Statement on Appropriate Data for RedistrictingApril 29, 2021
More than 50 Groups Release Statement to Outline Appropriate Uses of Decennial Census and American Community Survey Data to Draw Final Maps
Type:Letter / Comment
In light of the upcoming delayed release of redistricting data following the 2020 Census, the undersigned groups and individuals unequivocally affirm the basic principle that it is not appropriate to implement electoral district lines based primarily on American Community Survey (ACS) data. Data from the decennial census and data from the ACS are different, and each has appropriate applications, including in the area of redistricting.
Among other compatibility issues, the decennial census is designed to be a count of all of the nation’s residents as of one day every 10 years, while the ACS is a sample survey conducted annually on a rolling basis. Though there are specific and appropriate purposes for each data set in the context of redistricting, there are a number of reasons why ACS data are not fit to serve as the main source, or base, of total population for the purpose of drawing the final lines.
Congress enacted P.L. 94-171 in 1975 with the specific purpose of providing states with decennial census data developed for redistricting purposes. These data files include “block” level data by voting age, race, and ethnicity developed with the cooperation of state governments so that redistricting plans may comply with population equality and voting rights concerns.
ACS data are provided for other purposes, including to enhance and inform electoral line-drawing, but not as the primary population base for redistricting.
First, ACS data are estimates based on a small sample. They are designed to provide estimates, not specifics or counts, and can differ significantly from census data, which are designed to provide population counts. As an example, a comparison of the total population estimates from the 2005-2009 ACS (the most current at the time of redistricting in 2011) to the total population count from the 2010 decennial census show significant differences, such as county total populations varying from +36.6% to -51.5% or more between the two datasets.
Second, ACS data are dated and not current enough for purposes of redistricting. Census data are released one year after collection, while ACS data are released about two years later. ACS data, estimated from small samples, represent an average over a five-year period. For the 2021-22 redistricting cycle, states will have census data from 2020, but the most current ACS dataset available contains data collected between 2015 to 2019. This means the ACS data are on average 4 years old.
Finally, ACS data are not available at the census “block” level – the smallest geographic unit for which the Census Bureau reports data. Block level data allow for the building of legislative and other electoral districts with the requisite accuracy to meet constitutional requirements of equal population. ACS data are provided at the “block group” level. It is therefore nearly impossible to draw districts with equal population using ACS data as the base. Various systems to convert block group level to block level vary in methodology and in accuracy, and there is no consensus about best method.
As long-time census and redistricting stakeholders, we recognize that the delayed delivery of census data has resulted in a scramble by states seeking to address this adjusted delivery, compressing the redistricting process, including creating situations where states may be unable to meet state-established deadlines. The use of existing data, such as the ACS, as the primary basis to draw final maps is inappropriate. Line drawers should wait until decennial Census data is available to draw actual maps to be implemented in elections. We reject any attempts to rush the release of PL 94-171 data.