Methodology

News21 analyzed voter registration and removal lists in a dozen states and a handful of key counties. The lists included 49.3 million names of registered voters and 7.2 million names of voters who were struck from the rolls. They were compared with U.S. Census data for race, poverty level and millennial-age proportions by state, county and ZIP code. The list of purged names was also compared with U.S. Census data for the racial composition of 151,000 surnames.

News21 examined the aggregate purge and registration totals from the 2014 election for more than 3,100 counties nationally compiled by the U.S. Election Assistance Commission and reported in the Election Administration and Voting Survey files.

Due to gaps in some of that data, News21 obtained current full or partial lists as of mid-July for Florida, Idaho, Kansas, Nevada, New York, North Carolina, Oklahoma, Texas, Vermont, Washington and the District of Columbia. News21 also gathered voter lists from Cuyahoga, Franklin and Hamilton counties in Ohio; Vermillion County, Indiana; and Harris and Travis counties in Texas.

Many states refused to provide lists. Others charged high fees for the records.

Limited, but reliable. In the 2014 election, 190 million Americans registered to vote. Nearly 15 million registrations were purged. Broadly speaking, News21 data reflects half of the people purged and a quarter of those registered in states in all regions of the country.

First we looked at the statewide EAC data from 2014 and calculated the purge rates for all the counties for which we had data. In parts of the country where elections are handled at a more local level, we used geographic identifiers to group the communities in each county. We then incorporated data from the 2014 American Community Survey for each county. Then we ran statistical analyses to determine the relationship between purge rates and poverty, race and age. There were no correlations.

Next, we analyzed the states for which we had full voter and removal lists. We generated purge rates by county, ZIP code, surname and party affiliation. We joined the demographic data by ZIP code and county with the 2014 ACS data. From this we ran correlations and linear regressions to find relationships. There were none.

For the surname analysis, we eliminated names that showed up fewer than 100 times. Then we merged the purge rates for those names with U.S. Census data for surnames, nationally. The list includes only surnames held by 100 people or more, and includes the racial composition by name. For instance, 73 percent of the people named Smith are white, and 22 percent are black, with other races making up the remainder. We ran a linear regression between purge rate and racial composition by surname. No patterns emerged.

After sharing the News21 findings with political scientists and demographers, some suggested weighting the ZIP code analyses by population because the purge rate of a ZIP code with 100 people may not be equivalent to a purge rate in one with 1,000. We ran weighted analyses for New York and North Carolina and found no significant change in our findings.

Some came from anecdotal accounts, some from claims made and quantified in civil rights lawsuits, and some in making spot comparisons between two places.

In Hancock County, Georgia, for instance, the NAACP documented in a civil rights lawsuit that 174 people were slated for removal and nearly all people who were purged were black. This was corroborated by reporting in the town of Sparta.

In Cincinnati, there were clear differences between some ZIP codes. There, ZIP code 45216, Clifton Heights, has a 27 percent purge rate. The ZIP code 45248, or Cheviot, has a 9 percent purge rate. In Clifton Heights, black residents make up 30 percent of the population, and 26 percent of the population is below the poverty level. In Cheviot, the population is 97 percent white and only 6 percent live below the poverty line. The region’s poverty rate is 14 percent.

In New York, ZIP code 14209 in Buffalo had a 49 percent purge rate, with 24 percent of the population living below the poverty level. By contrast, in ZIP code 11941 in Eastport on Long Island, the purge rate was 15 percent and the poverty rate 12 percent.