Getty Images/Tetra images RF
Election data revolutionizes the running of campaigns
Political campaigns used to rely primarily on polls to inform decisions, but with advances in technology over the past 15 years they are now using advanced analytics techniques.
Data is changing the way politicians run election campaigns.
Polls were the primary source of data for decades, informing decisions prior to the start election campaigns.
Now, however, campaigns have a trove of data they can use to inform campaigns not only at the start but also throughout the election process leading right up to the election itself.
Census data plays a role in deciding who should run in a given election and what messages to make central to a given campaign, but beyond the census, information about the economic and racial makeup of a given state or district is now collected more than once every 10 years. The previous voting patterns of the electorate -- who votes and how they voted in the past -- is illuminating.
And social media data, the behavior of potential voters on platforms such as Facebook and Twitter, is now crucial and took off after the advent of smartphones in 2007.
Advances in technology, meanwhile, now enable campaigns to collect, manage and action all that data faster than ever.
Michael Cohen, founder and CEO of Cohen Research Group, a political consulting firm, is a veteran of numerous political campaigns having started his career as a field representative for the Republican Party in Florida after earning his Ph.D. from the University of Florida in 1996.
On June 15, Cohen's first book, Modern Political Campaigns: How Professionalism, Technology, and Speed Have Revolutionized Elections, will be released.
Cohen recently discussed the intersection of data and politics he explores in his book, and in particular how data has changed election campaigns over the past 15 years.
The title of your book references how technology has played a role in revolutionizing elections -- what are some of the specific technologies?
Michael Cohen: The core one, and the one everything rests on, is data. That's access to data, the connections between data sets, and how data is used. The revolution all goes back to data. It's not as sexy as social media or some of the other things, but social media is built on data. The way campaigns act in the current environment, they need data to make decisions, so that is essentially where everything flows from. Because of advances in computing, and the advances in the different ways to use data and combine data and work it together, data really winds up being where everything comes from. From there, you can put together an effective social media campaign, you can put together a traditional media campaign, you can put together where people need to be on Election Day, and [figure out] how to motivate people with certain messages.
Everything goes back to the strength of the data.
What data points are election campaigns using to inform their decisions?
Cohen: There are a couple of things. I'm biased here because I work in polling. For me the whole idea of understanding what people think, what they want and what drives them is extraordinarily important at the front end of a campaign. It helps define how you're going to approach this kind of election. What you also rest on is preelection data from previous elections, information about a district such as whether people have moved in or out, the composition demographically, the kinds of work people do. Those kinds of things together, and any consumer information you have, is super helpful. It gives a picture of what that district looks like, and tells whether it will actually be competitive or if the real competition is in the primary.
Then, once you identify the population you need to move, you need to find out what their opinions are and what drives them, and all of that comes together with social media data, economic data, any commercial data you can purchase, and certainly any data about their prior voting.
Michael CohenFounder and CEO, Cohen Research Group
What does all that data enable?
Cohen: All of that data together moves you to the next phase, which is the communications phase. It helps define what messages [to send] and gives you a unified reason to run. One of the things campaigns often miss is that they have a candidate who wants to run, but they don't really know why, so you have to give an overarching theme to the campaign and from there, dive down into other messages that might be interesting for voters. If you can come up with the overall theme from the front-end data, from there you can start picking apart which messages you need to go through, then which are the priorities, then figure out what people are reacting to and then how you get them to vote.
Campaigns, number one, are looking at messaging. But they're also looking at how they can raise money, and how they can raise people. Those are the three main categories of what they do with data -- understanding the electorate, communicating with the electorate and then moving them. Data is extraordinarily important in all phases of that, and on top of that, you're working in an environment that's much more dynamic than [two decades ago]. ... You have to be very targeted with understanding who you can move and who you can actually reach.
What was the data people were looking at two decades ago, and how has it changed since then?
Cohen: Fifteen or 20 years ago, the best data available was polling data. Now you have so much more, and because you have so much more it doesn't mean that the polling doesn't matter, but the polling can be put into context. You might look and say a race looks super competitive based on the polling, but then you look at other patterns and data you have -- what you're seeing online and what you're seeing in the economic data -- and see that maybe the race is not competitive. With all that stuff together, including the polling, that's how it's changed.
The first data that really revolutionized politics was polling. It brought us all the way up to the advances in computing, and I would even say mobile computing in 2007 when the first iPhone was released. We used to rely much more on polling, but now it's part of a suite of data that candidates are looking at.
What's a real-world example of a campaign using data to inform decisions that swung an election?
Cohen: The 2016 Republican primary. Ted Cruz was in a primary battling it out with some pretty tough sharks, including the future president, and he was trying to figure out how to win Iowa. His campaign was able to ... pinpoint exactly who they needed to communicate to, why and how, and what they needed to do in the field to actually get him over the hump. In Iowa, it's all about on-the-ground and moving human beings, as opposed to California where it's much more of a mass media campaign. Without the expert use of data that WPA Intelligence used, there was no way Cruz would have even been in the top three.
Because he had that, he was able to manage his time, manage his resources effectively and get himself to win that primary. Cruz was able to come out of nowhere and become one of the main alternatives to Donald Trump.
Has one party, or one region of the country, embraced analytics more than others when it comes to elections?
Cohen: At this point, the parties are on par with each other. The Democrats were way ahead at first. Barack Obama's team was really the first to advance data plus communications. A group called BlueLabs did all the work for him, and they were fantastic. They were some of the pioneers. After the Republicans lost [in 2008], they looked back and asked, 'What do we need to do?' and part of the answer was needing to get parity in data. I'd say that now, the folks on the Republican side are just as good, if not better in certain cases, than people on the left. Now, to a certain extent, the question isn't just how you squeeze more votes out of your base but taking the next big leap and figuring out how to get people in the middle or who move back and forth -- or even people who are so cynical that they don't even vote -- and bring them in. That will be the next big leap over the next 10 years -- how do we bring them into the [political] system, and how do we move them with effective persuasion based upon data.
With respect to Obama, who at the time he ran for president in 2008 was a freshman senator from Illinois, was his analytics operation more advanced than Hillary Clinton's that year?
Cohen: Absolutely, and it wasn't even close. In fact, his use of analytics probably affected the primary more than it affected the general election because he was up against the Clinton machine, and the Clinton machine was no joke. When you're looking at that matchup, going into it, Obama was the clear underdog. He probably shouldn't have had any chance at all, but his team was so sophisticated with data that they were able to overcome certain deficiencies. They knew who they needed to get. If you listen to David Axelrod and David Plouffe and that whole team, they will tell you that they had a better handle on their voters than Hillary Clinton had on hers.
You saw that also in the race between Trump and Hillary Clinton. She had other deficiencies as a candidate, but she never really got the message on data. I have students of mine who worked on her campaign, and they reported back to me that they would tell headquarters that they were knocking on the wrong doors. They would have sheets with data about people telling them that the people were persuadable, they'd knock on the person's door and get told, 'I'm voting for Trump, get out of my face.' They fed that data back to headquarters in Brooklyn, but headquarters believed in [its own] approach and didn't make shifts within the campaign that might have mattered. When the margins were small and she lost by only tens of thousands of votes in areas, the data meant more. The Trump campaign was better about data than the Clinton campaign.
Are campaigns using traditional analytics reports and dashboards or are they now using more advanced analytics tools like machine learning models?
Cohen: It's the latter. The really good firms are running pretty sophisticated models. Each of those firms do a pretty good job of understanding how to put all the data together, and the machine learning piece is extraordinary. They're teaching the data to figure out the right answer for them as opposed to being forced to interpret the data. It's getting very sophisticated, and that's why my hope is that data, and all the approaches to handling data, advances to the point where we're no longer just pressing on each side of the pendulum but realize that the real nuggets are in the middle, and that actually brings us closer to what campaigns used to be.
Now, they're very convinced that the base really matters. Trump understood, or was convinced, that the base is really important, and that's where consultants are right now. My hope is that in five or 10 years, the modeling is showing other options to bring more people together.
Are campaigns developing their own analytics tools or are they using tools from commercial vendors such as Tableau, Qlik, Microsoft and others?
Cohen: It's all over the place. I can speak from experience and know that certain firms on the Republican side have entire data science departments. At WPA Intelligence, they've put together different types of data products, and what you're seeing is that a lot of firms that started out as polling firms have added core data science to their offerings. It's the same thing on both sides, and there are multiple different firms doing a good job. The real leading edge ones, the difference between those and the ones who have bought platforms like Power BI or a Tableau, are actually informing data scientists. That's been a big leap in the last five to 10 years. They're realizing this isn't just another line of business but helps core research. You're getting a lot more data science.
Can a consulting firm develop its own analytics platform and sell it to different campaigns?
Cohen: Yes. But the stuff that we're talking about is not easily understood, so you actually need the consulting behind it. The hope is that some of the data these firms sell ends up as a consulting relationship as well.
What's the future of data and analytics in election campaigns?
Cohen: The modeling is going to get better. It's going to be more precise. Nothing works in politics like losing. What really teaches campaigns is when they lose, so when they lose they ask, 'Why didn't that model work?' or, 'Why didn't this approach work?' That's when they get a lot better. On the left, on the congressional side, they're struggling to figure out why everything didn't work and they could actually lose the House of Representatives. That's probably putting the right kind of science pressure on them to get better. On the right, they probably feel they have more of the secret sauce than the left. This will be one of those things where one side will get a little bit ahead, then the other will catch up. It will essentially result in much more precision.
The other piece of it is that they understand their bases so well that they need to start selling something else, which means that they will start looking at that fuzzy middle that they're ignoring because the people at the base are much more easy to motivate. If they can do that, that will become the next wave of products. I'm looking forward to that because I think that will be better for democracy.
Editor's note: This Q&A has been edited for clarity and conciseness.