APPENDIX
C

Data Requirements for Reapportionment and Redistricting

One of the first tasks of the panel was to examine the requirements for the decennial census to satisfy data needs for reapportionment of the U.S. Congress and the redrawing of congressional, state, and local districts for purposes of political representation. The Constitution of the United States mandates in article I, section 2, that "Representatives and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective numbers. … The actual Enumeration shall be made within three Years after the first Meeting of the Congress of the United States, and within every subsequent Term of ten Years, in such Manner as they shall by Law direct." Hence, the requirement to support congressional reapportionment, which, in turn, entails redrawing congressional district boundaries, represents the absolute bedrock on which the U.S. decennial census rests.1

The panel reviewed data requirements for reapportionment and redistricting (at the federal, state, and local levels) that stem from the Constitution (including amendments) and also from other statutes (e.g., the Voting Rights Act, Title 13 of the U.S. Code), judicial interpretation, and administrative practice. The panel sought to determine how open to interpretation the requirements might be, so that, in turn, it could consider the fullest possible range of census designs in the spirit of a "zero-based" assessment of the most cost-effective ways to conduct

With the exception of the information on the August 1994 ruling of the U.S. Court of Appeals for the Second Circuit on the 1990 census adjustment issue, this appendix appeared in the panel's 1993 interim report, Planning the Decennial Census: Interim Report.



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 239
Modernizing the U.S. Census APPENDIX C Data Requirements for Reapportionment and Redistricting One of the first tasks of the panel was to examine the requirements for the decennial census to satisfy data needs for reapportionment of the U.S. Congress and the redrawing of congressional, state, and local districts for purposes of political representation. The Constitution of the United States mandates in article I, section 2, that "Representatives and direct Taxes shall be apportioned among the several States which may be included within this Union, according to their respective numbers. … The actual Enumeration shall be made within three Years after the first Meeting of the Congress of the United States, and within every subsequent Term of ten Years, in such Manner as they shall by Law direct." Hence, the requirement to support congressional reapportionment, which, in turn, entails redrawing congressional district boundaries, represents the absolute bedrock on which the U.S. decennial census rests.1 The panel reviewed data requirements for reapportionment and redistricting (at the federal, state, and local levels) that stem from the Constitution (including amendments) and also from other statutes (e.g., the Voting Rights Act, Title 13 of the U.S. Code), judicial interpretation, and administrative practice. The panel sought to determine how open to interpretation the requirements might be, so that, in turn, it could consider the fullest possible range of census designs in the spirit of a "zero-based" assessment of the most cost-effective ways to conduct With the exception of the information on the August 1994 ruling of the U.S. Court of Appeals for the Second Circuit on the 1990 census adjustment issue, this appendix appeared in the panel's 1993 interim report, Planning the Decennial Census: Interim Report.

OCR for page 239
Modernizing the U.S. Census future censuses. Specifically, the panel considered the possibility that designs that made use of sampling and administrative records, which might offer cost savings and other benefits for the census, could satisfy data requirements for reapportionment and redistricting. REAPPORTIONMENT As noted above, the U.S. Constitution mandates the conduct of an "enumeration" every 10 years for the purpose of reapportionment of the U.S. House of Representatives.2 The interpretation of the word enumeration is obviously key to an assessment of whether census designs that involve sampling or administrative records could serve this fundamental purpose. A second important consideration is that reapportionment must be effected simultaneously for the entire country—one cannot reapportion in some areas in one year and in other areas in another. The Role of Sampling A legal review prepared by the Congressional Research Service (Lee, 1993) concludes that, for the purpose of reapportionment, there needs to be an attempt to account for every inhabitant in the country: a sample census, no matter how large, cannot satisfy the constitutional requirement.3 Similarly, "rolling census" designs, in which different parts of the population are surveyed each year without even a minimal census of the entire population at any one time, would not satisfy the requirement. (The rolling census designs proposed by Horvitz, 1986, and Kish, 1981, 1990, are in this category.) Other rolling census designs, namely, those that do include a minimal census every tenth year, would satisfy the constitutional requirement. (The designs proposed by Herriot et al., 1989, and Alexander, 1993, are of this type; see Chapter 6 for further discussion of the potential and problems of rolling census or continuous measurement designs that include a minimal decennial census together with rolling surveys.) Lee (1993) draws the conclusion that an attempt at a complete count is constitutionally required from a review of the meaning of the word enumeration at the time the Constitution was adopted and subsequent legislative and judicial history.4 In particular, two sections of Title 13 of the U.S. Code (which pertains to the Census Bureau) address the topic of sampling in the decennial census. Section 195, adopted in 1957, states that "except [emphasis added] for the determination of population for purposes of apportionment of Representatives in Congress among the several States, the Secretary shall, if he considers it feasible, authorize the use of the statistical method known as 'sampling' in carrying out the provisions of this title." Section 141(a) appears to be more liberal, in that it authorizes the secretary of commerce to take a decennial census every 10 years "in such form and content as he may determine, including the use of sampling

OCR for page 239
Modernizing the U.S. Census procedures and special surveys." Lee (1993) states that the courts have reconciled the two sections by holding that the Census Bureau may use sampling procedures in the census, but only in addition to more traditional methods of enumeration (see, e.g., Carey v. Klutznick [1980, Southern District, New York]; other relevant cases are referenced below). Although a sample census and some rolling census designs appear precluded on constitutional grounds, the use of sampling as part of the census process appears compatible with the spirit of the constitutional, legislative, and judicial history regarding enumeration, so long as the process includes an effort to reach all inhabitants. Specifically, designs that use sampling for the follow-up stage of census operations (after an initial attempt has been made to deliver a questionnaire to every household or person) and for coverage improvement programs (including adjustments based on sample surveys) would appear to meet the data requirements for reapportionment. Several court cases (in addition to Carey v. Klutznick) have explicitly upheld the constitutionality of an adjustment, citing the importance of having as accurate data as possible for reapportionment and redistricting (see Lee, 1993:18-20; NCSL Reapportionment Task Force, 1989:4-5).5 In City of New York v. U.S. Department of Commerce (1990, Eastern District, New York), the court stated that it "is no longer novel or, in any sense, new law to declare that statistical adjustment of the decennial census is both legal and constitutional. This Court has already recognized that Article I, Section 2 require[s] that the census be as accurate as practicable." Subsequently, the Eastern District Court in New York reaffirmed the conclusion that "the Constitution is not a bar to statistical adjustment" (City of New York v. U.S. Department of Commerce, 1993). Although the court upheld the constitutionality of adjustment, it ruled that the decision of the secretary of commerce not to adjust the 1990 census results followed the guidelines developed by the Commerce Department for the adjustment decision and could not be deemed to be arbitrary or capricious. The question of the legality of sampling for nonresponse follow-up has never been explicitly raised in the courts; however, language used in the court cases just cited clearly seems to support its use. For example, in Carey v. Klutznick (1980, Southern District, New York), the court held that the Census Bureau may use sampling procedures in addition to a traditional enumeration. In Young v. Klutznick (1980, Eastern District, Michigan), the court noted that, since 1970, the census has not been a "simple straight forward headcount" but instead "a relatively accurate estimate of the population developed through the use of self-enumeration by questionnaire, statistical techniques, and computer control devices." The court held that section 195 of Title 13 did not prohibit the use of statistical techniques in the census. (As Lee [1993:20] observed, the court appeared to treat "statistical techniques" as equivalent to "sampling.")6 There are precedents for designs that use sampling in the later stages of the census process. In the 1970 census, two coverage improvement programs were

OCR for page 239
Modernizing the U.S. Census conducted on a sample basis, and the results were used to add people to the census by an imputation procedure.7 In the 1990 census, a postenumeration survey was conducted of a sample of housing units for purposes of evaluating the completeness of the population count and developing adjusted counts on the basis of the sample survey results. These adjusted counts were not used for reapportionment, but the court-ordered process under which they were developed certainly contemplated that they might be so used (see City of New York v. U.S. Department of Commerce, 1989, 1990, Eastern District, New York). Most recently, in an August 1994 ruling, the U.S. Court of Appeals for the Second Circuit vacated the decision in City of New York v. U.S. Department of Commerce (1993) that upheld the secretary's decision not to adjust the 1990 census and remanded the case back to the lower court for reconsideration. The appellate court agreed with the lower court's view that Title 13 does not bar the use of sampling for a statistical adjustment of the initial census enumeration, and also with its assessment that statistical adjustment is feasible and would have improved the 1990 census counts and reduced the disproportionate undercounting of minority groups. The appellate court, however, disagreed with the basis on which the lower court upheld the secretary's decision against adjustment, namely, the arbitrary and capricious standard of review in the Administrative Procedures Act. The appellate court argued that a higher standard of review should prevail, stating that the decision not to adjust the 1990 census must be shown to be necessary to a legitimate governmental interest, or it cannot be upheld, and remanding the case back to the lower court "for further proceedings not inconsistent with this opinion." The court's reasoning was as follows: the states and the federal government are required by the equal protection clause of the Fourteenth Amendment and the due process clause of the Fifth Amendment to strive to ensure that the vote of every citizen counts equally; there was a demonstrated differential undercount in the 1990 census that disadvantaged identifiable minority groups in this regard; and the means by which to improve the count by statistical adjustment were available. The circuit court concluded that the "plaintiffs amply showed that the Secretary did not make the required effort to achieve numerical accuracy as nearly as practicable, and that the burden thus shifted to the Secretary to justify his decision not to adjust the census in a way that the court found would for most purposes be more accurate and lessen the disproportionate counting of minorities" (City of New York v. U.S. Department of Commerce, 1994, Second Circuit Court of Appeals). Role of Administrative Records The use of administrative records (e.g., income tax and Social Security records) to provide complete population data for reapportionment raises a different set of issues from the use of sampling. The use of records is probably not

OCR for page 239
Modernizing the U.S. Census consistent with most people's idea of a census, in that there would be no attempt to contact all the people on or close to a designated Census Day. However, such use could be viewed as meeting the constitutional requirements for reapportionment, if there were an administrative records system (or a combination of systems) that, when used for purposes of a census, could be determined to contain data for all inhabitants (or as close to the total of all inhabitants as has been achieved for traditional U.S. censuses), with the records assigned to the correct state of residence. See Chapter 4 for discussion of both the potentials and the problems of exploiting administrative records for the U.S. census. There is no body of legal opinion on which to base an assessment of the constitutionality of an administrative records census. However, both Lee (1993:30) and Barnett (1993) conclude that a census based on administrative records would be likely to be held by the courts to be constitutional if it could be demonstrated that the data were accurate. With regard to the more limited use of administrative records as part of the U.S. census process, there are several precedents. The nonhousehold sources program was an administrative records-based operation to improve coverage in the 1980 census (Citro and Cohen, 1985:94,200). It involved matching several lists to census records for selected census tracts in urban district offices. The lists used were driver's license records, immigration records, and public assistance records in New York City. Enumerators visited addresses of people identified from the match who might have been omitted from the count. The parolee/probationer check was an administrative records-based coverage improvement program adopted in the 1990 census (Ericksen et al., 1991:43-47). As part of this operation, probation officers in large cities and smaller cities with large minority populations were asked by census enumerators to verify addresses of parolees and probationers obtained from records. These addresses were matched against the census, and all cases of nonmatches were added to the census, with no attempt at a personal follow-up. In other words, this program added individuals to the census based solely on administrative records, as might be done for a census entirely based on administrative records. Both programs had serious problems of implementation that indicate needed areas for further research and development to improve the data quality and cost-effectiveness of coverage improvement efforts that make use of administrative records. The 1980 nonhousehold sources program had very low payoff in terms of additions to the census count. A total of about 6.8 million records were checked against the census, but only 130,000 people were added to the count as a result of the matching and field follow-up operations (Citro and Cohen, 1985:200). The parolee/probationer check had a high error rate. An estimated 53 percent of the additions to the census count as a result of this program (about 250,000 people) were erroneous enumerations, that is, people who were already counted or who should not have been included for some other reason (Ericksen et al., 1991:44).

OCR for page 239
Modernizing the U.S. Census REDISTRICTING ON THE BASIS OF TOTAL POPULATION Legal Foundations Reapportionment of the U.S. Congress or a state or local legislature carries the implication that district boundaries should be redrawn to accommodate changes in the number of seats allotted to the jurisdiction and, even if that number does not change, to accommodate changes in the distribution of population so that a vote in one district carries about the same weight as a vote in another. Indeed, in the nineteenth century, Congress typically passed a statute at the time of each census that required all states, whether or not they gained or lost seats, to redistrict and to establish single-member districts that were contiguous, compact, and as nearly equal in population as practicable (Durbin and Whitaker, 1991:4-5).8 After the 1920 census, however, Congress declined to reapportion the House because of the concerns of rural interests about the tremendous population growth in the cities, particularly from immigration. In 1929, Congress passed an act that provided for automatic reapportionment upon delivery of the population counts after each census, but it set no standards for redistricting. The courts held that the omission of such standards was intentional, and, since then, it has been up to the courts themselves to provide for any standards (Durbin and Whitaker, 1991:4-5).9 From the 1920s through the 1950s, the courts generally declined to intervene in the political thicket of redistricting, and congressional and state legislative districts became increasingly more unequal in population size. Many states chose not to redistrict after a census, unless they gained or lost seats, and those that did often paid little attention to achieving population equality across districts. Very large deviations in population, generally favoring rural over urban and suburban districts, were quite common. After the 1960 round of reapportionment and redistricting, the largest congressional district in the United States had over five times the population of the smallest district; the 20 most populous districts had a combined population of 14 million compared with a combined population of 4.6 million for the 20 smallest districts. Disparities among state legislative districts were even greater (Baker, 1986:258). A 1967 law required single-member congressional districts. Also, the Voting Rights Act, as interpreted by the courts and administrative practice, led to de facto standards with regard to the representation of minorities. (The Voting Rights Act was enacted in 1965 and has been extended and amended several times since then.) The landmark ''one-person, one-vote" Supreme Court decisions, beginning in the early 1960s, changed the requirements for redistricting drastically. In the first of these cases, Baker v. Carr (1962), which involved Tennessee state legislative districts, the court held that reapportionment and redistricting matters were subject to judicial review under the equal protection clause of the Fourteenth

OCR for page 239
Modernizing the U.S. Census Amendment. In Wesberry v. Sanders (1964), the court held, under article 1 of the Constitution, that congressional districts must be as nearly equal in population as practicable. In White v. Weiser (1973), the court rejected a Texas congressional redistricting plan in which the smallest district was about 4 percent smaller than the largest district, and in Karcher v. Daggett (1983), the court in a 5 to 4 decision rejected a New Jersey congressional redistricting plan in which the smallest district was only 0.7 percent smaller than the largest district. The court held that the state could have avoided such a deviation, as it had rejected a plan with a population deviation of only 0.45 percent. Furthermore, the court ruled, the state had failed to show that the deviation in its approved plan was needed to achieve a legitimate goal (Parker, 1989:61; see also Durbin and Whitaker, 1991:12; and Ehrenhalt, 1983:56-57). Whether the Supreme Court will continue to view virtually absolute population equality among congressional districts as an overriding constitutional requirement is, of course, not certain. Although Karcher v. Daggett (1983) is certainly indicative of a strict interpretation, the decision was a close one (5 to 4), and language in both the concurring and dissenting opinions at least raises the possibility that the court could modify its pursuit of absolute population equality in the future.10 Over the same period, the Supreme Court issued decisions that greatly affected state and local as well as congressional redistricting. In Reynolds v. Sims (1964), the court held that, under the Fourteenth Amendment, both houses of a state legislature must be apportioned on a population basis.11 Moreover, although mathematical exactness may be impossible, states should strive for population equality. Generally, however, the courts have allowed more deviation among state legislative seats than among congressional districts. The guidelines appear to be that deviations of up to 10 percent in the size of state districts are constitutionally acceptable, although they can be challenged on other grounds (e.g., racial discrimination). Deviations between 10 and 16 percent are presumed to be unconstitutional, but states can try to justify them; deviations above 16 percent are usually viewed as completely unacceptable (Parker, 1989:57-58; see also O'Rourke, 1980:22). Census Data for Redistricting After the 1970 census, the states could obtain population counts for geographic areas as small as city blocks, which were defined in urbanized areas and in other localities that contracted with the Census Bureau, and for enumeration districts in unblocked areas. However, no special data files or reports were provided specifically to meet redistricting needs. In 1975 Congress required the Census Bureau to provide decennial census population tabulations to state officials for purposes of legislative reapportionment or redistricting within 1 year after the census date (i.e., under the current schedule, by April 1 of census year

OCR for page 239
Modernizing the U.S. Census plus one) (Public Law 94-171; section 141(c) of Title 13). States can specify the geographic areas for which they require tabulations, provided that their requirements satisfy Census Bureau criteria and are transmitted to the Bureau no later than 3 years prior to the census date; if no special areas are identified, the Census Bureau is to provide "basic tabulations of population." In practice, basic tabulations have come to mean tabulations for individual blocks, the smallest area of geography identified in census data products. After the 1980 census, the P.L. 94-171 data file provided by the Census Bureau contained the following information: total population; white; black; American Indian, Eskimo, and Aleut; Asian and Pacific Islander; and other races; and a separate count of Hispanics. Data were provided for states, counties, minor civil divisions, places, voting districts (when specified by the state), census tracts or block numbering areas, enumeration districts or block groups, and blocks. The number of blocks identified in the 1980 census was 2.5 million, an increase from 1.7 million in 1970. Blocks were identified in all urbanized areas, all incorporated places of 10,000 or more population, and other areas for which a state or local government contracted with the Census Bureau to define block boundaries. The average population per block was estimated in the 1980 census at about 70 people, and the average population per enumeration district at about 600 people (Bureau of the Census, 1982:56, 67, 79). Voting precincts identified by the states were generally the size of an enumeration district or group of blocks. In 1990, the P.L. 94-171 file was expanded to include, in addition to the 1980 content, cross-tabulations of all items by age (under 18 and 18 years and over), a cross-tabulation of race by Hispanic origin, and a count of occupied housing units (included in response to requests from other users).12 The geographic areas identified in 1990 were the same as in 1980, except that enumeration districts as a concept no longer existed, and blocks—about 10 million in all—were defined for the entire country (Bureau of the Census, 1992:82-83). Although P.L. 94-171 requires the Census Bureau to furnish decennial census tabulations to the states, the courts have clearly held that the states may use other data sources for redistricting purposes. Over time, however, the states, on their own initiative and prodded by the courts, have come to rely almost exclusively on census data to prepare redistricting plans. In the 1960s and 1970s, several court cases held that a state could use other than decennial census data for congressional and other kinds of legislative districts (NCSL Reapportionment Task Force, 1989:12-13). In 1966, citing Hawaii's special military and tourist populations, the Supreme Court in Burns v. Richardson held that the state could redistrict on the basis of numbers of registered voters. However, this decision was reached after the court determined that the results would not have been substantially different from those based on total citizen population. In Kirkpatrick v. Preisler (1969), the court implied that the eligible voter population could be the basis for redistricting if identified properly

OCR for page 239
Modernizing the U.S. Census and applied uniformly. In Ely v. Klahr (1971), the court cautioned that a new plan for Arizona legislative districts could use registered voter data only if the results would not differ substantially from what would have resulted "from the use of a permissible population base." In the 1980s, the case law generally reflected the position that alternative sources of data could be used if they were applied uniformly and the results were comparable to those under a plan based on the total or total citizen population. For example, the District Court in Massachusetts upheld the use of a state census for legislative redistricting in McGovern v. Connolly (1986). However, a district court struck down a New Mexico plan that was based on number of votes cast, and, in 1982, the District Court in Hawaii struck down Hawaii's state legislative plan that used registered voters, finding that the results did not "substantially approximate" those based on total population. In the same case (Travis v. King, 1982), the District Court struck down Hawaii's congressional redistricting plan, also based on registered voters, as unconstitutional: "[P]ursuant to Article I, Sec. 2 of the Constitution states must depend on total federal census figures to apportion congressional districts within their boundaries." Recent cases dealing with the issue of adjustment of the census for coverage errors have generally upheld the view that the states need not use census data, at least not exclusively (Lee, 1993:5-6). For example, in reviewing Young v. Klutznick, the appeals court (1981, 6th Circuit) held that states are not constitutionally required to use census data supplied by the Census Bureau for redistricting, but can use adjusted population figures, so long as the adjustment was thoroughly documented and systematically applied. In City of Detroit v. Franklin (1992, Eastern District, Michigan), the court noted that Karcher v. Daggett did not hold that states must use census figures in redistricting but rather must use "the best population data available." In Senate of the State of California v. Mosbacher , the appeals court (1992, 9th Circuit) held that the Census Bureau was under no obligation to release adjusted data, but, if the state knew the census data were underrepresentative of the population, it could and should use noncensus data, in addition to the official count, for redistricting. In a recent case, City of New York v. U.S. Department of Commerce (1993, Eastern District, New York), the court found that there was a public interest in having available data tapes containing adjusted 1990 census counts down to the block level for the entire United States. The court ruled that the plaintiffs, who had acquired these tapes from the Census Bureau as part of the court-ordered process for deciding whether to adjust the census, could make the tapes publicly available.13 Over time, virtually all states have come to rely on census population counts for legislative redistricting. When states have used other bases, such as registered voters, they have generally had to obtain census data to demonstrate to the courts that their data would not give a substantially different result from census data. Although the court cases on adjustment noted above appear to give considerable discretion to the states in their choice of population data, they deal largely

OCR for page 239
Modernizing the U.S. Census with the question of whether to use the official census counts or adjusted figures that are also based on the census. It seems clear that population counts for redistricting are a practical, if not precisely a constitutional, required use of the U.S. census. Required Level of Geographic Detail Over time, the need for block data from the census for redistricting has also become a practical requirement: although not mandated, block data are treated by most parties as if they were mandatory. The driving force behind this focus on block data appears to come from the stipulations of the courts that there be virtually no deviation among congressional districts in population size and very little deviation among state legislative districts. In fact, not all states actually use block data in the redistricting process; many use data for election precincts or voting districts for which they have specified the boundaries to the Census Bureau (under the provisions of P.L. 94-171). These districts are generally the size of a block group or enumeration district in the old terminology. Commenting on the 1980 census P.L. 94-171 data program, officials of 33 states said they preferred working with block totals, and 15 states indicated they preferred working with voting district summaries. A total of 28 states, however, wanted to see voting district summaries for the entire state, and 8 states wanted to see such summaries for portions of the state in order to relate voting data to the block or precinct population data (Romig, 1983:9-10; see also Bureau of the Census, 1983; for a preliminary assessment of reactions to the 1990 P.L. 94-171 program, see National Conference of State Legislatures, 1992). The Census Bureau requires that voting district boundaries follow streets or other geographic features so that they align with census geographic boundaries; however, the precincts need not and typically do not align with geographic levels larger than the block (e.g., they often cut census tract boundaries). Hence, the Census Bureau has used block data to provide voting district summaries for the states that request them. As noted above, the courts have given primacy to the population equality of districts over such other criteria as compactness or contiguity. The result has often been the creation of very peculiar-looking districts, as state legislative majorities seek to redraw district boundaries in a manner that maximizes partisan advantage.14 Indeed, another force to retain blocks as the basic unit for redistricting is the widespread use of voting data (e.g., percent Democratic and Republican) in conjunction with census data to determine district boundaries. Typically, the people who work on redistricting want to look at several years of voting data because precinct boundaries change frequently: a common practice is to allocate precinct voting data to blocks and reaggregate the blocks to try out various redistricting plans to determine their political advantage. All of this manipulation of small-area data has been made possible by computers

OCR for page 239
Modernizing the U.S. Census and the advent of the Census Bureau's TIGER geocoding system. Indeed, one firm that assisted about half the states with redistricting after the 1990 census expressed the wish that census data could be provided for block faces (i.e., sides of blocks). The widespread use of computers as a data processing and mapping tool has made it easy for legislators to examine a variety of plans in order to determine the best one from their point of view. In turn, this behavior has made it necessary to have a database that can be reaggregated in many ways.15 Role of Sampling and Administrative Records It seems clear that the provision of population figures by block is, for all practical purposes, a burden that the U.S. census must satisfy. Even if the restrictions on allowable deviations in population size were to be relaxed somewhat in the future, small-area data would still be required to provide the basic units for definition of legislative districts. This requirement all but eliminates the sample census design, even if such a design were otherwise determined to be constitutional and feasible. The sample size would have to be so large it would practically preclude any cost savings compared with a complete count census. There appears to be no constitutional barrier to the use of sampling as part of the census process and hence no reason to rule out its use a priori, but questions have been raised about the quality of block-level data that might result under designs that incorporate sampling. For example, a factor in the decision not to use adjusted census counts in 1990 or, more recently, for intercensal estimates, was the belief that the adjustment factors developed on the basis of the postenumeration survey were not sufficiently reliable for small geographic areas (Bryant, 1993). Certainly, the merits of sampling as part of the census process require thorough research to determine its effect on total error and the costs of implementing various types of sample operations. It is possible, however, that the net effects could be positive: for example, careful sampling for nonresponse follow-up, at an appropriate sampling rate, might reduce costs and also reduce total error by decreasing nonsampling error more than the added variability due to sampling. Moreover, it is critical to keep in mind that block data are the input to redistricting, not the output. So long as the block data are of sufficient quality that, when aggregated to the congressional or other district level, the quality of the estimated population of the larger area is high, then the data requirements for redistricting would be served. Census designs that make use of administrative records might also meet the practical requirement for population totals at the block level for redistricting purposes, if a number of problems can be overcome. One problem for the purpose of redistricting—which is likely to be much more severe than the corresponding problem for congressional reapportionment—concerns the accuracy of

OCR for page 239
Modernizing the U.S. Census the addresses in administrative records systems, which may be out of date or pertain to a business or another individual (e.g., a tax preparer). REDISTRICTING ON THE BASIS OF RACIAL OR ETHNIC COMPOSITION All of the above discussion of data needs for reapportionment and redistricting has focused on total population figures. The civil rights movement of the 1950s and 1960s led to legislation, court decisions, and administrative practices that moved another requirement front and center, namely, the need for data on race and ethnic origin for purposes of legislative redistricting.16 Legal History The Voting Rights Act, originally passed in 1965 (P.L. 89-110) and extended and amended in 1970, 1975, 1982, and 1992, has led to the practical necessity for race and ethnicity data from the census (see Laney, 1992, for a history of the act). The act nowhere actually stipulates the use of census data, although it does require the director of the Census Bureau to make a number of determinations. But interpretations of the act by the courts and the Justice Department have virtually mandated the need for census data in redistricting. The original intention of the Voting Rights Act was to make it possible for blacks in the South to obtain the opportunity to participate in elections, an opportunity that was often denied them by unreasonable literacy tests and other barriers to registration and voting. The act, as enacted in 1965, prohibited (in section 2) under the authority of the Fifteenth Amendment the enactment of any election law to deny or abridge voting rights on account of race or color. It further specified (in section 4) that any state or county that had any test or device as a condition for voter registration on November 1, 1964, and in which the number of registered or actual voters fell below 50 percent of the total voting-age population in the 1964 presidential election could not use a literacy test or any other test or device to screen potential voters. Finally, it provided (in section 5) that any covered jurisdiction (i.e., any jurisdiction required to drop voting tests under section 4) had to submit "any voting qualification or prerequisite to voting, or standard, practice, or procedure with respect to voting" adopted after November 1, 1964, for "preclearance" to the U.S. Department of Justice or the U.S. District Court for the District of Columbia to determine that there was no abridgement of the right to vote on the basis of race or color. The 1970 amendments to the Voting Rights Act outlawed literacy tests and other devices in all jurisdictions, not just those in covered jurisdictions, and extended coverage to jurisdictions that had such tests in November 1, 1968, and in which there was less than 50 percent registration or turnout in the 1968 presidential

OCR for page 239
Modernizing the U.S. Census election. The effect of this provision was to cover subdivisions in northern and western as well as southern states. The 1975 amendments to the act included a major new provision that, on the basis of the equal protection clause of the Fourteenth Amendment, extended coverage under the act to protect the voting rights of language minorities, defined to be people of Spanish heritage, American Indians, Asian Americans, and Alaskan Natives. The preclearance provisions of the act (i.e., the requirement to submit proposed changes in voting procedures to the Justice Department for approval) were applied to any jurisdiction (counties and independent cities in most states and townships in others) for which the Census Bureau determined that more than 5 percent of the voting-age citizens were of a single-language minority, election materials had been printed only in English for the November 1972 elections, and less than 50 percent of all voting-age citizens in the jurisdiction had registered or voted in the 1972 presidential election. This provision covered the states of Alaska, Arizona, and Texas and political subdivisions in eight other states. The 1982 amendments to the act kept the basic provisions intact, but made some changes. The amendments extended the preclearance section of the act (likewise the provision for examiners and election observers in covered jurisdictions) to the year 2007, but also provided that Congress reexamine them in 1997. (Jurisdictions can petition for release from the provisions at an earlier date, but they must meet a stiff set of criteria for release.) Another provision stated that the standard of proof for judging an election law to be discriminatory under section 2 (as well as section 5) was no longer discriminatory intent, but, rather, discriminatory result. As somewhat of a counterbalance, still another provision stated that minorities had no right to proportional representation, but that the courts could consider the lack of representation as part of the totality of circumstances in cases brought under the Voting Rights Act. Finally, although not related to redistricting data needs per se, the act, as amended, included a provision (section 203) that is currently satisfied by using data from the census long form. This provision, as first adopted in 1975, required jurisdictions (counties, cities, or townships) to provide election materials and oral assistance in another language as well as English in areas for which the Census Bureau determined that 5 percent of the voting-age citizens were of a single-language minority and the illiteracy rate in English of the minority (defined as failure to complete fifth grade) was greater than the illiteracy rate in English of the entire nation. The 1982 act amendments asked the Census Bureau to investigate the usefulness of 1980 census long-form questions on mother tongue and English-speaking ability for determining coverage under bilingual assistance provisions. On the basis of the Census Bureau's research, of definition of a covered area became one in which 5 percent of the citizens of voting age comprised a group of people who spoke a single minority language and who

OCR for page 239
Modernizing the U.S. Census said they did not speak English very well and who had a higher illiteracy rate than the nation as a whole. The result of this change was to reduce the number of covered areas from about 500 following the 1970 census to about 200 following the 1980 census; about 300 covered areas were identified after the 1990 census. The 1992 amendments extended the bilingual voting assistance provisions to 2007 and made some additional minor changes to the definition (see Bureau of the Census, 1976; Kominski, 1985, 1992). With regard to data needs for redistricting, it is section 5 of the Voting Rights Act, with its requirement for federal review and preclearance—not only of tests for voting registration but of any "standard, practice, or procedure with respect to voting"—that has led to the practical necessity for census figures on race and ethnicity by block for redistricting.17 A key case that supported the use of section 5 to review many aspects of states' (and counties' and cities') electoral systems was Allan v. Board of Elections (1969), in which the Supreme Court held that such changes as moving from single-member to multimember districts were "practices or procedures" that were subject to review under section 5 because they had the potential of "diluting" the black vote. The Justice Department quickly moved to instruct legal officers in covered jurisdictions to clear every change in voting procedure. Whereas only 323 voting changes were received by the Department for preclearance between 1965 and 1969, almost 5,000 were submitted between 1969 and 1975 (Thernstrom, 1979:59).18 From 1965 to 1988, the Justice Department most often objected to three ways of setting up electoral systems on the grounds that they would have resulted in abridging or diluting the voting power of blacks, Hispanics, or other protected minority voters (Parker, 1989: Table 6.1): municipal annexations (1,088 of 2,167 total objections over the period); changing from single-member districts to at-large voting (472 objections); and redistricting plans that lessened the effectiveness of minority votes, for example, such schemes as dividing concentrations of minority voters into adjoining majority-white areas or minimizing the number of minority districts by placing minority voters in as few districts as possible (248 objections). Congress also specifically expressed an interest in having redistricting plans reviewed by providing in the 1975 amendments that jurisdictions covered as of 1965 could not seek release from coverage until 1982 and by extending that date in the 1982 amendments to 1984. These dates were enacted to ensure that there would be time for Justice Department or court review of redistricting plans based on the 1980 census in those jurisdictions (Laney, 1992:18, 24). Role of the Census As noted above, the data files provided to the states by the Census Bureau (under P.L. 94-171) include race and ethnicity counts by blocks—and voting districts if specified by the state—in addition to total population. The 1990 file

OCR for page 239
Modernizing the U.S. Census added a cross-tabulation of race by Hispanic origin. In a survey of reactions to the 1990 redistricting data program, 32 states said they used the race and ethnicity data, and 3 said positively that they did not (National Conference of State Legislatures, 1992:8). As noted above, not all states are covered by the preclearance provisions of the Voting Rights Act, although, potentially, redistricting plans in any state can be challenged under section 2. Overall, it seems clear that, for many states, the data are an important input to the redistricting process. Whether this situation will continue unchanged in the future is unclear. The preclearance procedures of the Voting Rights Act are scheduled to be reviewed by Congress in 1997 and, in any event, to expire in 2007. However, the language of the 1982 amendments, although not guaranteeing the right to proportional representation, states that minority representation is a factor to be considered by the courts. Also, historically, the preclearance provisions have been extended every time the act has come up for renewal. It would seem prudent to expect that the census in the year 2000 and most likely in the year 2010 will need to supply block-level data on race and ethnicity for purposes of legislative redistricting.19 Adding a characteristic such as race or ethnicity to census requirements raises the issue of measurement error and, for designs that make use of administrative records, the issue of data availability. See Chapter 7 for a discussion of several questions and concerns with regard to obtaining race and ethnicity data from the census. NOTES 1   The provision in article 1, section 2 (also in article 1, section 9) that required direct taxes to be based on the census was effectively repealed by the Sixteenth Amendment, ratified in 1913, which stated that "the Congress shall have power to lay and collect taxes on incomes, from whatever source derived, without apportionment among the several States, and without regard to any census or enumeration." 2   The current reapportionment formula, which uses the method of "equal proportions," was written into law at the time of the 1940 census (Anderson, 1988:189). 3   For the panel's review of legal issues, a report was requested from the Congressional Research Service (CRS). The CRS review was prepared by Margaret Lee (1993), an attorney with its American Law Division. Lee's report was reviewed internally by the Congressional Research Service. The panel then asked three constitutional scholars with expertise on census issues to review the CRS report and to offer their opinions on several census issues of interest to the panel (e.g,. the use of sampling in the census). The panel relied on the consensus of these scholars as well as its own judgment for its interpretation of the CRS report.

OCR for page 239
Modernizing the U.S. Census 4   Lee (1993:1) notes that the wording changed from ''census" to "enumeration" during the course of the Constitutional Convention. She could find no reason for the change but speculates that the negative references to censuses in the Bible may have been a reason (see also Anderson, 1988:10). However, the phrase "census or enumeration" appears in article I, section 9, of the Constitution, which deals with the levying of direct taxes. Lee (1993:2) notes that, in the dictionaries of the time, "enumerate" had the meaning of "to reckon up singly," and "enumeration" had the meaning of "the act of counting over." 5   These cases include Young v. Klutznick (1980, Eastern District, Michigan), City of Philadelphia v. Klutznick (1980), and City of New York v. U.S. Department of Commerce (1990, Eastern District, New York). Cuomo v. Baldrige (1987, Southern District, New York) also implicitly assumes that an adjustment would be constitutional. 6   In work commissioned by the panel, Barnett (1993) and Issacharoff (1993) agree with Lee (1993) that court decisions to date allow for the use of statistical methods, including sampling, as part of the census process. They go even further to offer arguments whereby the Constitution, which allows Congress complete discretion to determine the method for taking the census, might be interpreted to permit the use of methods that do not involve any physical enumeration at all. In their view, the key legal requirement is that the population figures be obtained by the most reliable methods possible. 7   The 1970 national vacancy check involved a resurvey of 13,500 housing units originally classified as vacant (about 0.2 percent of all such units). On the basis of the findings, imputation procedures were used to reclassify 8.5 percent of all vacant units as occupied and to impute persons to these units, amounting to 0.5 percent of the total population count. The postenumeration post office check was conducted in rural areas of 16 southern states. The Postal Service checked the address lists developed by enumerators for completeness, and Census Bureau staff followed up a sample of missed addresses in the field. On the basis of the results, census records were imputed for 1.3 percent of the population in areas, representing 0.2 percent of total U.S. population (see Citro and Cohen, 1985:189-193). 8   Until 1911, with one exception, Congress increased the size of the House of Representatives at each reapportionment, so that no state lost congressional seats. The exception occurred when the size of the House was decreased from 242 members following the 1830 census to 232 members following the 1840 census. In 1911, the size of the House was fixed at 435 seats. 9   See Anderson (1988: Ch.6) for a review of reapportionment and redistricting history and the relationship to the census. 10   One commentator (Baker, 1986:275-276) claims that a majority of the Supreme Court no longer truly supports the ideal of strict mathematical equality for congressional districts but has felt constrained by precedent. He argues that Congress should pass legislation that would permit a reasonable degree of population

OCR for page 239
Modernizing the U.S. Census     variance among districts and require other desirable criteria, such as compactness and contiguity. 11   This decision effectively abolished systems that assigned a minimum number of seats to each county or other jurisdiction no matter how small in size, although one exception occurred for a Wyoming state legislative redistricting plan. On the day the Supreme Court issued Karcher v. Daggett (1983), it also issued Brown v. Thomson (also a 5 to 4 decision), which upheld a plan that gave a seat to one county that was significantly smaller in size than the other districts. The court accepted the state's argument that this deviation from the ideal was justified to permit one isolated county to retain the seat it had been granted in 1913. Two members of the majority issued an opinion saying that they agreed with the decision only because the deviation from population equality involved just one county (Parker, 1989:61). 12   The additional items provided in the 1990 P.L. 94-171 file were available from the 1980 census as well but only in data files that were released at a later date than the 1980 P.L. 94-171 file; hence, they were not available for redistricting on a timely basis. 13   Previously, the Department of Commerce had released adjusted census data for the nation, states, counties, and cities; it had also released to Congress adjusted data for half of the census blocks. 14   In Davis v. Bandemer (1986), the Supreme Court held that partisan political gerrymandering was subject to judicial review, but it set a high standard for successfully bringing such a case, stating that plaintiffs must do more than show that a redistricting plan makes winning elections more difficult (Durbin and Whitaker, 1991:13-14). 15   This discussion of redistricting practices draws heavily on information provided to the panel staff by Kimball Brace, president of Election Data Services, Inc., Washington, D.C. See also Ehrenhalt (1983) for an overview of the redistricting experience following the 1980 census; and Henry (1989), who notes that half the states used computers for tabulation purposes in the 1980 redistricting cycle and that (as of 1989) several firms were ready with integrated computerized tabulating and mapping systems for the 1990 cycle. 16   Ethnic origin in the context of redistricting generally refers to Spanish or Hispanic origin. In addition to total population and information on race and ethnicity, many states want data on the voting-age population. These data were added to the P.L. 94-171 data file in 1990, and 28 states reported using them (National Conference of State Legislatures, 1992:8). The requirements for the census as the source of such data are not considered further here. It can well be argued that age—and sex—are essential characteristics to obtain simply to establish the existence of an individual and to help ensure that people are not double-counted. 17   Parker (1989:59-63) notes that challenges to redistricting plans on the grounds that they are racially discriminatory can be brought under section 2 of

OCR for page 239
Modernizing the U.S. Census     the act as well as the more frequently invoked section 5. Hence, although the preclearance provisions of section 5 currently apply to fewer than half the states (in some cases, just to selected jurisdictions in the state; see Laney, 1992), all states must worry about the racial composition of legislative districts if they are to avoid challenges under the Voting Rights Act. 18   See Davidson (1992), Grofman et al. (1992), Karlan and McCrary (1988), Kousser (1992), and Thernstrom (1979, 1987) for discussions of the history and implications of the Voting Rights Act from different perspectives. 19   Some analysts involved with redistricting efforts to comply with the Voting Rights Act have argued that it is also important to provide small-area data on citizenship from the census as part of the P.L. 94-171 data files (e.g., see Gobalet and Lapkoff, 1993). They find that citizenship rates among the Hispanic population vary greatly across small areas. Because voters must be citizens, legislative districts may not adequately represent the potential voting strength of the Hispanic population unless citizenship status is known. To provide data on citizenship in the P.L. 94-171 files would likely require asking for this information on the census short form. REFERENCES Alexander, C. 1993 Untitled [notes on continuous measurement designs]. Draft paper. Bureau of the Census, U.S. Department of Commerce, Washington, D.C. Anderson, M. 1988 The American Census: A Social History. New Haven, Conn.: Yale University Press. Baker, G.E. 1986 Whatever happened to the reapportionment revolution in the United States? Pp. 257-276 in Bernard Grofman and Arend Lijphart, eds., Electoral Laws and Their Political Consequences. New York: Agathon Press, Inc. Barnett, L.D. 1993 Comments on Congressional Research Service Report for Congress: Legal Issues for Census 2000. Paper prepared for the Panel on Census Requirements in the Year 2000 and Beyond, Committee on National Statistics, National Research Council. School of Law, Widener University. Bryant, B.E. 1993 Decision of the Director of the Bureau of the Census on whether to use information from the 1990 Post-Enumeration Survey (PES) to adjust the base for the intercensal population estimates produced by the Bureau of the Census. Federal Register 58(1):69-78. Bureau of the Census 1976 Language Minority, Illiteracy, and Voting Data Used in Making Determinations for the Voting Rights Act Amendments of 1975 (Public Law 94-73). Current Population Reports, Population Estimates and Projections, Series P-25, No. 627. Washington, D.C.: U.S. Department of Commerce. 1982 1980 Census of Population and Housing, Users' Guide, Part A. Text. PHC80-R1-A. Washington, D.C.: U.S. Department of Commerce. 1983 Stakeholders' Conference on Public Law 94-171 Program. 1990 Planning Conference Series No. 3. Washington, D.C.: U.S. Department of Commerce.

OCR for page 239
Modernizing the U.S. Census 1992 1990 Census of Population and Housing, Guide, Part A. Text. 1990 CPH-R-1A. Washington, D.C.: U.S. Department of Commerce. Citro, C.F., and M.L. Cohen, eds. 1985 The Bicentennial Census: New Directions for Census Methodology in 1990. Panel on Decennial Census Methodology, Committee on National Statistics, National Research Council. Washington, D.C.: National Academy Press. Davidson, C. 1992 The Voting Rights Act: A brief history. Pp. 7-51 in Bernard Grofman and Chandler Davidson, eds., Controversies in Minority Voting: The Voting Rights Act in Perspective. Washington, D.C.: The Brookings Institution. Durbin, T.M., and L.P. Whitaker 1991 Congressional and State Reapportionment and Redistricting: A Legal Analysis. Congressional Research Service Report for Congress, 91-292-A. Washington, D.C.: U.S. Government Printing Office. Ehrenhalt, A. 1983 Reapportionment and redistricting. Pp. 44-71 in Thomas E. Mann and Norman J. Ornstein, eds., The American Elections of 1982. Washington, D.C.: American Enterprise Institute for Public Policy Research. Ericksen, E.P., L.F. Estrada, J.W. Tukey, and K.M. Wolter 1991 Report on the 1990 Decennial Census and the Post-Enumeration Survey, Appendix A: The Census Process. Report submitted by members of the Special Advisory Panel to the Secretary, U.S. Department of Commerce, Washington, D.C. Gobalet, J.G., and S. Lapkoff 1993 Voting Rights Act Issues in Political Redistricting. Paper prepared for the Population Association of America annual meeting, Cincinnati, Ohio. Lapkoff & Gobalet Demographic Research, Inc., Oakland, California. Grofman, B., L. Handley, and R.G. Niemi 1992 Minority Representation and the Quest for Voting Equality. Cambridge, Eng.: Cambridge University Press. Henry, C. 1989 The impact of new technology and new census data on redistricting in the 1990s. Pp. 67-74 in William P. O'Hare, ed., Redistricting in the 1990s: A Guide for Minority Groups. Washington, D.C.: Population Reference Bureau, Inc. Herriot, R.A., D.V. Bateman, and W.F. McCarthy 1989 The decade census program—a new approach for meeting the nation's needs for subnational data. In Proceedings of the Social Statistics Section. Alexandria, Va.: American Statistical Association. Horvitz, D.G. 1986 Statement to the Subcommittee on Census and Population, Committee on Post Office and Civil Service, House of Representatives, May 1. Research Triangle Institute, Research Triangle Park, N.C. Issacharoff, S. 1993 Comments on Congressional Research Service Report for Congress: Legal Issues for Census 2000. Paper prepared for the Panel on Census Requirements in the Year 2000 and Beyond, Committee on National Statistics, National Research Council. School of Law, University of Texas, Austin. Karlan, P.S., and P. McCrary 1988 Book review: Without fear and without research: Abigail Thernstrom on the Voting Rights Act. The Journal of Law & Politics 4(4):751-777. Kish, L. 1981 Population counts from cumulated samples. Pp. 5-50 in Congressional Research Service, Using Cumulated Rolling Samples to Integrate Census and Survey Operations of the

OCR for page 239
Modernizing the U.S. Census Census Bureau. Prepared for the Subcommittee on Census and Population, Committee on Post Office and Civil Service, House of Representatives. Washington, D.C.: U.S. Government Printing Office. 1990 Rolling samples and censuses. Survey Methodology 16(1):63-79. Kominski, R. 1985 Final Report—Documentation of Voting Rights Act Determinations. Unpublished memorandum to Paul Siegel. Bureau of the Census, U.S. Department of Commerce, Washington, D.C. 1992 1992 Voting Rights Act Bilingual Ballots Determinations. Unpublished memorandum for the record. Bureau of the Census, U.S. Department of Commerce, Washington, D.C. Kousser, J.M. 1992 The Voting Rights Act and the two Reconstructions. Pp. 135-176 in Bernard Grofman and Chandler Davidson, eds., Controversies in Minority Voting—The Voting Rights Act in Perspective. Washington, D.C.: The Brookings Institution. Laney, G.P. 1992 The Voting Rights Act of 1965, As Amended: Its History and Current Issues. Congressional Research Service Report for Congress, 92-578-GOV. Washington, D.C.: U.S. Government Printing Office. Lee, M.M. 1993 Legal Issues for Census 2000. Congressional Research Service Report 93-177-A. Washington, D.C.: U.S. Government Printing Office. National Conference of State Legislatures 1992 Customer Feedback for the 1990 Census Redistricting Data Program—A Preliminary Summary. Prepared for Discussion at the National Conference of State Legislatures Annual Meeting, Cincinnati, Ohio. NCSL Reapportionment Task Force 1989 Reapportionment Law: The 1990s. Denver, Colo.: National Conference of State Legislatures. O'Rourke, T.G. 1980 The Impact of Reapportionment. New Brunswick, N.J.: Transaction Books. Panel on Census Requirements in the Year 2000 and Beyond 1993 Planning the Decennial Census: Interim Report. Committee on National Statistics, National Research Council. Washington, D.C.: National Academy Press. Parker, F.R. 1989 Changing standards in voting rights law. Pp. 55-66 in William P. O'Hare, ed., Redistricting in the 1990s: A Guide for Minority Groups. Washington, D.C.: Population Reference Bureau, Inc. Romig, C.L. 1983 Evaluation of the 1980 Census and the Legislative Reapportionment Process. Denver, Colo.: National Conference of State Legislatures. Thernstrom, A.M. 1979 The odd evolution of the Voting Rights Act. The Public Interest 55(Spring):49-76. 1987 Whose Votes Count? Affirmative Action and Minority Voting Rights. Cambridge, Mass.: Harvard University Press.