In this My Favorite Graphs episode, I revisit a 2016 New York Times visualization comparing the convention speeches of Donald Trump and Hillary Clinton. I explain why highlighting phrases across miniature pages can reveal themes in qualitative data that word counts and bar charts can miss. I share how I have used a simple version of this technique in my own work and discuss tools for creating similar views. Alicia contributes a voice note about the project's development, including the highlighted Google Docs behind the finished graphics. I also discuss how pairing an overview with example quotes can help readers see the broader context of an interview, focus group, or speech.
Resources
ggpage: Creates Page Layout Visualizations
Different Ways to Listen to the Show!
New Ways to Support the Show!
With more than 300 guests and 13 seasons of episodes, the PolicyViz Podcast is one of the longest-running data visualization podcasts around. You can support the show by downloading and listening, following the work of my guests, and sharing the show with your networks. If you’re interested in financially supporting the show, you can sign up for my Patreon platform, make a one-time payment via PayPal, or shop one of the show’s sponsors.
Transcript
00:00:07 Jon Schwabish
Welcome back to the PolicyViz podcast. I'm your host, Jon Schwabish. It's an intervening week here, folks. We've got another one of the "my favorite graphs" series here on the podcast, so this is the second time I'm doing it. I hope you find it interesting. These are just the graphs that I remember used to teach from, stick in my head. Not doing them in any particular order, just as I kind of feel like it and what comes up in top of mind.
00:00:33 Jon Schwabish
And for this week's episode, I'm going to talk about a 2016 piece from The New York Times: Wilson Andrews, Rebecca Lai, Irena Mikolajushin — I'm not sure I pronounced her name right — and Alicia Parla Piano. This piece came out in July of 2016. This was right after the nomination speeches by Donald Trump and Hillary Clinton at the respective Republican and Democratic conventions. This is a qualitative visualization, and there are not a lot of those out there. And this one sticks with me because, A, it's qualitative, which is kind of rare. B, it's really, really impressive work, even though it's, at its core, kind of like very straightforward. And, three, it is something that I have actually used in my work.
00:01:21 Jon Schwabish
Okay, so what is the piece? I'll put the link in the show notes, but the title is "Stronger Together and I Am Your Voice: How the Nominees' Convention Speeches Compare." And basically, there are a series of these visualizations. It's essentially, you can think of it as like a zoomed-out version of each speech. So Donald Trump's version on the left, Hillary Clinton's speech on the right. So you think of like a Word document zoomed all the way out. So you can see the thumbnail images of all the pages, and then phrases highlighted.
00:01:53 Jon Schwabish
In the first set of graphs, the phrases are highlighted in a pink color, showing what's wrong with the country. And phrases that are highlighted in green or shaded in green are what's good about the country. And what you can see right away in the first set of images is that Donald Trump's speech, nomination speech, has more phrases, more words about what's wrong with the country than Hillary Clinton's, which has a little bit more of a balanced approach, more green, I think, than pink. I think Trump only has like maybe three or four, two or three, two, three, four phrases in there that are what's good about the country, which probably doesn't really surprise most people who have listened to these two politicians.
00:02:43 Jon Schwabish
But what I love about this is that we take qualitative data and we show, or they show, the overarching trend, the overarching summary. And it's not like they're summing up the negative words, because how do you sum up negative words and then create a bar chart out of that, right? You'd have to sort of get into each, you know, a sentence can have mixed parts of it, right? So you have to get into the semantics of each sentence. And so I don't really see how showing a bar chart of the negative words versus the positive words actually works as well as this, where I can actually see the entire speech and see it fairly clearly shaded. And then, of course, because The Times has, you know, it's a web page, so it's, you know, virtually infinite scroll, they have a few of the quotes layered below. So you can actually sort of see how, you know, what they kind of identified in this case, in the first graphs of these two categories.
00:03:39 Jon Schwabish
And I've used this in my own approach and some of my own work. I've done it very simply, where I've literally done what I described, which is like a Word document. I highlight the words or the phrases or the sentences for the different categories that I'm looking at. So, you know, positive, negative, neutral, or yes, no, not sure. Highlighted those and then literally just zoomed all the way out and took a screenshot of it. And that was the easy way to do it.
00:04:08 Jon Schwabish
I know there's an R package called GG Page, which I've never used, that can do this. There's also the ability to do this on the Flourish data visualization tool, where you can put in your text and then set which words or phrases you want highlighted. So that's another way to sort of make these practically. But I just really like this because you get to see the patterns. And I think for, especially for qualitative data, we're not always looking for an exact count. I mean, that's sort of the richness of the qualitative data is that we're not always looking for a rich count of how many words or phrases are about, you know, this thing or that thing or the other thing. But we're looking for the overarching pattern, the overarching trend in the data.
00:04:49 Jon Schwabish
So I just want to also show that there were, or talk about that there were, I think there were two other sets of these they had. As you scroll down, another set, again, for Trump on the left part of the screen and Clinton on the right, highlighted in yellow, promises, vision, and policy proposals. So that's just like a series of yellow highlights. And then they did it also for the vice presidential candidates. So Mike Pence, Tim Kaine, where they had purple for praise for their running mates. And then they had, I think, at the towards the end here, I'm just scrolling through, criticizing each other. Again, that same approach. You can think of it as just a zoomed-out version and then just highlighted.
00:05:30 Jon Schwabish
And I've seen different versions of this over time. There was one, I believe, in the Washington Post for George Santos highlighting changes in his website as the sort of, this was a congressman from New York who was constantly lying about his credentials. So they sort of are highlighting all of the changes and when they occurred and for what reason. Again, I really like this approach. I think it's very clever for qualitative data visualization. But let's hear from Alicia, who sent in a little voice note about her recollections of this project.
00:06:11 Alicia
Okay, I don't remember a lot about this project. It was more than 10 years ago at this point, but I will do my best.
00:06:20 Alicia
Often for big speeches by presidents or candidates, we would visualize like most common words that they use to try to tease out the differences between like a Democrat or Republican speech or sort of summarize the issues the speaker was focusing on. But this approach sort of compares all these speeches at sort of a higher thematic level than just straight word counts.
00:06:49 Alicia
Looking at it now, I'm not actually remembering, but it's possible that we were struck by the negativity in Trump's speech that was sort of like atypical for a presidential candidate. He had spent a lot of time in his speech talking about crime and violence and sort of what was wrong with the country. And so I would imagine that we sort of use that as our frame or kind of like the starting point for this piece.
00:07:16 Alicia
And you can see that Hillary Clinton also pointed out some problems that the country was facing, but she didn't dwell on those for nearly as long. And she included a lot of examples about what was good about the country as well.
00:07:29 Alicia
I dug around a little bit to find the working documents for this project, and I did find a Google Doc where we basically highlighted the text of all of the speeches with colors that matched exactly what you see in the final piece. And I think KK Lai was the one who built the page, and she pulled in the categorization by sticking little prefixes in the Google Doc using ArchiML, which is the markup language that Archie, the New York Times graphic director, had recently created and we still use today.
00:08:03 Alicia
And then she rendered the text into these tiny images that sort of mimic the look of actual pages. So the final result is not really that different than what we were working with behind the scenes. And I think the result is really fun, and it sort of leaves an impression that something like bar charts breaking down the speeches would not have given.
00:08:23 Jon Schwabish
Thanks, Alicia. Appreciate you sending that in. Yeah, I mean, I, you know, I think she sort of like underplays personally. I think she underplays how valuable this approach is. I think this is one of the first times that I saw this approach of coloring phrases. And I think it's something that we can all do, right? It's pretty straightforward to do. And I hate to talk about AI in every episode, but if you think about how we can do this now, the AI tools are making this sort of thing even easier in terms of the analysis, right? To identify sentences, phrases, paragraphs that fall into different categories. It makes that even easier. And then to create this, you know, really doesn't take that much.
00:09:04 Jon Schwabish
So I think if you're a qualitative researcher or a qualitative data visualization communicator, I would definitely put this approach in your toolbox. Once you've analyzed the data to give people the full view of the transcript of the focus group or the interview of the speech so that they can see, oh yeah, okay, so this person talked about this thing a lot. So, and then give them maybe a couple of example quotes. So instead of just showing a single quote, which I think a lot of people worry about presenting because it's just that single quote, it's just that, you know, maybe the person cherry-picked it or maybe it's not really, you know, the entire, you know, perspective of the person speaking. You can sort of have both here.
00:09:50 Jon Schwabish
You can show how this person actually talked about, you know, this topic or used this sentiment or this phrasing across this entire, again, presentation, interview, focus group, speech, whatever it might be. So coloring phrases is one of the techniques that I throw in my qualitative data vis toolkit. And I think this is just a great example of it. So thanks to Alicia, thanks for The Times, and I hope you found this episode of My Favorite Graphs useful. Okay, that's all I've got. Thanks for tuning into the PolicyViz podcast, and I will talk to you next week.