Decennial Census This page intentionally left blank.