Why People Fight So Much About the Census
By Michael Teitelbaum and Jay Winter
After years of increasingly partisan wrangling, a special panel of three federal judges last week blocked the U.S. Census Bureau from using statistical sampling techniques to estimate for purposes of congressional apportionment the complete count of the U.S. population in the year 2000 census. The decision will surely be appealed to the Supreme Court and continue to resonate in Congress, but pending final resolution it might well be asked why an arcane question of technique and much else about the census has become such a political hot potato.
The answer lies in the energetic politics that have surrounded the collection and use of U.S. demographic data, especially since the 1960s. It involves some of the most contentious political issues of this period: the distribution of political power; the proper roles of race and ethnicity in public policy; and the trustworthiness of government. The debate over sampling makes clear how much real damage has been done to the credibility of the census, a bulwark of U.S. constitutional democracy.
It may now be time to consider new mechanisms to better insulate census issues from the partisan and interest-group politics that have been engulfing them.
The origins of this wrangle lie in the founding of the republic. The census was a critical element of perhaps the most difficult compromise needed to get agreement on the United States constitution in 1787: How would political power be apportioned among states with smaller and larger populations?
The Solomonic solution was a legislature consisting of two houses. In the Senate, each state would have equal representation irrespective of its population, so as to protect the interests of the smaller-population states. In the House of Representatives, seats would be based on each state's population as determined by a census, and reapportioned in the same way every 10 years, to protect the interests of states with larger or more rapidly growing populations.
Thus the U.S. constitution was not only the first written constitution, but also the first to base allocation of power on a regular census. This was a fundamentally political decision, but one that required that political considerations not be permitted to distort the basic data by which representation would be determined.
Yet politics found a way. While the census count determined the number of representatives from each state, the boundaries of electoral districts were determined by each state government. Clever politicians soon realized that boundaries could be drawn to political advantage. Massachusetts Gov. Eldridge Gerry achieved immortality of a sort when, in 1812, he contorted the state's district boundaries for partisan purposes; a political cartoon caricatured one of Gerry's state senatorial districts into the shape of a salamander, thereby memorializing the process as gerrymandering.
Public derision and judicial reversals caused such creative design to wane, but it was to rise again during the 1970s and 1980s, in the service of the higher political goals of civil rights and minority representation. The Voting Rights Act of 1965, adopted to end flagrantly discriminatory electoral practices in the Deep South, appropriately gave the Justice Department the right to review and to reverse states' redistricting decisions if these had the effect of scattering or otherwise minimizing the electoral impacts of minority voters.
This produced a new kind of creative boundary drawing. To increase the number of minority representatives, some state legislatures decided to concentrate geographically scattered minority residents into districts in which together they represented the voting majority. If this required truly strange boundaries, so be it.
The outcome was the creation of new electoral districts with complex shapes (a "snake," a "spider") that put Gerry's miserable salamander to shame. After extended litigation, the Supreme Court ruled in 1995 and 1996 that legislative districts purposefully carved out to increase congressional representation of minorities were unconstitutional if race was the predominant factor in their design.
Over this same period, race and ethnicity issues also became increasingly entangled with growing concerns about census undercounts, for the simple reason that analyses by the Census Bureau itself have shown that its counts are least complete for disadvantaged minority groups. For example, the 1970 undercount for blacks was estimated to be 6.5 percent, as compared with 2.2 percent for whites. And as affirmative action policies and identity politics grew during the '70s and '80s, the political importance of such undercounts increased dramatically, since affirmative action employment goals and targets and minority groups' perceived political clout were often based upon census data.
With even more at stake, minority advocates increasingly emphasized the importance of obtaining a more complete and accurate count of disadvantaged minorities, although a few seemed to have the cruder goal of maximizing the count.
But the undercount issue was not restricted to minority groups. The census undercount itself had moved by the '70s out of the technical domain to become a hot political issue. First, the Supreme Court during the 1960s began to require states to reapportion their own legislatures and congressional districts on the principle of "one man, one vote," meaning that sparsely populated rural areas could no longer maintain their overrepresentation in legislatures. Second, during the '50s and '60s, federal grants-in-aid to state and local governments grew rapidly, and were often allocated on the basis of formulas based on census data. Thus did the census count (and its undercount) come to be seen as synonymous with both power and money.
It was hardly surprising, then, that in the buildup to the 1980 Census, officials of cities with large minority populations began demanding that the 1980 count be adjusted for estimated undercounts. These proposals raised extraordinarily difficult technical and legal issues, and deeply divided responsible professionals involved in the census process. Would the survey data to be used for adjustment be any better than the census enumeration itself? Would it be correct to apply national or regional estimates of undercounts to quite small local areas?
In late 1980, after much internal debate, the Census Bureau director announced that the bureau would not "adjust the 1980 Census population to compensate for undercount unless directed by the courts." A flurry of lawsuits led by New York City ensued and the contradictory rulings and reversals that resulted demonstrated well the legal and technical ambiguities of the issues involved. Ultimately, the Supreme Court ruled that the Census Bureau had acted properly.
Undeterred, New York City (then joined by Los Angeles, Chicago, other cities, and the NAACP) filed its suit again after the secretary of commerce decided not to adjust the 1990 Census figures. The secretary acknowledged that, despite its best efforts, the census had failed to count 1.6 percent of the population, including about 4.8 percent of blacks and 5.2 percent of Hispanics at the national level, but concluded that adjusted figures would be less accurate than the original count. In March 1996, the Supreme Court, in an unusual unanimous ruling, rejected the claims .
It is in this context that the recent debate about census sampling a different technique than that debated earlier, and one supported by most statisticians has been conducted. Proponents argue that modern scientific sampling techniques that were unknown in 1787 would allow more accurate and complete data to be collected. Opponents argue that the constitution requires an "actual count" for apportionment of House seats, and that sampling techniques would allow political appointees at the Commerce Department to manipulate the sampling formulas to affect the basic data by which political power is apportioned.
On the issue of sampling, the 2000 Census appears to be on track for a major train wreck. However it is ultimately resolved, the debate has produced a perverse "lose-lose" situation: If sampling is disallowed by the courts, the data collected in 2000 without sampling will be seen as suspect by some. If sampling is allowed, the data's credibility will be viewed with suspicion by others.
The census politics of the past three decades and the current partisan row suggest that the credibility of future censuses so central to constitutional legitimacy may require oversight that is more insulated from the political fray. The Federal Reserve Board and the bipartisan Congressional Budget Office already provide such quality controls for economic, fiscal and monetary data. For census data, a permanent and non-political oversight panel might do the job; for example, a panel might be appointed by congressional leaders of both parties and the White House, but with all appointees drawn from a bipartisan list solicited from the National Academy of Sciences and from the established scientific societies in statistics, demography and economics.
Meanwhile, thoughtful observers might well ponder whether we have allowed the political stakes riding on census counts to rise too high to expect politically committed groups to restrain themselves from attempting to control the collection and tabulation of data. In demography, all numbers count. But in politics, some numbers count more than others.
Demographer Michael Teitelbaum and historian Jay Winter are co-authors of "A Question of Numbers: High Migration, Low Fertility, and the Politics of National Identity" (Hill & Wang).
© Copyright 1998 The Washington Post Company