Georgia's governor's candidates Stacey Abrams, left and Brian Kemp on May 20 in Atlanta. (John Amis / AP) Recently, there…
Recently, there has been a rebellion about Georgia’s attitude towards voter registration. The state’s “exact match” team passed last year requires that citizens’ names on their issued ID numbers must match their names just as indicated on the voting rolls. If the two do not match, that person’s voice is not counted. Georgia NAACP and other civil rights groups have filed a trial claiming that the measure effective since July 2017 refers to in the case of discrimination on racial minorities during the forthcoming mid-term election.
Georgia’s Foreign Minister Jack Kemp, a Republican Governor of Stacey Abrams Governor, has so far taken over 53,000 voters in view of unfair matches in the names of their voting records and other sources of identification, such as driving licenses and social security cards. If the action enters into force, voters whose information does not exactly match all sources must include a valid photo ID for voting on the polling day to vote. It may suppress voting, either because some voters lack ID or because voters are confused if they are eligible. Proponents of the rule claim that it is only intended to prevent illegal voting .
However, missing a hyphen, an initial instead of a full middle name, or just a contradiction in a letter in a voter, give ample proof that the voter is not who they say they are? How would we know?
Researchers often need matching records – and they need to get it right.
As happens, researchers often ask that question. When doing empirical scientific research, they often need to connect different data sets by any incomplete identifier – say agency names or individual addresses. While this may be boring, it is important to get the matches right. Match the wrong documents, and the analysis can be completely unreliable. This means that many data analyzes only maintain accurate matches.
However, even if incorrect matches can cause problems, items may be matched but have small differences. Eliminating these items can also corrupt an analysis.
Therefore, for the last three years, I’ve helped to develop an algorithm using a “fixedlink” probabilistic disk link that not only makes the data set quick and automated, but also tells the analyst how likely it is is that an inexact match of two entries is actually correct.
In a New Study co-authors with colleagues Ben Fifield and Kosuke Imai, we apply the algorithm to the issue of voter identification. The results give rise to serious concerns about Georgia’s exact match team – and its probability to prevent tens of thousands of valid voters from voting on polls.
How we did our research
We worked to link two nationwide selected files from 2014 and 2015 collected by L2 Inc a national non-partisan company providing voter data and related technology for campaigns . All active voters in 2014 appeared in dataset 2015, meaning we knew that a true match was always there. However, many records had typographic differences that prevent exact matches.
Our analysis showed that the “exact match” method would only link 66 percent of voters, who were actually identical, and correctly identified 91 million voters. In other words, “exact match” would exclude nearly 40 million entries actually referring to the same voters – disenfranchising quite a few Americans.
What does this mean to Georgian voters?
Georgia’s registry had a larger share of exact matches than we found nationwide – but 30 percent of the actual voters still failed to match exactly that situation.
However, we used our algorithm, which correlates with L2’s own matching record almost perfect (r = .99). We can match almost 127 million registered voters – or 93 percent of all voters in the 2014 data. Among those who did not match exactly, we matched that 25 percent have at least 99 percent probability of being correct matches, while 28 percent have at least 95 percent probability.
With our algorithm, in other words, 91 percent of those voting in Georgia’s voters would be cast for voting, or 3,441,342 voting citizens – while “exact match” cleans only 70 percent, disenfranchising 909,540 eligible citizens.
I also tried to link the voters in the 2016 US National Selection Study (ANES) with voter data in L2 data using two methods: exact match and an improved version of fastlink In recent developed .
The results are shown in the table below. As you can see, the exact match method misses a significant portion of valid matches. While our algorithm validated 60 percent of the voter account, “exact match” on average chose less than 30 percent.
And in accordance with the concern of the opponents of the Georgia action, non-white voters are especially likely to be dismissed. Match-match match rates are nine and six percentage points lower for each black and Hispanic voter than for white voters.
Georgia’s “exact match” team is the latest in a variety of voter identification actions that critics claim are thin disguised voters suppression tactics. Regardless of whether or not, Georgia’s “exact match” rule will disproportionately prevent minority voters from rolling their votes.