The first, critical stage of the 2000 census is over. Enumerators have finished knocking on doors and turned in their ID badges along with their tallies. Despite widespread complaints about intrusive questions, especially those concerning race and ethnicity, cooperation with the decennial census was remarkably high, and the overall effort appears to have been surprisingly successful, certainly compared to enumerations in 1980 and 1990.
Yet, as we now await the December 31 deadline for the Census Bureau to provide state population totals for congressional reapportionment, controversy will inevitably arise again. On this latest census, for the first time ever, individuals were permitted to identify themselves as belonging to more than one race. But it is yet to be determined how the Bureau will tabulate such responses. For example, how is someone who identifies herself as black and Asian as well as white to be counted? Civil rights groups fear that such individuals will not be considered minorities and that minority population totals will consequently decline and minority group solidarity be undermined.
Even more controversial will be the persistent problem of the minority undercount. For over 50 years we have known that the census disproportionately overlooks blacks and other minorities. In 1990, for example, the census missed about 5 percent of Hispanics and 4.4 percent of African Americans, compared to 0.7 percent of non-Hispanic whites. No matter how successful the 2000 census proves to be, it almost certainly will not overcome this gap.
To compensate for these uncounted minorities, it has been argued that census results be statistically “adjusted,” an approach supported by the Census Bureau itself. And while the Supreme Court ruled last year that adjusted census numbers could not be used to apportion Congress, it left open whether such numbers could be used for funding formulas and redistricting. With the Bureau now planning to produce two sets of numbers, both adjusted and unadjusted, continued wrangling over an otherwise successful census is assured.
Predicting the Unpredictable
Such controversies over the constitutionally mandated decennial census are of course important in their own right. But they are also useful as windows on contemporary politics and policymaking, especially with regard to racial minorities.
Although the minority undercount is real, its fiscal and political implications have been grossly overstated. Journalists are fond of repeating that $185 billion in federal funds are distributed annually on the basis of formulas that rely on census data. Yet in 1998, according to the General Accounting Office, just $449 million of that total (only 0.33 percent) would have been shifted as a result of adjusting the 1990 census.
How can this be? First, population is only one of several factors in federal grant formulas. Indeed, programs designed to help distressed communities often reduce funding when population increases. Second, to the extent that population gains do lead to grant increases, the critical factor for a given jurisdiction is not merely its absolute gain but its gain relative to other jurisdictions. In fact, many jurisdictions could register population increases from adjustment but end up worse off than they would have been without adjustment because other jurisdictions would experience even greater increases. Third, because such grant programs typically have funding ceilings, adjustment would result in a fixed pie divided up among more people. Per capita grants would actually decrease.
The importance of the undercount has also been exaggerated when it comes to congressional redistricting. The fact is, the partisan consequences of adjustment are impossible to predict. It is almost universally assumed that adjustment would advantage Democrats and disadvantage Republicans, on the theory that undercounted minorities are more likely to be Democrats. But the drawing of district lines in 50 different states is subject to myriad local political factors; quite aside from the demographic vagaries, nobody knows how each of these political scenarios would play out under adjustment. And weakened partisan loyalties would make the process less of an exact science than ever.
As Tom Hofeller, former staff director of the House Subcommittee on the Census, put it in 1989: “The gerrymander overcometh all. What demographics give, legislatures can take away in the dead of the night.” Then too, during the 1990 redistricting, Democrats controlled both legislative chambers and the executive mansions of 19 states, Republicans only three. Today the situation is significantly changed: Republicans dominate 14 states, Democrats only eight. In the states under their control, Republicans need not be threatened by additional minorities resulting from adjustment. After all, these could be packed into “majority minority” districts—a tactic that in the recent past has helped both minorities and Republicans, at the expense of nonminority Democrats. Republicans might also mitigate the impact of additional minorities by dispersing them among several districts.
Such uncertainty hardly gratifies House Republicans holding on to a slim majority. But this does not mean that adjustment necessarily benefits Democrats.
Sampling for dollars
For such unpredictable and relatively modest stakes, statistical adjustment of the census entails substantial but largely ignored risks. Characterized by opponents and proponents alike as “sampling,” adjustment actually bears little resemblance to the survey sampling and opinion polling to which advocates routinely liken it. To be sure, adjustment would rely on a huge sample survey (about 300,000 households) undertaken after the actual census. But the uses to which this post-census survey would then be put are quite different from what is commonly understood as sampling.
Typically, survey sampling extrapolates upward from a part to draw inferences about the whole. But this is not the case with sampling for census adjustment, which basically involves two “samples”: the original census and the post-census survey. In essence, census adjustment extrapolates sideways by matching the post-census survey responses of several hundred thousand individuals against the responses of those same individuals to the actual census several weeks earlier. This matching process is extremely complex and highly prone to error. For example, locating individuals who move between these two points in time is a major challenge.
Even the Census Bureau acknowledges that this facet of adjustment introduces substantial new error into census counts. Indeed, in 1991 the Bureau published adjusted numbers that were never actually used. About a year later it was announced that a computer coding error had improperly added about one million persons to those adjusted numbers. Had that set of figures been used to adjust the census in 1991, only to be deemed incorrect a year later, public confidence in the census would have been undermined.
Moreover, the sideways extrapolation that census adjustment entails compels statisticians to rely on assumptions and models that are invariably subject to challenge. Such complexities again sharply distinguish adjustment from conventional sampling methods. In other words, adjustment is hardly the straightforward scientific exercise that its proponents argue will provide reliable and objective answers to vexing political dilemmas.
But if the stakes of the undercount are so minimal and the risks of census adjustment so high, why are we tempted by such a bad bargain? First, we are told repeatedly by the Census Bureau that experts agree that adjustment is the way to go. And while it is true that many, even most, statisticians believe that the undercount should be fixed by means of adjustment, there is hardly a scientific consensus on this point. Indeed, as statisticians Thomas Belin and John Rolph write in Statistical Science: “We are far from sure that consensus is attainable on census adjustment.”
More to the point, census adjustment is hardly just a technical, scientific matter. Adjustment will create winners and losers; some groups and jurisdictions will gain from adjustment, others will not. And because such outcomes will depend on technical assumptions and logistical decisions about which reasonable people can and will disagree, the losers will have ample and justified opportunity to create controversy.
But none of this should surprise us. As originally conceived in Article 1, Section 2 of the Constitution, the census was an inherently and properly political undertaking designed to levy taxes as well as to apportion representation among the states. While the former purpose would encourage states to minimize their population counts, the latter would work in the opposite direction. Thus, the Framers relied on politics as the very mechanism by which an impartial count would be achieved.
The Right to Be Counted
Aside from the claims of scientific expertise, adjustment tempts us because we don’t want to say no to disadvantaged racial minorities, especially when such an apparently straightforward remedy is at hand. This temptation is all the greater because of the symbolic clout of the undercount for minorities, especially African Americans, who routinely associate today’s undercount with the Constitution’s original stipulation that slaves be counted in the census as three-fifths of a person.
Nevertheless, census adjustment may not be as salient to minorities as many observers believe. Minority advocates and organizations undeniably favor adjustment, but they have not invested heavily in it. Most of the political energy and resources advocating adjustment have come from state and local jurisdictions that are, to be sure, responding to minority claims but that have other reasons to push adjustment—not the least of which is that the Census Bureau is a weak federal agency, criticism of which involves no risk and the possibility of some additional resources, however exaggerated.
Then, too, saying no to minorities on adjustment is even more difficult when the issue is framed in terms of rights. Adjustment advocates have characterized being uncounted in the census as equivalent to being disenfranchised. Many have asserted “a right to be counted,” while still others argue that the undercount is a civil rights issue.
Yet far from being a right, participating in the census is a legal obligation. And while not being counted in the census is regrettable, it in no way prevents an individual from voting or organizing others to vote.
Finally, census adjustment is tempting because of the uneasy feeling that we are not sure exactly what can or should be done to aid or otherwise empower disadvantaged minorities. Yet in an era when voting has reached historically low levels, especially among minorities, we have come to rely on scientific expertise, rights-oriented advocacy, and administrative means to achieve political ends. Sometimes these work; often not. But in the controversy over the undercount, we have become so focused on statistical outcomes at the Census Bureau that we have lost track of the fact that empowerment ultimately comes from developing organizational and political strength at the grassroots.
Not only does adjustment ignore this truism, it actually eliminates one more incentive for the disadvantaged to connect up with society’s dominant institutions. For why would anyone bother to “make themselves count” if the Census Bureau, while urging Americans to cooperate with it, also indicated that adjustment would make up for their failure to do so?
One final irony. Despite the rhetoric of science and rights, adjustment would not of course eliminate politics from the census. The resulting politics would just be of a different order: arcane and technocratic. The process would be as open to view as a glass-partitioned, climate-controlled computer center. But the vast majority of us, especially the most disadvantaged and least sophisticated among us, would be on the outside looking in—rendered less able than ever to mobilize and challenge critical technical decisions affecting our communities. It is difficult to believe that the result would be other than increased minority disaffection and cynicism.
Peter Skerry is a nonresident senior fellow at the Brookings Institution and teaches political science at Claremont McKenna College. He is the author of Counting on the Census?: Race, Group, Identity, and the Evasion of Politics, from which this article is adapted.