blank.gif (51 bytes) The Right to Vote
A Sampler of Agendas

by Randall D. Lloyd, Ph.D.

resident Clinton asserted in his recent State of the Union address that "since every American counts, every American ought to be counted." This banality has as its premise that only sampling will count everyone. Technically, however, only a full person-by-person count will truly count everyone.

The reason offered by proponents of a census that would include statistical sampling is that in the past, and again in the future, millions of people in the United States will not be counted. They suggest that sampling will offer a better idea of the true population. Supporters of sampling assert that an undercount misses people in certain demographic segments of the population, particularly immigrants and poor people who may be hard to find or who are suspicious of government head-counters. Because the census is used for formulas for distribution of state and federal funds and for the decennial district reapportionment of legislative seats, undercounts have implications for policy that sampling proponents would like to see addressed.

Reasons given by opponents of sampling include both philosophical and practical concerns [see "Census Battle Has High Stakes for Nevada and the Nation," February Nevada Journal, pp. 20-21]. First, the constitutionality of a sampled census is in doubt. The Supreme Court has recently ruled that statistical sampling for use in reapportionment is not allowed under a 1976 law. The Court has not yet ruled on the constitutional acceptability of sampling; however, there can be little doubt that only a traditional census will pass constitutional muster. Second, opponents are concerned that statistical sampling will introduce politics into the process, with more powerful, populous states able to find people—real or imagined—and upset the relative balance. Sampling assumptions are apt to be based more on political considerations than on scientific ones.

Undercounting the Underclass?

allacies, and perhaps a hidden agenda, are at the heart of much of the current public debate over the proposed policy of using statistical sampling in the 2000 census. This debate has tended to be divided along party lines at the national level, with Republicans opposed to statistical sampling and Democrats favoring it. For electoral reasons, Democrats would like to see these uncounted people, who they perceive as natural Democratic supporters, included in the count. Republicans privately oppose sampling for the same reason. The divide is less defined at state and local levels for reasons related to federal and state funding. Distribution formulas for many programs are based on relative populations. Regardless of party, state and local politicians for reasons of funding are understandably more likely to favor a larger count, though with a statistical count many smaller states may lose funds to larger states.

Despite the belief that traditional procedures lead to inequities in the distribution of federal funds among the states, the problems are not as major as the rhetoric suggests. If federal funding were unlimited, then an undercount would shortchange the states. But funding is not open-ended. The pool of available dollars is finite. Money is fundamentally allocated as a state’s proportion of the total population. If any one state is undercounted and the rest are not, then the undercounted state would be shortchanged in its proportion. But, realistically, an undercount is not limited to a single state or group of states. More populous states like California, Texas, New York, and Florida have large poor and immigrant populations, many of whom will not be counted in a full census. If Texas, Florida, and New York miss the same percentage of their people in a count, whatever the actual number, they are on par with California and each other. Any state that miscounts by the same percent is also relatively equal. Where the problem lies is in the divergence from the average level of miscount. A state that misses 6 percent of its population while other states miss 5 percent is disadvantaged by 1 percent, regardless of the actual number, and not the 6 percent undercounted. For any single entity to profit, they must be the only one to find additional people. Underfunding will be evenly distributed for the most part, and only increases in overall program expenditures will meaningfully increase funding for any state or locality.

Donkeys and Elephants

he second concern expressed by proponents of statistical sampling is over missed electoral opportunities for Democrats. The assumption underlying this concern is that the uncounted people are natural Democratic supporters and will vote Democratic if counted. Therefore, goes the argument, the current process handicaps Democrats. But this reasoning makes little sense. The premise assumes that only people counted in the census can vote. That is not so. Suffrage is based on certain eligibility requirements, including citizenship, and not inclusion in a census count. If this were otherwise, people who became citizens between the decennial census would not be able to vote.

Perhaps what the proponents of statistical sampling really mean is that the number of congressional seats apportioned to the states, and the shape of congressional and legislative districts are adversely influenced by an undercount, mostly to the detriment of Democrats. However, under closer inspection, this concern is not supported. Yes, seats are allotted to states and localities based on population. But Supreme Court cases have established the requirement for exact population parity for congressional seats in a state and as practicably exact in population for state legislative seats. Each state is entitled to at least one congressional seat. Additional districts will be allocated for each population multiple of somewhere around two-thirds of a million people. However, since the number of congressional seats is unchanging while the number of people in each district increases with the population, the number of people required per seat will increase with the larger population found by sampling. If, as sampling proponents suggest, there are 15 million uncounted people, the population necessary to earn each seat will have to increase by 30-40,000 people over current estimations. California’s current 52-seat complement, for instance, will require finding an additional 1.5 to 2 million people just for the state to maintain its current apportionment. Additional seats will require population tallies beyond this and will be allocated on the new, higher number. Also, other states will find more people, thus offsetting any advantage that California might have had. For most if not all states, any newfound people will not automatically translate into more seats. State legislative district apportionment follows slightly different logic, and larger cities may be entitled to more representation, but Democrats are not likely at a current disadvantage.

Proponents of sampling may really mean that Democrats are harmed by the political makeup of the district with the lack of a larger enumeration. Imagine an inner-city congressional district with 650,000 evenly divided partisan identifiers who are counted, and another 100,000 Democratic-supporting uncounted people. Wouldn’t that be tragic for the Democrats? Wait a minute. A district that has an extra large number of Democratic supporters? Obviously the Democrats are not harmed in that district, since an overabundance of their supporters should help to ensure a Democratic victory. But suppose the people were counted and districts were gerrymandered to distribute Democratic supporters to other, more Republican districts. The undercount would disadvantage Democrats. Or would it?

Republicans are concerned about counting these people, because they believe, as well, that the uncounted will vote Democratic if identified and will influence electoral outcomes if distributed to districts. But why does anyone believe that these uncounted people will vote? We must assume that the only reason uncounted people—the poor and homeless, illegal immigrants, and drug-addicted derelicts—don’t vote is because the census does not count them. However, these are the people who don’t vote under any circumstances, whether counted or not. If found by a census, they still will not be likely to vote. The addition of these people will have an impact on electoral outcomes, however, and to the benefit of Republicans in both state and national elections.

If the numbers of people were increased by sampling, then they must be included in legislative districts. Once counted, a greater number of nonvoting, potentially Democratic supporters will be included. Imagine a district with 60 percent Democratic supporters, but where 50 percent of them don’t vote. This is a Republican victory in the making. Once included in districts, the effectiveness of the party with which they identify will be diminished. Rather than distributing the extra partisans to other, nearby districts, additional Democrats will need to be combined from those districts in order to assure a Democratic victory, thereby enhancing the prospects of Republicans in adjacent seats. The best outcome for any party is to exclude supporters who don’t vote.

Entitlement Search

ince the two arguments most often heard in favor of sampling are less than compelling, particularly in light of the constitutional and political concerns, there is likely to be another, unspoken reason for wanting to sample. One thing that a statistical count will do is to increase the number of people who are poor. For advocates of bigger government, showing a need for increases in the size and scope of government programs is a necessary first step in legislative battles to add funding for social programs. Since the 15 million uncounted people suggested to exist are likely to be poor, an argument can be made to justify an expansion of government. And if an additional 15 million poor people is not enough to propel the agenda, a census based on political assumptions can just as easily find 30 million missing people in need of government aid. Whether suggested for electoral reasons, funding formulas, or government expansion, let the debate at least acknowledge what a switch to statistical sampling will accomplish. An informed debate should consider more than the old, tired rhetoric. NJ

Randall D. Lloyd is a senior research fellow with NPRI and a lecturing professor in political science at the University of Nevada, Reno.


Join NPRI

Journal front | Search | Comment | Sponsors