What Data Visualization Can and Can't Tell Us About Pennsylvania's Voter ID Law
BY Sarah Lai Stirland | Tuesday, August 7 2012
A spot analysis of some 868,648 records of active voters in the city of Philadelphia suggests that Pennsylvania's new voter identification law may have a disproportionate affect on minorities and students.
But the analysis, performed by Philadelphia-based software company Azavea and limited in scope only to the company's home city, raises eyebrows as much because of what it says about the lack of access researchers have to quality data on voters as it does about who might be affected by new law.
"It appears that Pennsylvania's new strict photo ID requirement may be in effect a racially discriminatory voting procedure," Tamara Manik-Perlman, Azavea's spacial data analyst, wrote in a recently published blog post.
But she admitted that the findings are limited in scope because the analysis is confined to the city of Philadelphia, the only geographic area for which Azavea has all the relevant datasets.
Still, the debate around voter ID laws is national in scope, and Azavea's latest work reiterates a glaring lack of reliable data informing those discussions.
When Pennsylvania Governor Tom Corbett signed a new voter ID law this March, the swing state became one of several in the past year to require stricter identification for voters hoping to cast their ballots. In Pennsylvania's case, voters generally need a state-issued driver's license with a photo. If they've voted in previous elections, they can apply for a non-driver's license photo ID. Students at accredited universities can use their student IDs if the IDs have an expiration date. Passports are another acceptable form of identification.
Democrats have generally decried the laws as discriminatory against students, the elderly and minorities. Republicans say that the new laws are necessary to crack down on voter fraud. Neither side has generally offered much quantitative evidence to support any of those claims.
Azavea performed the analysis by comparing voter registration data provided by Philadelphia's city commissioner, Stephanie Singer, with a list of individuals from Pennsylvania's Department of State that showed people without a valid ID under the state's new law. That list included voters who did not appear in Pennsylvania's Department of Transportation database, and voters with drivers' licenses that expired prior to Nov. 6, 2011. Azavea mapped that information to 2010 census redistricting data containing information about race and ethnicity.
The company performed the analysis after former Philadelphia Inquirer reporter Tom Boyer, volunteering in Singer's office, reached out to them.
"It's just a first step but I think it'd be pretty hard for anybody to knock it down," Boyer said in an e-mail. "There is a strong racial pattern to this. Everybody assumed there would be, it's not a surprise, but it's important to have it established statistically."
Azavea came up with a map that showed by wards the percentage of voters without valid ID. The map also shows what percentage of each ward is white, black, Latino and Asian. The map shows higher percentages of "active voters" who are blacks or Latinos than whites in many of the wards, and high percentages of those voters also either have expired IDs, or no valid PennDOT IDs.
"The map makes clear that the spatial distribution of those who lack ID is non-random," Manik-Perlman writes. "Voters without IDs are heavily concentrated around the University of Pennsylvania and Drexel University in West Philadelphia, as well as parts of North, West, and Southwest Philadelphia."
Manik-Perlman acknowledges that the PennDOT data isn't perfect — some people who do have valid voter IDs don't show up in PennDOT's database, as the Philadelphia Inquirer recently reported, because the computer system has trouble recognizing certain kinds of names. But she responds:
"There are two data sets-- the larger is actually the set of voters whose names matched in the PennDOT database but whose IDs are expired. These are positive matches-- the voters WERE found in the PennDOT database so the name misspellings are not an issue here-- and the pattern of disparate racial impact is very strong.
We didn't want to cherry-pick the data, though, so we included the second set in the analysis. The second data set, which is of the non-matches and has the problems that you describe-- has a much weaker pattern of racial correlation, but it is still there. If we were really talking about random name mismatches there would be no correlation.
Azavea's findings could add further fuel to a fight currently underway in Pennsylvania. The American Civil Liberties Union is leading a court challenge over the legality of the Voter ID law in state court, and is seeking an injunction against its implementation. Arguments in that case wrapped up last Thursday, and the judge overseeing it has said that he plans to issue a ruling next Monday. The Justice Department has launched an investigation into whether the law violates the federal Voting Rights Act, which prohibits voting practices that effectively discriminate against minorities.
Pennsylvania's Department of State spokesman Ron Ruman said that the department doesn't have any comment on Azavea's findings. Meanwhile, the Justice Department is conducting its own research.
However, Azavea's map and scatter graphs will only add to the perception that voter ID laws are discriminatory in their effect. What is barring a more thorough analysis into the real impact of these laws nationwide, Manik-Perlman says, is a lack of standardized, digitized information.
"For instance, the ward division (voting precinct) boundary files that we used are something that we have maintained manually in-house using text descriptions and non-geospatial maps," she writes in an email. "Current precinct boundary files are extremely hard to come by, and are often maintained on a county-by-county basis and may not be available in a digital format at all ... I think doing this analysis for all of the voting precincts in states where these voter ID laws have been enacted would be a monumentally difficult task in terms of gathering the necessary data."