You are viewing:

ArchivedContent

Information released online from January 20, 2009 to January 20, 2017.
Note: Content in this archive site is not updated, and links may not function. External links to other Internet sites should not be construed as an endorsement of the views contained therein.

printable banner

U.S. Department of State - Great Seal

U.S. Department of State

Diplomacy in Action

Social Media Trends: Election Edition

Robert D'Onofrio, Director of Data Communications, Facebook
Cleveland, OH
July 21, 2016




Date: 07/21/2016 Location: Cleveland, OH Description: Dr. Robert D'Onofrio, Facebook's Director of Data Communications, discusses social media trends during this presidential election. - State Dept Image

4:00 P.M. EDT

THE REPUBLICAN NATIONAL CONVENTION, CLEVELAND, OH

MODERATOR: I would mention that if any of you are publishing stories based on our briefings, we would love to get links to your stories. As you send them to us when we’re back in New York and Washington, we would really appreciate that. It helps with our work.

For our last briefer we have Dr. Robert D’Onofrio. He is Facebook’s Director of Data Communications and he is here to talk about social media data and trends during this RNC. He is also going to be available at the DNC, so those of you who are going there.

I should note that our speaker’s views are their own and do not represent the views of the U.S. government. After the briefing, we’ll have a Q&A. Please do state your name and media organization when you ask a question. Again, thank you so much for coming to our space at the Republican National Convention. We hope to see you at the DNC, as well. Thank you.

DR D’ONOFRIO: Hello, and good afternoon, and thank you for spending the next half hour with me as we walk through Facebook’s data.

Like Monica mentioned, I’m the Director of Facebook’s Data Communication, and that means I have a team of researchers that focus on big events around the world and try to tell stories by using data to help journalists, help the media, help consumers, and really anybody out there just sort of get a sense of understanding what is happening on Facebook.

The way we do this, and first we’ll take a look at the candidates in the election, is by developing a key word list that allows us to find out, like if somebody is mentioning Mr. Trump, if someone is mentioning Hillary Clinton, if someone is mentioning an issue related to the election specifically. And we sort of build off of that to come up with a couple of different ways of looking at the data, and I’ll walk you through an example with Mr. Trump to help you understand the way we kind of think about social conversation and the way we reflect that conversation through the use of this touch screen.

So, most of the data I’m about to show you is over the last 30 days on Facebook. We have been updating this throughout the convention, so you will see some spikes over the last couple of days in conversation, and I’ll explain those and walk you through those, as well. But for the most part, what you’re going to see is about one month’s worth of Facebook information.

If you look at Donald Trump over the last 30 days, the first thing I would note is about 31 million people have been talking about him. That’s 31 million unique people on Facebook in the last 30 days have specifically mentioned something related to Donald Trump. The number below that, interactions, 266 million interactions, is the total of all of the post, comments, likes, and shares that those 31 million people made.

So, one way to think about this is, you know, there was this conversation that’s been taking place on Facebook over the last 30 days. 31 million unique people posted or commented or liked or shared something related to Mr. Trump. And that came to the tune of 266 interactions. That’s kind of like the scale of the conversation that the 31 million people had.

We also take a look at a few basic demographics, and of course this is all done in aggregate and anonymously, so it’s a very privacy-safe way of taking a look at these trends. About 57 percent of the conversation came from women on Facebook. Another important note here: we don’t distinguish between positive post or negative posts or neutral posts. So, this is just any mention of Mr. Trump. About 57 percent of the conversation came from women, and you can take a look at the age categories to see which age groups were participating in that conversation as well.

What I really find interesting is the line at the bottom where you look at those last 30 days. That’s the overall volume of conversation and the numbers you see on the left hand side - 13 million, 6.5 million - those are the number of people on any given day talking about Donald Trump. Not surprisingly, the last few days as we’ve headed into the convention, there’s been a sharp increase in that conversation.

Going back to the candidates, let’s take a look at Hillary Clinton. Over that same time period, she’s had about 25 million people talking about her, and also well over 200 million interactions. So, a rich conversation about both of the candidates. Not a surprise that we’re at the RNC so we saw a little bit of an increase in Donald Trump’s buzz, so 31 million people talking about him. We would expect this number to increase a bit at the DNC in Philadelphia next week. And as Monica mentioned, I will be there, as well. I’m happy to share this information with you there, as well.

The demographic split: pretty similar to what we see for Mr. Trump. And, again, what we see with these spikes on the -- whenever we see a big jump like that it always correlates with a new event. So, as the news came out about no charges being pressed with the government ethics scandal, and other sort of points along the way we’ll see some of this jump up next week, as well.

Any time there’s a bit increase in Facebook conversation, it’s being driven by a news event. And I think that one of our big learnings as we think about the 2016 election and the way social media has enabled and facilitated participation in the conversation about the election, it’s a really rich conversation. 89 million people since the start of the year in the United States have talked about the election in some way, shape or form, and we’re still months away. We haven’t even had the general election debates. I’m really - we’re fortunate to be able to look at such a rich conversation, because that enables us to then look at the issues.

These are the top five issues people have been specifically talking about on Facebook over the last 30 days, and these really shouldn’t come as a surprise. There are explanations behind each of them. So, crime and criminal justice, number one, I think with a lot of the tension between police officers and Black Lives Matter, not a surprise that that’s on there. Government ethics, number two, connected with the Hillary Clinton email scandal. Racial issues is actually kind of interesting, was around six or seven heading into the convention a few days ago, now we’ve seen it pop into the top five, just based on the last few days of conversation on Facebook. So, a lot of focus on racial issues. Guns is number four. Then homeland security and terrorism number five. Again, with the attacks in Nice. I think these things are very much top of mind.

These change over time. It will be interesting to see how this shifts next week as the Democrats sort of go through their agenda and focus on the issues they would like to highlight. We have the ability, though, to look at a number of issues. So, this is the sort of holistic picture of all the issues that we’re sort of able to take a look at that can pop into that top five based on what people are talking about on Facebook.

If we wanted to drill in a little bit more, let’s say into crime and criminal justice, we can also see the demographics behind that conversation as well as the age distribution of who’s participating in that conversation. About 8.3 million unique people over the last 30 days have been talking about crime and criminal justice and that’s our number one most discussed policy issue on Facebook. And, again, you can see how that sort of spikes over time. And with all of these issues you truly can look up what happened on these days when you see these spikes, and it’s a news story that’s driving it.

Another way to visualize this, and this is really one of my personal favorites is our time line. That’s what we had up on the screen as people were sitting down as we started the briefing. This is one of the few visualizations that goes beyond 30 days. We actually go all the way back to last August when we were all here for the first debate.

What this shows over time is first for the candidates and then for the issue, how the conversation grows and shrinks about each candidate or issue over time. As the words pop out and get bigger, it means many, many people are talking about it; and then as it gets smaller, that means the conversation is starting to shrink a little bit. And as candidates drop out of the race the sort of drop out of this animation.

The bottom sort of graph plots out the conversation about all of the candidates. So, again, when you tend to see some of these spikes you’ll see a few of the names pop out and it will be directly connected to news events. And you can see the field thinning here a bit until we’ll eventually have just the main nominees.

QUESTION: I’m probably breaking the rules here. I’m Kris Ronneberg with the Norwegian newspaper Aftensposten.

I saw Mike Pence there throughout. Is that something that you added afterwards? I’m guessing that he wasn’t part of the conversation the whole time.

DR D’ONOFRIO: We actually have been taking a look at a lot of different potential candidates and a lot of names along the way. So we started out with a really kind of comprehensive list of who might enter the race, even if they didn’t. But we did sort of focus on, for the purpose of visualization, just people who are actively candidates in here. So, we added Pence in once that nomination was announced, and we’ll do the same with Hillary’s VP pick once that’s announced, as well.

So, we have the ability with this framework to sort of add in new entities like that, and the same thing on the issues front.

So the same thing with the issues, I won’t play the entire time line, but I’ll just let some of it play here starting in October, and you can see out of those 25 or 30 issues overall that we have the ability to take a look at, how different news events drives conversation about those issues on Facebook.

The last thing I’ll show before we open it up to Q&A. I talked about gender, I’ve talked about age, we’ve talked about overall volume, but we haven’t talked about geography, yet. So, one of the really interesting things about this election is going to be which states will help decide who’s going to win. What are those swing states? And within those states, what are the counties that are going to matter the most?

So, we have the ability at the county level to break out that conversation both about the candidates as well as the issues. So, if you start by looking at the candidates, the darker areas represent above-average levels of conversation in those counties, in this case about Donald Trump. So, not a big surprise, the consistent pattern has been he has a lot of above average support along the border in California and Arizona, parts of southern Florida, and also parts of the eastern seaboard. That doesn’t mean that the rest of the country’s not talking about him. Again, there’s millions of people overall talking about Donald Trump. It just means that the rest of the country is talking about him at a pretty normal pace, but along those border states, the conversation is particularly high.

When you look at Hillary Clinton, you see that the support is a little bit more widespread. She has less of those areas of combination that kind of pop out, but you do see some sort of lighter areas in California. I think that as we look at Jill Stein, the Green candidate, what we kind of see here are probably former Bernie Sanders supporters talking about Jill Stein as an alternative, and trying to think about what they’re going to do with their vote.

We can do the same thing, it’s pretty straightforward, Gary Johnson, former Governor of New Mexico. Not surprising that’s where most people are talking about him. And Mike Pence, we can again guess that Indiana will be where the conversation is strongest. And we can do the same thing on the issues. So we can look at any of the issues that we have the ability to look at the county level to identify where people are especially talking about that issue over the last 30 days.

The final thing I’ll wrap up with: another map that is at the state level that helps us understand the state specific conversation, again about the candidates and the issues. Let’s just take a look at Ohio. Over the last 30 days, you can see the conversation is pretty balanced between the two candidates, with Donald Trump having a little bit of a lead in terms of people and interactions. We can also look to see what issues are resonating within Ohio specifically, and you’d note that it’s different than the national picture.

So, if you wanted to take a closer look at some of those swing states and what the issues are in some of the swing states, you know, the idea is that this can help people get a sense of, you know, supplement the other information that’s out there by learning what people are talking about real time on Facebook.

With that, I’d love to take your questions.

QUESTION: Hi. Just to start off the conversation, I guess. Yashwant Raj, from India newspaper, The Hindustan Times.

So, the last slide that you showed about what people are talking about, so these are search terms, or these are actually conversation topics? When you say North Korea, for instance, so, are people talking about North Korea or Cuba? Or, are they looking for what the issues are on Cuba or North Korea? I understand criminal justice at the top trending topic, but North Korea and Cuba I’m a little intrigued about.

DR D’ONOFRIO: They are based on conversations, so it’s not anything that people are searching for through the Facebook search box, but the actual posts that people are making that mention something related to North Korea by name, whether it’s a rocket, a missile test for example, or with Cuba with the relaxation of some of the policies, with visiting opening up a little bit. Any of those specific terms that are connected with that policy issue is what contributes to how we measure things. So it’s not based on searches; it’s actually based on somebody has to type key words that are directly related to the issue in order for us to measure that.

QUESTION: My name is Hiro Aida, Japan’s Kyodo News.

I’m just wondering whether is there any way for you to get a picture of not only conversation how many times people are talking about, but how many times people are talking about, for example, the name of Donald Trump, negatively or positively. Is there any way for you to do that?

DR D’ONOFRIO: That’s a really good question. Sentiment analysis is something that we’re definitely interested in exploring, and it’s something that is a natural question to ask here because it helps you better understand that conversation, especially the demographics of the conversation. It’s not something that we currently make available, but it’s something that we’re definitely looking at and would like to be able to think about doing at some point in the future. But it’s not something that we plan on doing for the 2016 election.

QUESTION: Hi. I’m Anna from Brazil. Folha de Sao Paolo.

I was wondering if you could track any other candidates, third party candidates.

DR D’ONOFRIO: Yes. I didn’t show them. I should have scrolled over. So we actually are tracking; so, we can see the conversation around Mike Pence, the VP candidate. But we do have both Gary Johnson and Jill Stein in here. So, with Gary Johnson, a little bit less than two million people over the last 30 days have been talking about him. But note the little increase here over the last couple of weeks. I think as he and Bill Weld announced their ticket. They started putting out some of their information. We see a little bit of an uptick here. Then, similarly, with Jill Stein, over the last couple of weeks, I think a growth in interest there. I expect this also to grow next week at the DNC. So, we definitely want to track anyone who is currently a candidate for the general election, we’re going to continue to track. And especially if Gary Johnson ends up hitting the threshold to be in the debates, we will enhance those efforts as well.

QUESTION: Hi. My name is Robert Poredos from Slovenian Press Agency.

Maybe it’s too early or not, but did you have any problems with [inaudible] of bias maybe from one or the other side?

DR D’ONOFRIO: The answer is ‘no’ with this research and I think that part of the reason why is because we’ve been developing this approach for a few years now and we’ve wanted to just let the data reflect the conversation. There’s no need for any sort of you know, subjectivity here. So, what we do is just build a framework that is key word-based, based on key words and hashtags that would be directly connected to the conversation. And then we let the system anonymously and in aggregate figure out what the numbers are.

To me, the litmus test is seeing how the volume shifts when new events happen. And it really captures the fact that Facebook is where people are talking about these things real time. And any bias that may exist, it’s simply just reflecting what’s out there in the world. We get this asked sometimes in terms of the sports contests. Like, look, I can’t control who won the Euros, I just reflect who was talking about Portugal or who was talking about, you know, France. So it’s the same kind of thing here. It’s as the election cycle evolves over the next few months we’re just going to continue to reflect that conversation.

QUESTION: Again, Hiro Aida from Japan.

I’m just wondering whether the Facebook users’ political orientation, do you think they’re rather liberal or conservative? The political orientation of the Facebook user.

DR D’ONOFRIO: Our data science team a few years ago did some research into this because there was a thought out there that, well, maybe Facebook just kind of was this echo chamber that reflected your own beliefs back to you. What we actually found is the conversation really crosses the spectrum of ideas, so it really is kind of a balanced conversation where you’ll see some things from friends you agree with, friends from people you disagree with. Now that’s not always the case, but that’s the typical user experience. And when we have looked at political ideology, it’s pretty much mirrored a lot of the other estimates that are out there in the U.S. in terms of party affiliation, liberal or conservative.

I think when you look at our penetration in the U.S. market, we have so many users even on a daily basis in the U.S. that it would be hard for that not to be reflective of the population at large at this point.

MODERATOR: Thank you so much for coming.

# # #