It’s been a long road, but we’re finally here: The first version of the ideology.media mapping project is ready. I know no one cares but me, but: Hooray.
In all seriousness, I’m glad to share this first map. Without further adieu, let’s dive into the first version of ideology.media’s Mapping Project!
I’m reposting the ‘working definitions’ that I used to rate articles below. Later, I’ll share some thoughts about how these definitions worked out. For now, however, this was the system that was in use for our first map.
For more information about how these ratings look in action, please check out this linked blog post.
But what does that mean? To make it a little easier to understand, I’ve included the exact same map below with a slider that lets you compare it to a quick-and-dirty positional map of prominent modern American politicians.
What Do They Mean?
Broadly speaking, the average of the fifteen media outlets I examined, which ranged from mainstream news sources like the Associated Press and the New York Times to smaller fringe sites like The Palmer Report and InfoWars, is slightly conservative and slightly elitist. There were also large swaths of the ideological map that were completely unrepresented by the chosen Western media outlets.
The outlet that came closest to the statistical average of all these outlets was The New York Times, though they were slightly more progressive and slightly more populist than this average. Keep in mind, however, that this average was weighted, as the conservative outlets were much further to the right of the map than the progressive ones were to the left, and there were more of them.
Part of the reason for this is that the progressive outlets during this time period were largely (with a single exception) more about Donald Trump than they were about progressive politics. For instance, compare these two maps, on the bias and frequency of coverage on Donald Trump compared to Government Aid (such as unemployment benefits).
A Quick Note on Methodology
Every outlet chosen went through the same process to select articles, though there was some growing pains early on. To break it down:
- See if they have a print publication. If so, go with the top 5 articles on the front page.
- If they don’t have a print publication but they do have a ‘most popular’ or ‘most read’ metric, the five most popular articles from that date.
- If they don’t have a front page OR an easy sorting metric, then just 5 random articles from that day.
This means I read 55 articles from each source, breaking them down into keywords on subject and type of media and rating each article. Once I had all 55 articles rated, I found the average rating for that site during that time period. That is the number on the map above.
What Comes Next?
In the coming weeks and months, I will be analyzing the data from this first round and creating maps that give a slightly different look at the information. For instance, how does a single outlet cover different topics? Do the biases of the New York Times differ when covering Trump compared to when covering, say, the Portland protests? And what kind of content do these sites put out?
By the end, each of the outlets covered this way will have a distinct page you can visit to get basic information about their biases and coverage habits.
I’ll also be using my analysis of the media to answer some common questions. Does the media have a liberal bias? What does that mean? Is there any unbiased news source? And is that even something you should want? Those will be available in the site’s blog section.
I want to end by acknowledging the flaws in the project. I had a small number of test readers who used these metrics to rate articles that I had rated. I hid the outlet from them, so they wouldn’t bring potential biases that were unavoidable for me.
What I found was that, in many cases, my test readers often agreed with me on the Elitist/Populist scale, while their ratings on the Progressive/Conservative scale were more random. This suggests to me that the E/P scale is more coherent, its rules more clarified, while the P/C scale is more subjective.
There is always going to be a degree of subjectivity in a project like this; I do not want to fool people into thinking that this or any media venture is unbiased. What I do want, however, is for my scales to be consistent. Right now, one of them is more unbalanced than the other.
I also want to do more work on my keyword system. Right now, it feels scattershot. A more robust system would let me build better accompanying maps.
I want to thank Cee Harker, Scott Nicolson (@scottlnicolson on Twitter), L. Woods, and others who gave me feedback on the rating system, who helped check articles, and who talked to me about how visually ugly my earlier versions of this were.
You tried your best. I’m so sorry it turned out like this.