Democracy Works & TurboVote, Co-founder Seth Flaxman talks about using data and technology to help people vote via mail. Their main product, TurboVote has had great success due to their team’s ability to gather the most accurate list of election locations in the country. They have made voting via absentee ballot as easy as getting a movie from Netflix, which is the first step of their larger plan to make voting easier in the US.
- Mechanical Turk – you can use this service for large, repetitive tasks that require a human eye.
- Odesk.com and screen scrapers – you can hire developers to create simple Python screen scrapers to aggregate data from the web.
- TurboVote.org – great tool for setting up a voter registration and absentee ballots on your site.
George: This is Using the Whole Whale. A Podcast that brings you stories of data and technology in the nonprofit world. This is George Weiner your host and the chief whaler of wholewhale.com. Thank you for joining us.
George: I’m gonna go out on a limb and say it’s probable that you have missed the chance to vote in more elections than you’ve actually voted in elections. And this is something that really frustrates today’s guest, Seth Flaxmen who’s a co-founder of democracy works. Welcome to episode thirteen where we’re gonna be talking with Seth about how they use big data and aggregated polling locations in order to make it easier for people to vote via absentee ballot. This project gained the attention of little known companies like Google who actually contacted them for that database. So we’re gonna talk with them and see where this project is going.
George: Seth thank you for having us here today on the podcast. You work for and you are the founder of Democracy Works. Tell us a bit more, who are you and what do you do?
Seth: Co-founder actually.
George: Yeah I already got one wrong. Nailed it.
Seth: I have to make sure to give Katie credit you know. Owning half the organization.
George: So tell me what do you guys do?
Seth: So the idea’s to use technology to make a seamless modern voting system for everyone. That’s the sort of 30,000 feet deal. And turbo vote is the main sort of program we have a site and the idea there is that anyone should be able to sign up online one time and then get all the help they need to stay registered to vote and voting in all of their elections for the rest of their life. So an entirely voter centered design voting system. But we just launched this year our second program called Ballot Scout and it’s a tool for local election offices. We did a nine month technology study to figure out what tech election offices actually wanted and would spend some money to acquire. And what they really want is a way to track all the mail ballots that they’re starting to send people more and more. Cause in 32 states now everyone can vote by mail. Every year more people vote by mail, but the tracking system isn’t that good so ballots get lost. People call them constantly ask them where their ballots are, have them returned. So that’s our second program.
George: So it’s pretty crazy to me that something as fundamental literally baked into our constitution the right to vote is so goddamn complicated. And one of the things I love about your initial approach is that you’re trying to make it, as you said, as easy it is to watch a movie on Netflix. That’s how easy voting should be. And you started actually by gathering a whole bunch of unstructured data. And can you tell us a little bit about, you know your process of saying. Tackling voting is a beast. Where did you start?
Seth: Yeah although I should say unfortunately in the U.S there is no right to vote in the constitution. This is something–
George: Oh good, but I can get bare arms, right?
Seth: Right yes exactly. I mean the U.S is the only country in the world that doesn’t have a federal election system. All of our elections are run by counties and towns. And it’s one of the reasons why every–
George: I have a right to badger my electoral college.
Seth: Right yeah.
Seth: Yeah I mean this is the problem with having the oldest continually functioning democracy in the world is yous till have all of these sort of patches. like our whole democracy is a patch upon a patch upon a patch, for like 250 years. But the question about the data stuff is interesting.
Seth: So there’s a challenge if you want to build a single system that anyone in the country can sign up online and get5 all the help they need to stay registered and vote in all of their elections. That means that you need to be collecting data from 8000 different local election offices. So into even the first iteration when we were doing the pilot in 2010 and we were just gonna be helping a few hundred college system use the system Even during the pilot we had to have the contact information of to the mailing address information for all 8000 of those local election offices. So that was the first data challenge that we faced. And we needed all of that address information because one of the ways we make registration, voter registration easy, or vote by mail easy is that when you sign up on the site, if you need to register or vote by mail there’s a form associated with that. So we’ll mail you that form already printed out, already filled out with a pre-addressed, pre-stamped envelope. So in order to pre-address that envelope, we had to collect all of this data. And there’s no central repository. There’s like no nationwide data setup. So we had to screen scrape every secretary state offices website.
George: Was this a manual or automatic process?
Seth: Automatic. Well we had to build screen scrape.
Seth: Build screen scraper for every site
George: So what exactly is screen scraping? Well it’s kind of like it sounds. You’re crawling through a site to find the relevant information you want and saving that to one central database. But instead of having a person do this ie: the worst intern task ever. You actually have a little bit of code that crawls the website and looks for the parameters that you set. Such as give me an email address that you’d find on here. Or give me the location and save that back to the main database. And this is not only time saving but can create a pretty valuable database of information. If you’re interested in doing this for your organization there’s a couple different ways. You can obviously Google screen scraper see what comes up. But you could also get a developer. You can potentially find them through odesk. You can find that developer and give them parameters. I want all of this kind of information from all of these kinds of sites. Have them build that algorithm for you. And then let it run and see what kind of data you come back with. Alright let’s get back to the interview.
Seth: After we had it, we then had to feed it through a postal service address checking api just to make sure that they were all real places and several of them were not. I would think like five of the states. Well like Virginia there was Text, what was it, Text Virginia was one of them. Was one of the towns.
George: Text Virginia? Is Text a real place?
Seth: No. (laughs)
George: It’s just a developer with a fake P.O box somewhere. So how did you figure that one out?
Seth: The address checker but then we had another town in Massachusetts and we sort of called them, like you know your address the postal service doesn’t know where you are. And they’re like of course they know! It was Barnstable, Barnstable Massachusetts. Of course the post office knows where we are, we’re right across the street! And so there’s a complete, like we don’t need an accurate address in their minds because they’re across the street from the post office in this small town in Massachusetts. So why would a correct address matter.
George: Frank knows exactly where we are!
Seth: Right exactly. So that was the first sort of level of challenge. And now what we did is we actually turned all the screen scrapers into a tool that automatically screen scrapes every month, called Dog Catcher. And that’s how we see whether these addresses are changing and stuff like that. Yeah there’s a bunch of those challenges. Like election dates. You have to collect every single election date in the country. Something a lot of people don’t know is there is an election every single Tuesday. Somewhere in the country. Somewhere in the country there is an election every single Tuesday. So we’ve tried to bunch different ways to do that. Right now it’s strangely enough the best way to do it is just setting up hundreds of Google alerts for election, like election data county this. Election date county this. And we have a full time researcher, and this is part of his job. We spent a whole summer trying to experiment with Mechanical Turk to see if we could collect all of the election dates in a sort of more automated way. But the quality wasn’t good and it was too hard. So we’re still looking for ways to do it more automated.
George: Mechanical Turk is actually a bit of software that allows people from remote locations to actually complete simple, what they call human intelligence tasks, or HIT’s. And this is all run through Amazon, so you can look up Mechanical Turk and Amazon and see the types of tasks, such as, can you look through these images and find all of the things in this image. Write them down. Give them back to me. And so it’s kind of like these things that need a little bit more human intelligence to accomplish. All itemized so that people anywhere, remotely can work on them. Mechanical Turk actually has a pretty hilarious back story as well. It came from someone in 1770, a guy named Wolfgang VanKemplan, who created this Mechanical Turk or automatic chess player to empress Maria Theresa of Austria. And this automatic, chess playing illusion traveled the country and played people like Napoleon and even Benjamin Franklin. Turned out that it was a hoax and there was a chess master hiding inside it to play. So that’s a little bit of backstory in there. Alright let’s jump back into the interview.
Seth: For now it’s just blood sweat and tears.
George: No substitute. Accept no alternative’s. So you have all of this data. Tell me a little bit how your impact measurement, like so what. It’s nice but so what. I think there have been, I’ve heard of actually other projects that have been able to collect this data. How have you turned this into impact, and what are you measuring?
Seth: We collect all of this data. Election dates, forms, deadlines, I.D rules, addresses. In order to build Turbo Vote. And in Turbo vote the goal is, are we actually helping people register to vote and are we actually helping them vote. Are they actually voting. And right now, there’s a huge data challenge to answer that question, because although whether someone votes is publicly accessible information. So you can actually see whether your user did register and did vote, the system is so bad. The only folks who have the, there’s like sort of two or three firms that collect this data nationwide. Catalyst, Target Smart. They only update this information once every two years about. Right after a federal election. So we only have these brief windows to check in, once every federal election to see whether our folks voted. And it’s good, like we had very high numbers in 2012 that we were really happy with. So of the people who we registered to vote who were first time voters, 76% voted in the 2012 election.
George: What was the total number on that?
Seth: The total number of voters?
George: Yeah, for you guys.
Seth: Oh so we actually I don’t remember what number that is off the top of my head. I mean we signed up almost 200,000 folks to use Turbo Vote that year. But yeah I mean honestly what I would love to see. The metric I would love to have for turbo vote is the percentage of elections someone is voting in over the course of the year, over the number of elections they could vote in. Because there’s so much obsession in our space with the federal elections, when really the way democracy most has the most direct impact in someone’s life is through their local elections. There’s lots of people who care about quality education and then don’t vote in school board elections. And so we want to increase that sort of ratio and make people realize that a lot of democracy occurs between those four years when presidents are elected. But it’s so hard to get that data. I don’t have a plan for that yet. (laughs)
George: Well at least you know what you don’t know.
Seth: Yeah right.
George: Puts you at least somewhere ahead of where you were before.
George: And that’s a great way of looking at it sort of like an election penetration number.
Seth: Right and also cause if you can get someone to vote in even like two or three elections in a row, it changes their habits and become regular voters for life.
George: Like brushing your teeth.
Seth: Yeah exactly if you just sort of almost remember every four years you’re never gonna become a regular voter.
George: So talk to me about what’s coming down the pipe. What are you trying to work on to extend your mission?
Seth: I sort of alluded to it a bit before. We have these two big programs. One is Turbo Vote. In a sense that’s everything you can do to make voting easier before you work with government. So you just, take all the information government has, the raw material and you turn it into an application that sits on top of the platform of government. And now what we’re trying to do is to actually sort of become that platform. So to build our first tools that government election officials could use to make voting easier. And one of the reasons why we’re doing this and why someone else hasn’t done this already is what I was saying before. All of our elections are run at the county level. And in eight states the town level. And so no VC’s are investing in companies that build great affordable technology for county government. There’s no real market there. And so we’re stepping in as a nonprofit tech startup to build that technology and sell it to them at sort of an affordable rate. And just because we want elections to be run better.
George: So this is hilarious. In order to be this type of solution, you realize after looking at the marketplace and gathering this information, you found a problem and you’re like if we can fix the infrastructure of this. Then we can get back to our ultimate goal which is to get more goddamn people to vote. We got to fix literally the goddamn plumbing of everything that’s going on.
Seth: Right exactly it’s like we have to get all the gunk out of the pipes in our democracy and then see where we actually are. Right now we don’t have a clear picture of what our participation rate actually are because the whole system is designed to decrease turnout. As opposed to make turnout easy.
George: I feel like that’s a topic for a whole other podcast in terms of ease of voting and disenfranchising.
Seth: Oh my god I would love to geek out on the history of voting. (laughs)
George: What is your favorite tidbit that you think people don’t know about the history of voting?
Seth: So why do we vote on a Tuesday? That is my favorite question. There’s a whole nonprofit that’s called, Why Tuesday? It’s such an interesting question.
George: No clue.
Seth: We vote on a Tuesday because in the 1700s, Sunday was church day so you couldn’t vote on Sunday.
George: That’s gods day.
Seth: That’s gods day, you’re at church. Then Monday you would travel to the county seat. If you were in George Washington’s district you would stay over at his house. You’d get a lot of beer, you’d party at his house. Then Tuesday morning you would go to the town square. Two candidates would line up and one at a time the white wealthy landowners would shake the hand of the candidate they wanted to support in front of everyone. And you’d have time to then travel back home that afternoon in time for market day on Wednesday. So from the beginning of the U.S voting was supposed to fit the way we live. And be really seamless and convenient. And instead we’re just still voting like we live in the 1700s in a farming society.
George: Although I feel like it’d be pretty damn cool to shake Obama’s hand. Be like, yeah I got your back.
Seth: Yeah exactly (laughs)
George: Not terribly effective.
Seth: One of the few things people know is that the highest participation in voting happened in sort of the Lincoln era, the mid 1800s we had like 70-80% participation in elections. And what you would do is you would grab, the parties would hand out ballots like already filled out, and you would choose the ballot you wanted. And then you would drop that filled out ballot in the ballot box. And it was important to have that because so many people couldn’t read. So you had to be able to grab a filled out ballot. After the end of the Civil War and the sort of freeing in the first Black Americans voting, that’s when we invented the ballot booth in the ballot. Because it kept illiterate people from voting. Specifically black voters from voting. So even the invention is called the Australian ballot. It was taken from there and then adopted nationwide almost over night in like 1870s. So it’s just I love these stories like this because it shows that our voting system is not set in stone. We change it for all sorts of political reasons over time. And we could continue to change it.
George: We gerrymandered the heck of it.
George: Tell me a little bit about how it seems like you do think with a plumbing mindset, right? You have to look up the river for types of solutions, how do we look at the system as a whole. Tell me a little bit about your working with API’s.
Seth: One of the things we’re excited to be doing next year is we realize that our whole, the whole Turbo Vote application runs on all of this data. And we’re going to open up each of those sort of data feeds as it’s own individual API so all of their people can build interesting tools on top of it. So we have it so we’re gonna be building API of every election date. So that if you want to build your own type of reminder tool of that election date.
George: Are you going to be charging for use?
Seth: No not for the raw election data.
George: It’s mind blowing that nobody does this.
Seth: It’s mind blowing.
George: The government should be freaking–
Seth: Well actually I forgot to finish my story here. Even when I, from before when we did the pilot in 2010 we built this data set of local election offices.
Seth: We were just sharing it online some labs list being like, here is this list. It’s useful we put it together for our pilot. No one else has this. And someone from Google reached out to us, was on that list and said. That’s amazing, that’s the cleanest most complete list of that type I’ve seen. We’re gonna use and we’d like to give you a grant. So–
George: How do you do?
Seth: Yeah, ooh! And actually I thought that’s how fundraising worked after that that’s like not true at all.
George: Just get out there do build all this stuff and then Google just drops you a check!
Seth: Yeah not true.
George: Yeah that’s not how the world works
Seth: But it was just sort of amazing at that point because it’s just three of us working at that point while we were in grad school part time and we put together a data set that Google wants to use because no one else had done it.
George: That’s fantastic. So talk to me, you’re opening up these API’s and instead of doing what I think a for profit would have done, which is, it’s mine it’s ours and if you pay us an exorbitant fee we will this information to you. I love it. That’s why a non for profit sector is I think are very stoic where technology needs to go because it’s exactly this type of story.
Seth: Yeah this is like public data it belongs to everyone and it should reinforce the whole civic tech space. And a thousand different ideas.
George: Yeah data.gov. If you’re listening. Come on!! And final question here. You’re gathering these data, you’re creating systems. What is it going to take you think to move the needle on turnout? What is the tipping point and what is the role of technology in hitting this tipping point?
Seth: So we work a lot with college students. Right now we have 125 college partners. And what that means is they actually pay us annual partnerships fees as part of our model and we build them a version of Turbo Vote with their logo and everything on it. and then they put their site by class registrations. So everyone who signs up for classes signs up to turbovote. Or could be freshman orientation or something else. So like a bottleneck point that every student has to pass. And college students are an amazing population for this question because in 2010 of the college students who didn’t vote, 66% said they didn’t vote because of process issues. They forgot, you were working, you were out of town. and these are problems that we can solve. This are the low hanging fruit of turn out. You know build a system that makes vote by mail easy, that sends reminders so that you can take out all of the process excuses for people.
Seth: So that’s one part of the equation where I think we can increase turnout. Because voting, when you talk to a political scientist, there’s a benefit cost analysis for every voter. Does the benefit of voting outweigh the cost of voting. And what politicians do is they usually try to up the benefit by promising you a bunch of shit. But what we’re doing is we’re lowering the cost of voting on the other side of the equation. And I think that’s the first step to massively increasing turnout. But there’s a second part of it where voting is a social behavior. You vote because your friends vote, your family votes, you know other people are voting, you don’t wanna be seen like someone who’s not voting. And I think my hope is that we can increase turnout enough by making the process easier that we can then sort of create more social pressure for people to vote. Especially by making vote by mail easier. I think it actually a lot of young people especially they can vote next to their laptop. So they like, have their ballot out and they’re on Facebook. And they’re like, hey guys what the hell is this person’s name doing this. Like what’s this ballot initiative. It actually makes voting more social and more informed.
George: I’m forced to agree with you and playing with the variables of ability, motivation, and trigger. If you moving down the sort of cost and opportunity cost of voting, that’s one solution. Aside from keeping the U.S Postal service in business, obviously by mailing. How far out is the online ballot?
Seth: The online voting is tricky because it has to be both secure and anonymous. So you can bank online because it doesn’t have to be anonymous. The bank knows every single click you’re doing and so it can be secure. And they can also if you make a mistake and 100,000 dollars moves out they can put in a different 100,000 dollars. Money’s fundable in a way votes are not. And so it’s very challenging. I don’t think that technology exists now to build a system an online voting system that is both secure and anonymous. So one massive issue, I actually think that the bigger problem beyond the technology. The technology’s just one big problem and I don’t see any solutions out there right now. I think trust is the bigger problem, especially after 6 months after all this NSA revelations and heartbleed. This has been a bad year for trust and the internet and I think that is actually gonna set it back even longer. So for now for me I think the best we can do in the short term and in the medium term is seamless vote by mail. Your ballot comes automatically in the mail . You get a text message, reminding you to mail it back. We’re gonna know if you don’t vote.
George: I will find you like Liam Neeson.
Seth: (laughs) Kidnapped. And that’s the first step. And then the second step is gonna be building trust and figuring out what the technology is and building trust in it and I don’t know how long that’s gonna take.
George: That’s a hard thing to measure. Seth, thank you so much for joining us. How do people find you? How do people help you?
Seth: If you, there’s lots of ways to help us.
Seth: If you want to give us a grant, you can give us an email email@example.com and we will find you. But if you want to help us partner with a college that isn’t already partnered with us. We’re trying to get to 300 colleges this year from 125. And you can email firstname.lastname@example.org.
George: Do you guys have Stanford yet?
Seth: We do, we just signed them last week.
George: We got a couple Stanford listeners, nice job guys.
Seth: Yeah good job! (laughs) And how else can you help, I don’t know, just follow us on twitter and we’ll figure it out.
George: Perfect and you’re at?
George: Seth thanks again.
Seth: Alright thank you!
George: Democracy Works and Turbovote are not just doing another get out the vote initiative which we’ve been doing for years and I think seen the results and the upper limits of what happens. They looked at the issue from afar and said, what can we do to make it easier? How do we reduce the activation energy that it takes to vote to go out there and go on location especially for a college bound displaced population. And they landed on absentee ballots and they said, you know, we’re getting movies via mail through Netflix. That’s easy enough. Why can’t we make it as easy to get a Netflix video to our doorstep and back, as it is to send an absentee ballot. And as they chased that question, you see where they kind of ended up. They actually had to create the database. Now create the technology in order to reduce the activation energy, And I do believe that there’s an inevitable step toward electronic voting. But as Seth said it’s gonna definitely hinge a lot on our relationship of trust as it comes with technology and data security. And it’s been a rough year, that’s for sure. But I’m still hopeful that true democracy should be working best when it’s people are able to vote and also be able to read the ballot of course.
George: Well we’ll leave it there. Resources on today’s podcast can be found at wholewhale.com/podcast. Thanks for joining us. This has been using the wholewhale. For more resources on today’s show please visit wholewhale.com/podcast and consider following us on twitter @wholewhale. And thanks for joining us.
George: I see you’re still listening so I’ll keep talking. Special thanks today if the audio sounds a bit better it’s because we have a new member. She’s Sarah Finn who’s helping us with audio for the podcast. And in the true tradition and spirit of using the wholewhale I enlisted our front end whaler Simon Gringregg on the banjo. And so we had a banjo recording session. And so thank you to both of you.