Voter ID Stats | Stats and Stories at JSM / by Stats Stories

Kelly McConville is a survey statistician who develops estimation techniques that combine complex survey data with big data sources. Her work is used to estimate official statistics, related to canopy cover or occupational statistics, or to assess the impact of voter ID laws. She enjoys teaching her students how to learn from data and introducing them to R (an open source statistical software program). She also involves her students in her work and co-chairs two national programs: the Undergraduate Statistics Project Competition and the Electronic Undergraduate Statistics Research Conference

+ Full Transcript

Brian Tarran: Hi this is Brian Tarran from Significance Magazine and I’m reporting live from JSM 2019 in Denver, Colorado, and I’m here today with Kelly McConville of Reed College. Hello Kelly.

Kelly McConville: Hello.

Tarran: And we’re talking about photo I.D. laws. So earlier this week you gave a talk about the impact of photo I.D. laws. But before we get into that, do you want to tell listeners a bit about yourself?

McConville: Yeah. Great, thanks Brian. So I am an assistant professor of statistics at Reed College. I’ve been there for a year and so actually the study I’m going to talk about today was one I did when I was at Swarthmore College which is right outside Philadelphia.

Tarran: Okay. And so photo I.D. laws- so if we’ve got listeners outside the U.S. you might want to explain photo I.D. laws. Give us some context to that.

McConville: Yeah, great. So photo I.D. laws can take many forms but in their purest form, it’s just that someone goes to the polls to vote and they have to provide some document to verify their identification. So this could be a government-issued I.D., it could be a non-government photo I.D., it could be a bank statement, it really varies from state to state what the particular law is.

Tarran: Okay so, have there been more recent changes to these laws?

McConville: I would say, in terms of changes, it’s been more that many more states are taking up the laws. So when we did our study there were 33 states that had voter I.D. laws, and if we go back, even a few decades there were very few states that had them and they were much less strict, in terms of what identification you needed, so it’s typically stricter laws now. So at this point now, there are 35 states with voter I.D. laws.

Tarran: And so the idea of the laws is to prevent fraudulent voting or people who might want to vote more than once, or people who are not entitled to vote from actually doing that. But what was the question that you were looking to address through your research?

McConville: Yeah so that would be the primary argument for the laws. But a standard argument against the laws is that it makes it difficult for people who really have no good means to get an I.D. to be able to vote and that, unfortunately, this tends to impact specific groups more than others, minority groups in particular. So we really wanted to know, are people being affected by these voter I.D. laws where they aren’t able to vote and then also is there disparate impact?

Tarran: So how do you go about answering a question like that?

McConville: First I should say, this was not done by myself. I had two fabulous collaborators, so the person who first posed the problem was Mary Gray at American University and then myself, and Lynn Stokes from Southern Methodist University took up the charge to join her on this. So what we did is we did exit polling around our institutions. So Mary did exit polling around Northern Virginia, I did it in Philadelphia, and then Lynn did it in Dallas County. And I’m saying “we”, but really actually our students- our undergrads and graduate students- they are the ones who are going out to the polling locations and asking people about their voting experience.

Tarran: And would that be whether they were able to vote or not, or whether they were turned away or sent home to get I.D.?

McConville: Yes, so we did not ask them who they voted for, we were just asking them whether they voted or not. In Pennsylvania we were particularly interested in whether or not it was their first time voting in the district or not, because actually the Pennsylvania law that came into effect in 2012, which was a voter I.D. law, was actually struck down at the beginning of 2014, so in that case the only people who are supposed to be asked to show I.D. are first time voters in the districts. So, we, in particular, wanted to know was this being enforced correctly.

Tarran: So what were the findings?

McConville: So what we found was that sort of across the three sites, half a percent and 2% of people who went to the polls and tried to vote were turned away due to voter I.D. issues. And in particular, we found at some of the sites that there was a disparate impact. So, in particular in Dallas County we found that female voters were 12.6 times more likely than male voters to have issues with voter I.D. issues, and that black voters were four and a half times more likely to have voter I.D. issues than non-black voters. So again we’re seeing this sort of impact is not impacting everyone equally.

Tarran: Were your students able to – through the questionnaire that they had- were they able to pick up some of the reasons why they might me this disparate?

McConville: We didn’t ask any questions specific to that. One thing that I should note is the most common reason that someone was not able to vote at a particular place was just because they went to the wrong polling place. So we did have people who had voter I.D. issues, but that was not the most common reason that they were turned away.

Tarran: Okay so this was three sites that you were looking at- or was it multiple sites within?

McConville: Yes exactly. So all three sites used a two-stage sampling design where we took a random sample of polling locations, and then we did a systematic sample of people. So the students at my site were interviewing every third person.

Tarran: Okay, and so what can you say more generally than based on the sample that you’ve got the analysis approach that you took, is there anything- any wider conclusions that you can draw from this?

McConville: Yes, as a statistician I would be very cautious in drawing too large a generalization, and say that more work needs to be done. More work needs to be done in different locations. I would say this is a great time for me to plug that I would love to see more academics take this up for multiple reasons. One being that I think it was a wonderful learning experience for our students. So, I said I did it when I was at Swarthmore College, and Swarthmore is an undergraduate only institution, so I had 100 undergrads in their introductory and intermediate stats courses, learning about things they just typically wouldn’t have learned about, like survey sampling and non-response errors and ethics, and sure, these might be things that we would touch on, but we touched on them much more deeply than we usually would so I think it really enriched the course. And the other thing is I think there’s just a feasibility thing, right? So we trained the students, I’d say it took about a week of course time to train the students but then we were able to send them out and do this exit polling and if we were trying to hire 100 people to do that, the sticker-price on that would be much higher than in sending students out. So I would say it’s hard to generalize based on just our three sites, which, again, all three of our places had different voter I.D. laws, and so I would love to see more people do it. Another thing is we did it in a mid-term year, but it would be great to have people doing this in a presidential election year.

Tarran: Is this- the sort of questions- are they asked as part of the exit polls that media organizations might run? I mean I’m talking from a U.K. perspective. I assume that they do exit polls in the U.S. in the same similar sort of way?

McConville: Yeah that’s a really great question, I feel like what I see in exit polling is just estimates of who people voted for, right? So that we can have that information more quickly and so I don’t really know about people asking these questions and I guess it would be lovely if they were also asking about whether or not people had issues and the difficulty of getting to their polling place and if part of the process was confusing or anything like that.

Tarran: And that so that you’re able to do this future work and the results kind of extend to other areas, you find a similar sort of patent. What other kind of implications of the findings, for the voting process for democracy – you know, what would be the next step after that for addressing that through policy?

McConville: I think that’s a lovely and a hard question. So I guess that I would ask our policymakers to think about really trying to broaden their definition of the document required to show who you are, right? So the narrower that we make that qualification the harder it is, right? The more it excludes people. So if we could think more about what forms of identification do more people have, then I think that would be the way to ensure that everyone gets to vote.

Tarran: Okay, and there’s two more years until the president- what is it one year? Two years? How long is it?

McConville: Well, we already have plenty of campaigning going on, so it’ll be here before we know it.

Tarran: That’s right, so if people do want to help out they should contact you quickly.

McConville: Exactly, and I would love to say that if people are interested we’ve got lots of materials, we had scripts we wrote for our students we can help with sampling designs- all sorts of things we can help with. People thinking about logistics about how they would want to do this.

Tarran: Fantastic. Well, thank you for talking to us today Kelly. Good luck.

McConville: Yeah, can I say one more thing?

Tarran: Of course you can.

McConville: Okay, so I just want to end by saying- so we found that it was half of a percent to two percent that was maybe affected and I worry that people are going to go, “well alright, well then this is not really a big deal, right?” And so the one extra thing I want to add to that is if we just look at the 2014 Federal election there were two races that were decided by less than half a percent. There were four that were decided by a margin of less than one percent. So even if it’s a small effect- even if it’s a small number of people who were impacted, it still can have a pretty practical significance.

Tarran: Excellent. Thank you very much again for talking to us today and good luck with any future studies that may come out of this.

McConville: Thank you.

Brian Tarran: My name is Brian Tarran and I’m the editor of Significance Magazine. Find us online at For this special JSM series of podcasts we’re collaborating with Stats and Stories. Stats and Stories is a partnership between Miami University’s Departments of Statics, and Media, Journalism and Film and the American Statistical Association. Follow us on Twitter, Apple podcasts or other places where you can find podcasts. If you’d like to share your thoughts on our program send your email to, or check us out at And be sure to listen for future editions of Stats and Stories, where we explore the statistics behind the stories and the stories behind the statistics.