Mapping the Electorate - Liveblog from #PoliticalAnalytics2016 | MIT Center for Civic Media

Mapping the Electorate - Liveblog from #PoliticalAnalytics2016

From the #PoliticalAnalytics2016 website: Political Analytics is a one-day conference at Harvard University featuring top minds from media, politics, and academics. We are starting an exciting conversation about the growing role of data and analytics in determining the winners and losers in politics. Our goal is to promote new methods, technology, and discussions for the improved analysis of politics.

Speaker bios from the #PoliticalAnalytics2016 website:


Kirk Goldsberry is a sport and political geographer and writer specializing in spatial reasoning, visual communication and representation. He was a visiting scholar at Harvard University and a staff writer at Grantland.


Amanda Cox is an editor for The Upshot and is an expert in visual journalism involving graphics, interactives, and other multimedia on web platforms. Her work has vastly improved the reader experience for The New York Times to follow election returns and election coverage.


Karl Gude is a professor for the School of Journalism at Michigan State University. He has an extensive background in information graphics, working in the field for news outlets such as Newsweek, The Associated Press, and The National Sports Daily. He is currently spearheading the first information-graphics program at MSU.


Andrei Scheinkman is a director of data and technology and a deputy editor for FiveThirtyEight. He has worked for The Huffington Post and The New York Times and is an expert in developing data-driven interactive features for online publications.


Kirk introduces the panelists and asks Karl to lead things off. Karl shows a scene from an old United Press International newsroom in the 1970s. He started as a cartoonist and shows a picture of his desk while he was covering the Reagan campaign.


Image of Karl's desk in 1979.

He shows the base maps that they used at the time. They would cut everything out and hand pattern it and then photograph it. When you went to print you would wrap it around a drum and convert the graphic to pulses which would interpret them to information that could be sent around the wire. Then they made a small improvement to the fax machine instead of the drum. Finally, they moved to an early Apple computer. But even when the photos were high-resolution they had to be transmitted via fax or drum.

He moved to work at the Associated Press and tried to get people to use Apple computers but the older people refused. The younger people did adapt to it. In 1988, he covered the elections with an Apple computer in his backpack.


Graphic by Karl from 1988.


Now they produced better graphics that could be transmitted via modems. They had better basemaps. He moved on to graphics at Newsweek. They had a nice staff and that's when mapping really grew up. He wanted to build things for the web but they were told that that would have to be for their personal time. He offers that that might be a reason that Newsweek went out of business.


He's now teaching graphics at Michigan State University and shows some of the complex, modern maps that his students have been making.


Map by a student of Karl's.


Kirk notes that the medium is always a character in this process of mapping the election. Now that we have computation at our fingertips, things have really changed.


Amanda shows one of the favorite maps of people at the New York Times. The reason they like this map is because it's high resolution (maps precincts) but also lets you filter by income. Another reason they like this map is that users picked it up to do other things like see who voted for which mayors in unplowed areas during a snowstorm.



The Mayoral Primaries, an interactive by the New York Times in 2013.


She shows how maps reveal patterns. What look like electoral patterns now - such as a crescent shape of voting patterns across the South - are also maps of cotton production 100 years ago and are also maps of where the coast was prior to that so the soil was rich for production of cotton.


Kirk transitions to Andrei from 538.


Andrei oversees graphics at 538. They use that term broadly. They have a staff of 8 people who have experience in design, data, mapping and more. He shows an image from 538:

The Facebook Primary, a project by fivethirtyeight.


They worked with Facebook to get data from the primaries. They started looking for patterns for what people were sharing on Facebook in regards to the elections. For example, if you turn on Cruz you see that people in Texas support him. But they found the most interesting results at the local level. If you click into Austin, TX, you see very granular data at the zip code level.


Here the point is about the map as an interface to let you drill down into the data.


Kirk states - That's a real shift right? When Karl was working you knew the story. But in this day and age you don't necessarily know all the stories in your own data. You can give people the Choose Your Own Adventure but there is still a story to tell. He asks Amanda to describe the Times' wind map.


Amanda describes her role on the map that shows county-level change over time as wind. She says it works as an impressionistic map of change in political leaning.


Map at the New York Times, 2012.


Kirk asks - How do you blend mapping and storytelling and where do you see it going, particularly on mobile?


Andrei - What is hard is making this work in 3 inches of screen real estate. There's a lot of technical work that has to go into it. We balance exploration and storytelling by giving people some scenarios and interesting maps that are in the data. So you present them with curated stories that come from the data. That was something we couldn't replicate on mobile.


Amanda - Some of our mappers are better at this. Navigational maps have taught people the basic things about mapping. The interesting things are less about how I scale this and more about how we interpret things. There is lower hanging fruit. I often enjoy the mobile version more because it forces us to be better editors.


Kirk says he tries to teach students about choropleth and other thematic maps that can try to show more of the nuance and complexity of geographic patterns. He asks - "How can we be better about maps? How do you guys approach the day after the election?"


Amanda - She says their favorite new trick is mapping how votes come in live. But the general thing is to try to explain why something happened. How do we understand that?


Karl - The audience day after will be wondering who their neighbors voted for. We are leaving people behind who don't like interactive experiences, who don't appreciate data. A challenge down the road is how to include the great percentage of Americans who are not as engaged. His audience has always been the small guys, the audience in Helena, MT. They are not as well-informed. Are they being left behind? Does this all come down to storytelling? Rolling over things is not the same as storytelling. A bunch of data coming up on a hover might mean nothing to them. Maps are beautiful and wonderful but people have to find these interactive things and seek them out.


Audience Questions


Laura Amico - Where do you see stories not being identified? What about editing?


Karl - As news producers we make a lot of assumptions about the readers. Just showign a graphic - it's often missing context. There's an assumption that readers will always be able to understand the context, why we are showing you this. Beginning, middle, end. Cause and effect. Showign the storyline is really important as part of the graphic. They can often confuse more than offer clarity.


Professor - A question related to drilling down. How many people are actually drilling down? Do you track that?


Andrei - Just looking at how many people come to a thing is not enough. We do track how people interact with our projects. Maps are pretty high in number of interactions and in time on page. Average is 3-5 minute range which we think means they are interacting with it in depth.


Audience - Question about AZ primary. Decreased precinct locations. If you could use maps to determine where polling locations should be how would you do that?


Amanda - Great question. I would try to make a model for where people live and the time of day they are likely to vote.


Audience - What fields outside politics inspire you the most?


Andrei - We do a lot of sports and mapping around sports. Kirk did maps about the intersection of sports and culture and the geography of fandom. Our economics writers do a lot of graphics. What subjects that we cover have geo data available?


Amanda - The foreign coverage is the natural home for NYT maps and the data isn't as substantive. Geography really still matters.


Karl - I like looking at Buzzfeed and seeing what the young people put out there. Seeing what younger people make for each other is always very inspirational. The fringe folks doing these crazy things and they are fun and engaging. There are ways to connect with different audiences.


Kirk - There are incredible datasets out there behind locked doors - at Google, Facebook, and so on. Occasionally we get a peek into them. But there is an incredible thing going on where these platforms are harvesting all this data and they don't share this data. That's what I want to see more of.


Andrei - Us too. We tried really hard to get Facebook to share more stuff - about ads, about interactions - and they have a lot of concerns about protecting user privacy but also about protecting their business.


Kirk - There are good reasons they are not sharing but I wish we could meet halfway.


Audience - Could you talk about the population distribution problem? Are any alternative formats


Andrei - Josh from Govtrack made a critique of 538 and said you should only use cartograms in order to demonstrate the population. I feel like that underestimates your audience. When you do cartograms there is a place for that. But they are complicated. But geographic maps are more readble.

Kirk - We tried to build a congruent depiction of whatever world we are trying to map. Often we are time to map the electoral college, not the American country.

Karl - There's a lot of free stuff out there for generating maps. There is not one for cartograms.

Audience - What are the tools now for doing this?

Andrei - We use a lot. QGIS, R, d3.js, HTML/CSS, Python.

Karl - Plug-in for Adobe Illustrator called Map Publisher. Adds a maps tab to illustrator and it will generate the map. Stays continually linked to the data. Great for basemaps.

Amanda - What Andrei said. NYT code on top of d3, a Javascript library written by Mike Bostock.