This essay examines briefly some of the problems inherent and issues involved in the census of 2000, notably the interstate problem of apportioning congressional seats on the basis of an admittedly inexact enumeration of the nation's population and the subsequent intrastate problem of drawing district lines for both congressional and state legislative seats. Beyond the political ramifications, the census methodology used in 2000 may also signify major implications for the allocation of federal benefits for the first decade, if not longer, of the twenty-first century.

The federal census of 1990 has the dubious distinction of being the first modern census to be less accurate than its predecessor. With an admitted undercount of at least four million persons - in reality, an estimated undercount of 8.8 million less an estimated 4.4 million "double-counted" individuals, the 1990 census was reportedly the most inaccurate headcount since post-census evaluations began in 1940 (Pope 1998). 

The U.S. Constitution requires a census of the population every ten years to determine representation in the U.S. House of Representatives: members are to be "apportioned among the several states . . . according to their respective numbers." Census counts, however, have been less than perfect since the first enumeration in 1790. In that year, Thomas Jefferson, who as secretary of state administered the census, thought the nation's population had been seriously undercounted. Ironically, Jefferson reportedly was among those counted twice, once in Philadelphia and again at Monticello (U.S. News 1998). Further, the actual method of allotting seats among the states has been controversial from the beginning: the first presidential veto in 1792 was of a bill "for an apportionment of Representatives among the several states, according to the first enumeration" (Kimmitt 1978).

Virtually every census has generated complaints and partisan disagreements, but apportionment of House seats proved to be of little difficulty during the nineteenth century. As new states were admitted to the Union, Congress simply increased the size of the House to accommodate the expanded population. Following the 1910 census, however, Congress fixed the size of the House at 435 (excepting the temporary increase to 437 for the 1960 election to accommodate the admission of Alaska and Hawaii). But Congress failed to pass a new apportionment act following the 1920 census, the first to indicate a majority of the nation's residents lived in urban areas. These results were "so devastating to the hinterlands that the House went through the next decade in a state of official denial" (Elving 1997, 1478), and powerful members from rural areas prevented a redistribution of districts among the states. 

In 1929, Congress eventually formulated an "automatic" reapportionment plan that required the Census Bureau to calculate the seats assigned to the states. After each state receives one seat as constitutionally required, the Census Bureau employs the formula of equal proportions to allocate the remaining 385 seats in succession on a priority basis. Within the fixed number of 435 seats under this apportionment method, a state can gain population but actually lose representation if its percentage increase is less than that of the nation as a whole. 

No census calculation formula has escaped criticism, and the equal proportions method has engendered continued controversy. The initial census conducted under the system signified a serious degree of population undercounting. The number of men reporting to Selective Service in 1941 proved substantially greater than the number expected on the basis of the 1940 census (Pope 1998). In the 1990s, several states have challenged census procedures. As one example, Montana argued unsuccessfully that the equal proportions method did not achieve the greatest possible equality (U.S. Dept. Of Commerce v. Montana, 1992). 

Moreover, as Supreme Court decisions have required greater mathematical equality in districting at all levels of government, minority populations and urban officials have expressed additional concerns over possible undercounts. Their concerns have gone beyond loss of numerical representation in Congress and in state capitals to include reductions in both federal and state funding, declining urban tax bases, and possible shifts in the balance of political power to less favorably disposed public officials (Heslop 1980).

A census is a snapshot of the population at one point in time when a government attempts to count its residents and gather information about them. In the case of the United States, this point in time is April 1 of the closing year of each decade. As is usually the case, the 1990 snapshot was somewhat out of focus: the census missed approximately four million residents, or an estimated 1.8 percent of the population. The 1990 census tally was based on returned questionnaires and follow-up canvasses plus a post-enumeration assessment, employing sampling techniques. This approach produced two conflicting sets of population figures. The results of the sampling suggested a significant undercount, particularly in larger cities and among African -Americans, Hispanics, Asians, and Native Americans. 

In the aftermath of the controversy surrounding the "two-number" 1990 census, Congress adopted the Decennial Census Improvement Act of 1991. This law directed the National Academy of Sciences to determine means by which the most accurate census count could be achieved, including the appropriate use of sampling methods in the acquisition or refinement of population data. On the basis of this law (and perhaps influenced by the 1994 Republican takeover of the House of Representatives), the Clinton administration proposed supplementing conventional census methods in 2000 with statistical sampling. This action would move the dispute over census methodology into the political arena.

Citing the conclusion of an independent panel that "it is fruitless to continue trying to count every last person with traditional census methods of physical enumeration" (Gruenwald 1997, 1435), the Census Bureau initiated plans to reach 90 percent of households, either by mailed questionnaires or personal contact with nonrespondents; make population estimates for the remaining 10 percent of nonresponding households from these replies; and conduct a second random sampling of 750,000 households nationwide to validate its findings and make any necessary adjustments.

Supporters of statistical sampling argued it would decrease the cost of the census in 2000 (estimated at $4 billion) but, more importantly, reduce the customary undercounts and improve its accuracy. Opponents not only challenged claims that sampling would provide a more accurate census but also contended the procedure was unconstitutional: the U.S. Constitution demands an "actual Enumeration" for apportionment purposes. 

Publication of the Census Bureau's plan to conduct the 2000 census prompted two separate legal challenges. On February 12, 1998, the Southeastern Legal Foundation, identified as a conservative group from Atlanta (Dao 1999a), filed suit claiming the planned use of sampling violated both the Constitution and the Census Act, which had been amended in 1976 specifically banning the use of sampling for purposes of apportioning members of Congress (Glavin v. Clinton, 1998). On February 20, congressional Republicans sued the Commerce Department, also claiming the proposed use of sampling for apportioning House seats violated both the Constitution and the Census Act (U.S. House of Representatives v. Dept. of Commerce). In each case, a lower court panel ruled unanimously in favor of the plaintiffs. 

The U.S. Supreme Court expedited the administration's appeal of these decisions, hearing oral arguments November 30 and handing down its decision January 25, 1999. The Court dismissed the House suit but ruled 5-4 in the administration's appeal of the Glavin case "that the Census Act prohibits the proposed uses of statistical sampling in calculating the population for purposes of apportionment" (Clinton v. Glavin, 1999). Having ruled on the basis of statutory law, the Court found it "unnecessary" to consider whether the use of sampling also violated the Constitution. Four justices in the majority suggested that sampling would be unconstitutional even if permitted by law; the four dissenters held it was not a constitutional problem; and Justice Sandra Day O'Connor, who delivered the majority opinion, expressed no view (Greenhouse 1999).

To some observers, the decision was less than definitive. While it barred sampling for apportionment, a White House spokesman suggested the ruling was a "limited decision" that affirmed the use of scientific sampling for other purposes, such as allocating federal funds to state and local governments (Carelli 1999). Others indicated the decision appeared to permit state legislatures the option of using statistically-adjusted census data for drawing district lines (Dao 1999b). However, a redistricting expert from the National Conference of State Legislatures cautioned that providing two sets of population data would create immense political headaches for state legislatures and place them in a very difficult position: "They would have to choose one . . . and whichever unit they use, they will be sued" (Dao 1999a).

Confirmation that the partisan struggle over the census of 2000 would continue came less than a month following the Supreme Court decision. The Census Bureau unveiled its revised plan February 24 with Director Kenneth Prewitt announcing that the agency would, in effect, produce a "two-number" census. First, the bureau would conduct the census in accordance with the Court's ruling, striving to count each of the nation's residents to provide figures by December 31, 2000, for congressional apportionment. Then, employing a statistical sampling method it labeled the Accuracy and Coverage Evaluation, the bureau would attempt to develop a more accurate population count by April 1, 2001, for states to use in drawing congressional and state legislative district lines and for distribution of federal funds (Schmid 1999).

Each census produces changes in the allotment of congressional seats among the states. For example, after the 1970 census, 11 seats changed hands among 14 states, with five states gaining seats and nine losing seats. Following the 1980 headcount, 17 seats switched among 21 states, with 11 gaining and 10 losing seats. On the basis of the 1990 census, 19 seats were reallocated, with eight states gaining and 13 losing seats. Over this time frame, the big gainers have been California (+14), Florida (+11) and Texas (+7), while Arizona (+3), Colorado (+2) and Washington (+2) have also benefitted. The principal losers have been New York (-10), Pennsylvania (-6), Ohio (-5), Illinois (-3), and Michigan (-3). Iowa, Massachusetts, New Jersey and West Virginia each lost two seats during this period. These shifts have materially benefitted Republican successes in gaining congressional seats.

Projected shifts of congressional seats following the census of 2000 indicate a slowing in the number of changes, although the direction of the shifts - from the northeast to southern and western states - would continue. Interestingly, the distribution of seats based on the estimated 1998 populations favors the Sunbelt states, while the allocation of seats based on projected 2000 populations benefits states in the "interior West" - Colorado, Montana, Nevada and Utah. In either case, Montana is expected to regain the seat it lost in 1990, which would make it "the first state whose delegation size was reduced to a single Representative to do so" (Huckabee 1998, 4).

In considering the possible loss of seats, both the 1998 population estimates and the 2000 population projections are in agreement that New York and Pennsylvania will continue their downward spirals, with each losing two seats; further, Connecticut, Mississippi, Oklahoma, Ohio, and Wisconsin each will lose one seat. However, where the 1998 population estimates foresee no changes in either Illinois or Michigan, the 2000 population projections indicate each state will lose another seat in Congress. Illinois is considered particularly vulnerable, retaining the 435th seat on the priority list by just over 33,000 residents.

Drawing boundary lines for congressional and legislative districts is never a simple task. In 2001 this assignment will be complicated more than usual by several factors. First, the possibility exists, in the absence of a definitive judicial ruling to the contrary, that legislatures will have two sets of population figures to work with in drawing district lines. As a redistricting expert has noted, "There is no bright-line legal precedent to help. And whichever unit [legislatures] use, they will be sued" (Dao 1999a).

A second factor complicating the legislatures' assignment is that statistically-estimated figures are expected to increase city populations substantially - at the expense of surrounding suburban areas. Demographers project sampling could add at least 250,000 to the population of New York City, suggesting the city could lose no congressional seats and possibly gain seats in the state legislature - while suburban Nassau and Suffolk Counties on Long Island would lose representation on both levels (Dao 1999b). Likewise, in Chicago, where the estimated 1990 undercount was as much as 159,000, sampling could add up to 200,000 to the city's population in 2000. This figure would help Chicago retain its present congressional allotment and increase its delegation in the Illinois legislature (Greenburg 1999).

Similar situations exist in California, Florida and Texas. In California, observers claimed a 1990 undercount of nearly a half-million people, mainly in Southern California, cost the state one House seat, and $500 million in federal funding; a failure to employ sampling in 2000 could deny the state two additional seats (Los Angeles Times 1999). In Florida, observers calculated the state's 1990 undercount at 258,000, with much of it in heavily Hispanic Miami-Dade County (Miami Herald 1999). In Texas, where more than 486,000 residents - 2.8 percent of the state's population - were not counted in the 1990 census, the Republican state administration filed a brief in the Glavin case in favor of sampling (Camia 1999).

On the other hand, while advocates of sampling see urban populations adjusted upward by use of this technique, recent Census Bureau projections of population growth suggest a contradictory circumstance. Figures released in March 1999 indicate the nation's fastest-growing areas are "counties around large cities," not necessarily the city or the metropolitan county itself (Census Bureau, 11 March 1999). Should the actual enumeration confirm this forecast but adjusted figures on the basis of sampling techniques indicate greater urban populations, the districting conflicts facing state legislatures in 2001 may surpass any confrontations experienced in the past.

Finally, several states have yet another complicating factor: the necessity to draw majority-minority congressional and legislative districts, creating constituencies favorable to the election of African Americans and Hispanics. Observers are aware of the major controversies throughout the 1990s over districting in the states of Georgia, Louisiana, North Carolina, and Texas, and the lesser conflicts in Florida, Illinois, New York, and Virginia. Although the Supreme Court has finally ruled that states cannot engage in "racial gerrymandering" (Abrams v. Johnson 1997), it will not permit systems which deny minorities an equal opportunity to participate in the political process and elect representatives of their choice. The probability of gains in seats for Florida, Georgia, and Texas and the likelihood of losses of seats in Illinois and New York suggest the certainty that these struggles will continue in the first decade of the next century.

Gaining political advantage is, of course, at the heart of this conflict, with control of the House of Representatives and state legislative chambers the immediate goal. Democrats anticipate statistical sampling to increase the count of metropolitan, especially inner-city, residents, improving their chances of winning congressional and legislative seats. Republicans fear a loss of as many as 24 House districts and control of several state legislatures if the Census Bureau is permitted to employ sampling techniques (Pope 1998; Dao 1999a). The implications for racial and ethnic minority representation are also at issue, and not only in those states under close judicial scrutiny during the 1990s. Further down the road, of course, is the distribution of Electoral College votes in 2004 and beyond.

Not without consideration, however, is the distribution of at least $180 billion (1996 figures) in federal funding based on population. At least one supporter of statistical sampling argues that apportionment of House seats is "probably the least important aspect of the census for most Americans [who] have a real stake" in how federal funding is allocated (Greenburg 1999, 1:8).

Others disagree as to the ramifications of a census undercount, either political or economic. One observer suggests that a reordering of House seats in 1990 based on statistical sampling would have shifted only three districts, with California, Georgia, and Montana benefitting at the expense of Oklahoma, Pennsylvania, and Wisconsin. While the implications for these states are obvious, the overall effect would have been relatively modest. Further, he suggests that possible population errors would result in only minimal shifts in the distribution of federal funds to state and local governments, perhaps three-tenths of one percent, or $600 million out of possibly $200 billion in federal spending (Samuelson 1999).

The problems are apparent, but a solution is not. Republicans have criticized the potential "two-number" census as "hypocritical and confounding" and "a recipe for disaster" and promised to reduce Commerce Department funding unless the proposal is rescinded. On the other hand, supporters of statistical sampling have announced plans to sue, if necessary, to force the Census Bureau to use sampling at least for distributing federal funding to state and local governments. At the midpoint of 1999, the only certainty is that the debate over "actual Enumeration" versus statistical sampling as the appropriate method to determine the nation's population will continue.

Camia, Catalina. 1999. "Supreme Court bars census sampling plan." Dallas Morning News, Jan. 26 (www.dallasnews.com).

Carelli, Richard (AP). 1999. "Court: Census Can't Use Estimates." Washington Post, Jan. 25 (www.washingtonpost.com).

Dao, James. 1999a. "Census Ruling Reignites a Partisan Battle." New York Times, Jan. 27: A19.

Dao, James. 1999b. "Split Decision Sets Stage for Battles." New York Times, Jan. 26: A20.

Elving, Ronald D. 1997. "Rarely a Consensus on the Census." Congressional Quarterly Weekly Report, June 21: 1478.

Greenburg, Jan Crawford. 1999. "Sampling for census restricted." Chicago Tribune, Jan. 26: 1:1, 8.

Greenhouse, Linda. 1999. "In Blow to Democrats, Court Says Census Must Be by Actual Count." New York Times, Jan. 26: A1, A20.

Gruenwald, Juliana. 1997. "Statistics Stir the Passions of Politics." Congressional Quarterly Weekly Report, June 21: 1434-36.

Heslop, Alan. 1980. "The Redistricting Tangle." American Demographics, May: 25-34.

Huckabee, David C. 1998. House Apportionment Following the 2000 Census: Preliminary Projections. Washington: Congressional Research Service, The Library of Congress.

Kimmitt, J. S. (compiler). 1978. Presidential Vetoes, 1789-1976. Washington: Office of the Secretary of the Senate, U.S. Government Printing Office.

Los Angeles Times. 1999. "Backward Step for Census," Jan. 26 (www.latimes.com).

Miami Herald. 1999. "Undercounting noses," Jan. 27 (www.miamiherald.com).

Pope, Charles. 1998. "Census: A Political Calculation." Congressional Quarterly Weekly Report, July 11: 1873-76.

Samuelson, Robert J. 1999. "Politics and the Census." Newsweek, Feb. 8: 48.

Schmid, Randolph E. (AP). 1999. "Criticism greets announcement of 2 census counts." Indianapolis Star, Feb. 25: A5.

U.S. Department of Commerce. 1999. "Fastest-Growing Counties Are Southern, Western and Predominantly Metropolitan, Census Bureau Reports," March 11 (news release).

U.S. News and World Report. 1998. "Justices to study census sampling," Dec. 7: 14.

Abrams v. Johnson, 117 S.Ct. 1925 (1997).

Clinton v. Glavin, No. 98-564 (Jan. 25, 1999).

Glavin v. Clinton, 19 F.Supp.2d 543 (1998).

U.S. Department of Commerce v. Montana, 503 U.S. 442 (1992).

U.S. House of Representatives v. Department of Commerce, 11 F.Supp.2d 76 (1998).