MOD: Masters of Data

Bringing the human to the data

Carin Robinson: Data, Democracy, and Politics

Associate Professor of Political Science, Hood College

June 26, 2018

33:52

A campaign is interested in whether or not they can persuade you.

We are going to talk about how data, and data analytics, is an essential part of today’s political campaigns, and the implications for the future.

Welcome to the Masters of Data podcast, the podcast where we talk to the people on the front lines of the data revolution about how data affects our businesses and our lives.

Our guest today works in political science. This field might seem like it’s miles away from some of our other guests, but the reality is that data and data analytics has penetrated every element of our society, even politics.

Carin Robinson, an associate professor of political science at Hood College in Maryland, is an expert on today’s political system in the United States. We’re going to talk about how data is an essential part of today’s political campaigns and the implications for the future.

I was super excited when you agreed to be here because I’ve become a total politics nerd over the last couple years. I started listening to NPR Politics and listening to podcasts, and it’s been an exciting few years I would say, right?

Oh, for sure. There are so many opportunities now to stay up to speed on what’s going on in interesting and fun ways, whether it be following the personalities in politics or following the numbers. There are many websites, blogs, and podcasts that I’m encouraging my students to pay attention to help them make sense of our current political environment.

Carin Robinson’s Start in Political Science

I know you, but the listeners don’t. I’d love for you to talk a bit about how you got to where you are. I know you grew up in Wisconsin, you went to George Washington University, got your Ph.D. at Georgetown. Tell us a little bit more about why you went into political science, what took you that direction?

I never anticipated going into political science. In fact, I have vague memories as a child being annoyed when the State of the Union came on because it was interrupting me watching The Cosby Show on a Thursday night. I stumbled upon it when I graduated from college in 2000. I had a degree in communications and got a job working as a journalist for a small newspaper.

width=

I started covering local city council races and at 21, I held a position of great importance in the political process of those small towns.

I wanted to better understand the relationship that media has with politics and look at it in a more objective analytical way., so I applied to programs, and ended up at George Washington University. I did a master’s in media and political theory there and then wanted to pursue a Ph.D. which I did at Georgetown.

My focus really became political behavior, the larger picture. Not just in media but all the ways in which people are motivated to participate in politics and the ways in which they’re not motivated to participate in politics.

Some of my research looked particularly at citizens who have religious beliefs that motivate them to participate in politics. I looked at the relationship between religion and politics in the 2000 election and have followed it ever since. I did a large survey experiment for my dissertation. I loved survey data and I’ve taught classes on US politics and teach a class on elections every fourth when we have a presidential race. I also teach the class on research methods.

I’m very passionate about people understanding politics through a systematic lens. We have a lot of people who have opinions about politics and opinions about our President and our Congress, and often not favorable. I love encouraging people to look at it in a systematic way that helps to evaluate things objectively, and not have these knee-jerk reactions based on our predispositions and our opinions that are not often founded on data itself. I love research methods even when the students don’t because I think it helps refine how they process what’s going on in our world today.

It makes a lot of sense. You know, I was in kindergarten when Reagan was running against Carter, and I remember they came to our classroom and asked who we wanted to vote for. I said, “Goofy.”

width=

That’s was my “political weirdness” at that time.

How Political Awareness Has Changed

During the 2016 election, my daughter was only in kindergarten yet she had a very specific opinion about who she should vote for. With your kids and with the students you teach, do you think the political awareness has changed in the last decade?

Well, the sources of their political information have changed. Yes, their parents, environment, and social networks inform their understanding, but we’re also seeing a lot of satire. There is The Daily Show and various social media postings that tend to emphasize everyone’s predispositions. What we’re finding is the opinions that people are espousing are ones that they have always had and then they seek out information that just confirms their predispositions. We’re not seeing a lot of grace, deliberation, or refinement as people come to hold these attitudes.

I’m careful with my own children to ask some questions and help train them how to think about it intelligently, deliberately, as opposed to just adopting some funny remark that they heard someone else says.

People seek out information that confirms their predispositions. I am careful to help train my children to think about the political process intelligently and deliberately.

Now there are so many funny remarks about our leaders. I try to steer clear of that just because I think there’s an opportunity to look at this with a mature perspective. I want them to think of American politics in a way that will help them make sense of it 10 or 20 years from now, when we’re looking at a different group of representatives and Senators. Maybe not an entirely different given the likelihood of being reelected these days and incumbency advantage, but there’ll be a different President, and there will be different issues and conversations that we’re having. I want them to have a framework that works for the long term and doesn’t just get them a laugh on their social media post.

width=

I like that. Maybe you’ll help me set up a correspondence course for my kids.

Right, okay. I’m not quite there yet. Still just focusing on the 18- 19-year-old people in my class who need some work as well.

I think that everything you’re saying about the kind of echo chambers that we live in is astounding.

What Happened in the 2016 Election

It seems like in this last election we got it so wrong. I mean the data just seemed completely off. What’s changed in our understanding of the process in the last couple years?

I think we’re just now starting to get our bearings on what happened, and I it’s going to take another cycle or two for us to feel more confident about correcting some errors that did take place.

In the 2016 election, the polls were not as off as people want to think that they were. We’re so consumed with the Electoral College that we kind of lost sight of the fact that Hillary Clinton did win the popular vote by two percentage points. Most polls had her winning the popular vote by about three percentage points. They were off on average 0.9 % on the popular vote.

In most circumstances and other elections, we wouldn’t bat an eye. We would have said they were spot on, but given the discrepancy between the Electoral College vote and the popular vote this time, there is not much celebrating going on in terms of how accurate our pollsters were.

To their credit, public opinion researchers, pollsters from newspapers and from the Academy have done a lot of evaluating about their processes that were in place. A couple things that they’ve concluded are accurate conclusions that will help us moving forward and really speak to the fact that we need to pay a lot more attention to the poll results at the state level. So, it wasn’t so much that national polls were off, it was that polls in Pennsylvania, Michigan, and Wisconsin were off. They were off by a pretty substantial degree, so why were they off?

One, it’s been noted the last couple years, how college-educated citizens were much more likely to vote for Hillary Clinton than they were for Donald Trump. That correlation was more pronounced in 2016 than it was in previous cycles. We also know that in survey research, college-educated

citizens are much more likely to respond to surveys.

So, they are over-represented in surveys, and there are weights that are used already to account for that. Because we knew that, we used weights to account for that, but in 2016 that college education meant you were way more likely to vote for Clinton. Because that variable wasn’t weighted properly, the polls were a bit skewed.

width=

The other thing that the reflective pollsters and public opinion researchers have noted is that there are a chunk of voters who were undecided, about 13% in Pennsylvania, Michigan, Wisconsin, and Florida. In past elections, those undecided voters have evenly split and they just really haven’t been the game-changer that they ended up being in 2016.

Those undecided voters swung to Trump in a large margin. In one state, I believe it was Florida, it went 30 points more for Trump than for Clinton, and so those late deciders made more of a difference in these state polls that we could do prior to the election.

Those polls said that Hillary Clinton was likely to win, but had we taken a poll the morning of the election it might have tapped into those late deciders who ended up swinging for Trump.

Those are the two big explanations that there’s a lot of enthusiasm for. There was some speculation about a shy Trump voter effect. There were voters that just didn’t want to admit they were going to vote for Trump because of social desirability bias, but there isn’t a lot of evidence for that right now.

Moving forward we’re going to pay more attention to polls done at the state level. The interesting conundrum there is that we rely on the media industry for these polls, and with the decline of local newspapers and even state news media, we are not going to have access to some good polling done at the state level. It’ll be interesting to see how that can be corrected in the future.

After the 2016 election, we’re going to pay more attention to polls done at the state level.

I was reading an article in The New York Times the other day, that said there were a lot of people questioning the efficiency of these exit polls, and that maybe they skewed a certain age group.

Yes, the exit polls are interesting. They’re done by Edison Research. That’s one company that most media sources then subscribe to and rely on. They are brainstorming how this can be done differently, even distributing a website that people can go home log into and fill out. That way they are not relying on face-to-face conversations immediately as people leave the voting booth. I think that’ll be re-examined as well.

How Data Is Used in Politics

I remember when Obama was first elected, it seemed to me that data had a much more prominent part in that campaign. Didn’t he have a CTO or something on the campaign? So, talk to me a little bit about how you see data being part of these campaigns today and how they’re dealing with it.

Numbers have always been part of the story but what’s exciting about right now is that everybody wants to talk about the numbers, and we’re fascinated about this process. I believe we’re longing to bring some concreteness to a world that has become a bit unhinged when we’re caught up in Twitter and random posts, and the sagas and the dramas that are going on behind the scenes.

I think we’d love to have these numbers to really talk systematically and quantitatively about politics today. But, yes, President Obama is credited with really ushering in data analytics to campaigns. Data has always been around, but that campaign really did hire some key quantitative strategists that helped solidify data analytics in campaigns today.

width=

It’s interesting to compare the number of supporter email addresses that Democratic presidential nominee John Kerry had in 2004, with the number that President Obama had in 2008 and 2012. John Kerry had 3 million email addresses of his supporters, who he could send emails to, and contact. President Obama had 13 million in 2008 and 20 million in 2012. That’s how dramatically the landscape has changed.

Data analytics are a key part of campaigns today. We’ve always been interested in running campaigns that are cost -efficient and resource-effective. States, since the Constitution was written, have had a large say over how elections were conducted, and for quite some time they have always collected information about who registers to vote and who votes.

States don’t keep a record of who we vote for, but they keep a record of who votes and who doesn’t vote. Now, those records were on paper for quite some time. There was not great reliability over how those were kept. When we had the election of 2000, Bush vs. Gore brought to light all the inadequacies of our voting structure and systems, the variability across states, the inconsistencies of how many people we thought had voted, and then how many votes were actually cast.

Congress got involved. They passed the Help America Vote Act of 2002, and that really streamlined voting records. It actually required states to keep digital records of who voted, who is registered to vote, the names and addresses of those people, and in some cases contact information and gender.

For political scientists, it’s exciting that all that is required by law, because all that data is available. All of a sudden, those sources of data became much more reliable. Now states have the option of how accessible they want that data to be. In some states it might cost you two dollars to access to that data, in other states it costs thousands of dollars, and you have to be a candidate or a representative from the state party. There’s great variance as to how accessible that data is, but in theory, it’s publicly available. The Help America Vote Act in 2002 ushered in access to these voter lists, and we call them voter files.

Now we are getting that information from secretaries of states, and then we’re combining that with other data. The census, also freely and publicly available, can be added to your voter files. Then, we’re working with consumer organizations and commercial firms that are adding all sorts of consumer behavior, credit scores, home ownership etc to the data. So, these files are just getting huge.

width=

I heard a little bit about that Carin, but that actually sounds a little scary. So now they are looking at my credit score, and my credit card purchase history to determine who I’m going to vote for.

There’s a big question over privacy. It’s kind of ironic because you do see some candidates such as Ted Cruz in 2016, running for the Republican nomination. Cruz was concerned with government intrusion and privacy issues, but then he was noted for using some pretty big data files that have a lot of information. Yet he justifies that in the name of wanting to run an efficient campaign and contact who he thinks he should contact.

Privacy is a big question here. I highlighted the sources of the data, but I failed to mention that citizens themselves are making their own data available as soon as they offer an email address to a campaign, or to some partisan group entity. The group has access to a lot of things just by virtue of having your email address. So, we are at fault as well for giving information about our allegiances and our associations via our email addresses or if we like them on Facebook.

Let’s say I sign up for some campaign email list or some cause I’m interested in and they have my email. So, they use that to connect me to other databases they have. Is that what you mean?

That’s what I have been reading about. I haven’t worked on the campaign directly so I’m trying to keep up with the pace of this world as well. My understanding is that the email is a goldmine for any association that that email has with other organizations.

Companies can start tracking where that email pops up, and that helps them build a profile that says who you are, what you like, what kind of messages might appeal to you, how likely you are to donate money, or how likely you are to volunteer for them. They build a profile and they try to make sense of who you are to know whether or not you are worth their time.

Psychographic Micro-Targeting

When you and I were prepping for this, you sent me this term which I love. “The psychographic micro-targeting.” It sounds like something out of a spy thriller.

I wrote it in big letters so I would make sure I spoke of it accurately because I botch it all the time. Psychographic micro-targeting. I say “micro-targeting” for short.

When I heard that I was thinking of something like psychographically micro-targeting with extreme prejudice. Something I would hear in a spy movie, but what does it actually mean?

width=

What it means is that in some campaigns today there is interest by some candidates and parties to get into the nitty-gritty of citizens and who they are. As I was alluding to earlier, the campaign is interested in whether or not they can persuade you.

To that end, they want to know are you an agreeable person, are you a neurotic person, are you very open and engaging, are you an extrovert, are you an introvert? That information is gleaned by your purchasing behavior, and what groups you belong to and what magazines you subscribe to. Using that information, they then decide if they contact you, and send you some kind of persuasive direct mail.

There was a lot of media coverage on the Ted Cruz campaign, in 2016, when he hired Cambridge Analytica to help him conduct this psychographic micro-targeting. If there was an individual who was known to be fairly neurotic, fairly fearful, a campaign would then run an ad featuring a home burglary and have it include a message about being pro-gun, and the candidate being in favor of the second amendment. That would be a uniquely tailored message that would be then sent to this person who the campaign knew was slightly neurotic, and their thinking was this is going to be more effective in earning this person’s vote, or in encouraging them to volunteer.

Now similarly, if you find someone who’s very agreeable, they would maybe respond better to an ad that emphasized family values, and they would be quite turned off by some ad that featured a gun and a home burglary.

There’s a lot of work now being done to determine the effectiveness of this, and I wonder, does this really matter? There’s not a ton of evidence to suggest that this gets you much further than a more generic voter file would.

What matters more is if a person votes on a regular basis, and if they are Republican or Democrat. We don’t really know if all this is adding a whole bunch more to the story, but it’s being attempted.

Ted Cruz did not win the Republican nomination in 2016 so at the very least that sample size doesn’t suggest that this is absolutely necessary to win a campaign. But some people are pursuing the effort to really zero in on certain personality types and market messages in such a way to appeal only to those personality types.

Not coming from the political science academia myself, it does sound remarkably like the kind of marketing efforts that you’re seeing in large internet companies, particularly what Facebook has been doing.

They’re basically micro-targeting me with ads and that’s what they propose to advertisers just like in Obama’s election where they were bringing in social media and they were able to show that it was effective for campaigns. I guess it shouldn’t be that surprising that we would try this in political campaigns, because it’s been so successful for commercial advertisers, right?

Right. As a Political Scientist, I have to question if this is really worthwhile. I think the jury’s still out on that. As the data continues to increase, and consumers and citizens provide that data by making our social media profiles so public, at the end of the day it’s still about if someone is actually going to go and vote, and what drives that vote choice.

As a Political Scientist, I have to question if psychographic micro-targeting is really worthwhile.

Sometimes it’s as simple as if someone is Republican or Democrat and if they vote. Because we find that we don’t often get people who vote in one election and don’t vote in the next. We have people who vote all the time, or we have people who don’t vote at all.

width=

Campaigns try to muster a few of those people who sit in the middle because sometimes those people are the people who end up determining the election result, but I don’t know if we’re really tapping into something that profound by being able to analyze our potential voters to this extent. And of course, there are privacy concerns involved as well.

There’s a risk in profiling someone as an agreeable person, or a neurotic person, or an extrovert, or a conscientious person, and then forming an ad to target them. Sometimes it will totally have a boomerang effect and not accomplish what you wanted to accomplish. I don’t think citizens enjoy being pandered to.

Again, I think more research needs to be done for us to know if this is absolutely necessary and effective, and of course, there’s the philosophical question of if this is good for our system of government.

There’s risk in profiling someone’s personality and forming an ad to target them. It can have a boomerang effect and not accomplish what you wanted it to.

Tribalism in Data and Politics

It seems like part of what we’re talking about is tribalism. We tend to exist in tribes, and those tribes have opinions. I really enjoyed reading Truman’s biography a few months ago. The book talked about how Truman and Eisenhower essentially picked their parties based on their families. It’s not a new thing. This is the way we operate.

No, it’s not new. In fact, when I applied to graduate school I wrote my statement of purpose which encompassed what I wanted to pursue research-wise. I looked back at Madison’s campaigns from back in the day and how he went after certain religious citizens because he was a member of a particular church and he wanted to mobilize the citizens of the church that he belonged to.

I was doing this in the early 2000s applying to graduate school. I mentioned how we thought George W Bush, and even Al Gore, speaking of their faith in such a strategic way wasn’t so great. But from the dawn of time, people have been tapping into these social identities to mobilize citizens.

I know for some that “tribalism” was the word of the year in 2016 because it really does govern so much of our political behavior. We know what Republicans or Democrats like to eat, where they shop, what they like to buy, what celebrities they like, what soda they like to drink, all sorts of things that are different if you’re Republican or a Democrat. This data has come with these communities that have formed and this team mentality of winning and losing.

We definitely saw a lot of that with President Trump’s rhetoric in the 2016 campaign. He promoted an us-versus-them rhetoric which some were fans of and some were not, but it was consistent with this notion of social identity governing how we approach politics. What we look for in a candidate is really an affirmation of what we think to be true about our social identity.

“What we look for in a candidate is really an affirmation of what we think to be true about our social identity.”

I have to go back to something you said. So, you’re seriously telling me that Republicans and Democrats drink different sodas?

width=

They totally do, and I wish I had the data on that. There are differences in what cars they like to buy, and if they prefer Wii or PlayStation. There are distinct differences, because with those choices come a host of other things. It’s socioeconomic status, it’s your race, and in some cases, purchases are driven by your age.

You can connect all sorts of dots and see that it’s not so much that this group likes Dr. Pepper and this group likes Pepsi, it’s the cultures that they’re in. It’s more about if it’s rural or urban, from Wisconsin or New York City, and there are a host of things that are connected to those consumer choices.

Those cultures like certain things.

Yes. It’s kind of fun to highlight that, and to put yourself into those lists, and say, “Oh yes. I guess that is what I am because I like that soda and that brand of coffee.” So, yes, there are differences that are that mundane but seem to really register as being quite an accurate description of the two-party partisan identities.

Unique Data Strategies in Politics

With you being in the middle of this, what are you seeing different, what are you looking at in the midterms from your academic perspective?

I am teaching campaigns and elections in the fall. I am monitoring some of these primary races to see what strategies are employed, if we do see more of the micro-targeting efforts, or if we’re seeing more traditional efforts to just simply get out the vote.

I just wanted to highlight one thing that I took note of last week. There was a gubernatorial primary race in the state of Georgia, and the woman who won Stacey Abrams is using a unique strategy of targeting black voters.

She rightly noted that about 1 million black voters did not vote in the 2016 election, and we do know that there was a lower turnout amongst demographic groups that we thought were going to turn out for Hillary Clinton and there was higher turnout for groups that favor Trump in 2016.

She’s really making it her mission to turn out these black voters, and really banking her campaign on it. It’s a unique strategy because her opponent in the primary really wants to appeal to the more moderate center. Usually in a red state like Georgia you need to appeal to some Republican voters. Stacey Abrams is asserting that maybe not needed.

I also think this midterm election, and future elections, are going to reaffirm that even with all the data that we have access to there’s still the nuts and bolts of campaign work that we’re never going to escape. There is always going to be door-knocking, there’s always going to be phone calls, there’s always going to be rallies, there’s always going to be volunteers holding signs, and putting signs in their front yard, or bumper stickers on the car.

As much as observers of political campaigns are getting excited about data analytics and how the field is changing, I think we still need people who are going to be walking and talking that campaign work.

While I expect to see in the midterm some unique and different strategies, I think given 2016 results, nothing is taken for granted. With Democrats in particular we’re going to see some variation as to how they attempt to make sense of the blue-collar voters who swung to Trump, and how are they going to be drawn back into the Democratic Party.

Those are some questions that I think the midterms might help us answer. We shall see, and we’ll see if there’s the Democratic wave that people are expecting. I think there’s a lot of time left on the calendar for things to change, so I’m looking forward to it.

I believe it was on NPR, there was some piece about a woman running for Congress in Montgomery County where you live. She targeted specific people based on very basic data. She knocked on certain doors to talk and not others based on the data she found.

That’s what I mean. Before we had access to the digital databases from the states following the Help America Vote Act, we had very crude data. Many campaigns used a simple scale of 0-4, and that represented how many times a person voted in the last four elections and whether the person was a registered Republican or Democrat. That alone would tell a campaign which doors to knock on.

Because you don’t need to waste your time on people who have voted in the last four elections for the opposite party, and you might not even need to spend a lot of time on the people who voted in the last four elections for your party. You might want to spend some time on the people who voted in the last two or three elections for your party, and you might even want to spend some time on the people who voted in the last four elections, but they split their vote. So those are two very simple variables that really tell you if that person is worth the campaign’s time.

You may use the data to guide you there, but it’s still people talking to people voting for people.

And these are people who change their minds. I think of a die-hard Republican who all of a sudden has some medical condition that really, really needed Obamacare. All the data might have pointed to the fact that that individual would have voted for Trump or for Romney in 2012, but that health event may have pushed them to the other side. So, we don’t want to oversimplify the world, we don’t want to oversimplify a person.

Yes, absolutely. Carin, maybe we’ll reconnect after the midterms to get your view on how that all went through.

Yes. Give me a couple of days. I couldn’t really talk on the 2016 election, because everything I’d been teaching the two months before went out the window, and I had to regroup and my students had to regroup. That speaks nothing of my partisanship, it just speaks to the fact that academically speaking I never would have expected us to elect someone with the lack of experience, with the three marriages, with all sorts of things that traditionally have meant the person is not a very viable candidate.

If nothing else, it is going to generate lots of work for you in the next couple years.

Well, Carin, I really appreciate you taking the time to come on with me. This has been a great discussion.

The guy behind the mic

Ben Newton

Ben Newton

Ben is a veteran of the IT Operations market, with a two decade career across large and small companies like Loudcloud, BladeLogic, Northrop Grumman, EDS, and BMC. Ben got to do DevOps before DevOps was cool, working with government agencies and major commercial brands to be more agile and move faster. More recently, Ben spent 5 years in product management at Sumo Logic, and is now running product marketing for Operations Analytics at Sumo Logic. His latest project, Masters of Data, has let him combine his love of podcasts and music with his love of good conversations.

More posts by Ben Newton.

Listen anytime, anywhere

Available to stream or download via these and other prodcast apps.