Presidential Members Report: Implications for Minority Voters in 2001 January, 2001 << Back to Introduction

Introduction: The Undercounting of Minorities in the United States Census

It is not possible to count fully and accurately the population of a large and diverse nation. No matter how heroic their efforts, census takers inevitably miss some persons who should be counted. Other persons may be double-counted and the enumeration process itself may erroneously add persons to the count. The net undercount is the difference between persons missed in the census and persons double-counted or erroneously included.3 Since 1940, the Census Bureau has acknowledged a net undercount that is not randomly distributed across the population. The undercount, measured in percentage terms, is substantially larger for minorities than for whites. The differential undercount for minorities, moreover, is not a function of the total undercount in the population. In 1940, the net undercount for the population was 5.4 percent, compared to less than 2 percent for 1980 and 1990. Yet the differential undercount of about 4 percentage points between whites and non-whites has not varied significantly during this 50-year period.4

Traditional census procedures are not effective in reducing the differential undercounting of whites and minorities. According to recent studies conducted by two panels of the National Academy of Sciences, "differential undercounting cannot be reduced to acceptable levels at acceptable costs without the use of integrative coverage measurement," which is a statistical sampling method that adjusts census data for the undercount. Three panels of the National Academy of Sciences recommended including such an adjustment in order to create the most accurate possible census and to reduce the differential undercounting of whites and minorities.5

Although census data corrected for such statistical sampling were not included in the official data released for redistricting purposes in 1991, the Census Bureau did eventually report the corrected numbers, which substantially reduced the undercounting of minorities. Table 1 demonstrates that for each state studied, the percentage undercount of members of minority groups is substantially higher than the percentage undercount of non-Hispanic whites. In one state _ New York _ the 1990 Census even slightly overcounted non-Hispanic whites, despite undercounting members of minority groups. Table 2 demonstrates that the undercounted population in each instance includes a much greater percentage of minorities than a state's overall population.

The failure to adjust for the undercounting of minorities in the drawing of legislative districts can affect opportunities for minority voters to participate fully in the electoral process and elect candi dates of their choice to legislative positions. Differences in the numbers and percentages of minorities can be significant in areas of minority concentration, where legislative districts with substantial percentages of minorities are drawn. Given differences in the candidate preferences of white and minority voters, small differences in the minority composition of districts can be consequential for the opportunity for minorities to elect candidates of their choice in such districts.


The state-by-state analysis that follows below will explore the implications for minority voter opportunities of using corrected as opposed to uncorrected data from the 1990 Census. At the time of the completion of this report, the Census Bureau had not yet released the block level data used for the post-2000 redistricting and other purposes. Existing legislative districts, however, serve as a baseline against which to assess minority voter opportunities when new districts are created in each state. For those involved in drawing redistricting plans, the 1990 experience can also be an important guide for assessing the impact on minority voter opportunities of the use of corrected versus uncorrected data for the post-2000 redistricting.

This analysis will not attempt to redraw legislative district plans or demonstrate that a shift to corrected data in the post-1990 redistricting process would necessarily have produced different configurations of districts. It does not suggest that districts should be based primarily on race. It does not attempt to reassess comprehensively the redistricting plans of the states included in the study. No claim is made here that the districts identified in the state-level analyses are the only ones in which there had been a potential to enhance minority-voting strength. Rather the analysis explores only whether the use of corrected data would have had the potential to expand minority voter opportunities in the post-1990 redistricting. The analysis is limited to plans drawn statewide for Congress, State Senates and State Houses of Representatives. It should be noted, however, that as compared to such statewide redistricting plans, the use of corrected census data would likely have had an even greater potential impact on the relatively less populous districts drawn for municipal and county governments within each state. The web site of the National Committee for an Effective Congress (NCEC) provides data on the demographic composition of each state's congressional, state senate, and state house districts.6 For states that have altered their district plans during the 1990's, the NCEC data is for the most recently available plan. District maps are available from the NCEC and from the web sites of most state governments.




Summary of Findings






North Carolina



New York


End Notes

Report Home

To top.

U.S. Census Monitoring Board
Presidential Members
4700 Silver Hill Road
Suite 1250 – 3
Suitland, MD 20746
Phone: (301) 457-9900
Fax: (301) 457-9901