April 1999 What's Next in the Census Sampling Case? By Douglas A. Samuelson As OR/MS Today reported in its hastily added update on the Census sampling case [February 1999], the Supreme Court decided to prohibit the use of sampling-based methods to correct for potential undercounts in the decennial census, at least for the purpose of apportioning representatives among states. As OR/MS Today also reported, however, more controversy is sure to follow. OR/MS analysts should find both the analytical issues and the political and legal implications interesting. The first noteworthy thing about the decision is that the Court was far from unanimous. Five Justices held against the Census Bureau's sampling-based plan, while four favored it. Upon careful reading, however, the Court was even more divided: The decision was not really 5-4, but more like 4-3-1-1. Justice O'Connor, writing for the majority, based the decision solely on her reading of the laws, refusing to reach any broader public policy questions. She carefully distinguished between the count for apportionment, which is specifically covered by the 1995 and 1997 Congressional Acts, and other uses of the count. Since the 1976 Act directs the Secretary of Commerce (of which the Census Bureau was and is part) to use all methods which are not provably invalid (sampling was the primary method the Congress had in mind), this decision can be viewed as a mandate to use sampling for all purposes other than apportionment among states, and the administration has already announced that it will view it this way. No other Justice fully concurred with Justice O'Connor's opinion. Justice Scalia wrote a concurring opinion raising broader grounds for excluding sampling; Justice Thomas fully concurred with him; and Justice Kennedy and Chief Justice Rehnquist wrote additional concurring opinions which substantially followed Scalia's. Justice Stevens wrote the primary dissenting opinion, with Justices Souter and Ginsburg mostly agreeing with him in short dissenting opinions of their own. Relying heavily on the 1976 Act, Justice Stevens argued that the most complete and accurate methods of counting best serve the Constitutional goal of equal representation, so sampling is allowable for all purposes. Justice Breyer also dissented, on somewhat narrower grounds. He contended that the legislation barred only sampling-based plans that replace rather than supplement attempts at full direct counting, so the administration's proposed method would be legal for all purposes. Fertile ground for analysis Where does this leave operations researchers? House Republicans, including the chairman of the relevant subcommittee, have indicated they will oppose any use of sampling and may seek to withhold funds for any sampling-based activity in connection with the decennial population census. Various groups representing cities and ethnic minorities filed amicus curiae ("friend of the Court") briefs in the action the Court just decided and might decide to bring new actions challenging any decision not to use the sampling-based adjustment. No matter what funds the Congress attempts to withhold, it seems nearly certain that the Census Bureau will produce at least some kind of sampling-based estimates of the errors in enumeration, and then the various parties will clash over what to do with these estimates. A critical quantitative question, often neglected in the legal controversy, is the actual effect of using or excluding sampling. Conventional wisdom holds that ethnic minorities in central cities and in highly rural areas are most likely to be undercounted by traditional enumeration, and that these people tend to vote Democratic, so sampling-based adjustments help the Democrats. In fact, however, the situation is more complicated. The people who are unlikely to be enumerated are also unlikely to vote; in fact, among those most likely to be missed are illegal immigrants, who would be ineligible to vote. Sampling-based adjustments, therefore, help states and districts within states which have larger enumeration-based undercounts, but not necessarily the party for which these undercounted individuals would have voted if they voted. California, Texas, Florida, Arizona and New Mexico are among the states likely to have large undercounts. Four of these five states have Republican governors and legislatures. Therefore, barring the sampling-based adjustment may end up depriving these states of House seats which would more likely have been Republican than Democratic. On the other hand, apportionment within states has historically swung more House seats from one party to the other than apportionment among states. When the Democrats captured the governorship of California in 1998, some political analysts predicted that this alone — assuming the Democrats could maintain their current hold on both houses of the California Legislature — could shift six to eight seats from the Republicans to the Democrats. This would be enough to return the Democrats to majority status in the House. Giving states the option of using sampling-adjusted numbers for internal reapportionment provides more ways for the party in power to juggle districts in its favor — and more basis for the other party to contest the result. A few analysts applying optimization and forecasting methods in key states are likely to prosper by helping to draw district boundaries in ways which favor one party or the other. Federal funds A closely related issue is the difference undercount adjustments would make in allocating federal funds. In early March, the General Accounting Office (GAO) released an estimate that for 15 major federal programs, less than 1 percent of the funds would be affected, with nearly half the reallocated money going to California and about another one-fourth going to Texas, Arizona and New Mexico. Again, which states get the increased or decreased funds is a different question from who actually benefits. This, too, is fertile ground for analysis. Finally, although the Supreme Court did not reach the broader issue in this case, it is part of a pattern of politicians increasingly belittling scientific expertise when they don't like the consequences of the expert advice. This behavior calls for increasing vigilance, in turn, by scientists and scientific organizations, and for pointed responses when appropriate. The American Statistical Association has been vocal in defending the scientific validity of sampling, while carefully avoiding taking sides on the political questions; if the OR/MS community does not choose to speak out on this issue, it should at least be prepared to defend its methods against political attacks in the future. Sources William Jefferson Clinton, et. al., Appellants, v. Matthew Glavin, et. al., decision of the Court and concurring and dissenting opinions, obtained from the Cornell Law School Web site, http://supct.law.cornell.edu/supct/html/98-404.ZX.html. "Adjusted Census to Slightly Alter Federal Funding," The Washington Post, March 3, 1999, p. A21. "Blue Ribbon Panel on the Census," American Statistical Association Web site, http://www.amstat.org. "National Academy of Sciences Convenes New Census Panel," National Association of Development Organizations (NADO) and NADO Research Foundation Web site, http://www.nado.org, Nov. 2, 1998. (This organization describes itself as advocating rural development and expresses a concern that, without sampling, some rural populations may be undercounted.) "5-4 Court Decision Bars Sampling for Purposes of Congressional Apportionment; Political Debate Shifts to Use of Sampling for Other Purposes," Census 2000 Initiative Web site, http://www.Census2000.org. (This organization describes itself as a non-partisan educational group, not affiliated with any of the parties, interested solely in informing the public about the issue. This site has announced that it may soon close down, as the case is now decided.) Douglas A. Samuelson is the president of Infologix and Division E director of INFORMS Table of Contents OR/MS Today Home Page OR/MS Today copyright © 1999 by the Institute for Operations Research and the Management Sciences. All rights reserved. Lionheart Publishing, Inc. 506 Roswell Street, Suite 220, Marietta, GA 30060, USA Phone: 770-431-0867 | Fax: 770-432-6969 E-mail: lpi@lionhrtpub.com URL: http://www.lionhrtpub.com Web Site © Copyright 1999 by Lionheart Publishing, Inc. All rights reserved.