If you’ve ever thought that the quest for more clicks is affecting the sorts of articles that get published in the media, Wharton marketing professor Pinar Yildirim wants you to know that you’re right. But it’s not quite the overarching impact that you might expect.
In this interview with Knowledge@Wharton, she talks about a new working paper funded by the Mack Institute, “Clicks and Editorial Decisions: How Does Popularity Shape Online News Coverage?” The paper, which was co-authored by Ananya Sen, a doctoral candidate in economics at the Toulouse School of Economics in France, teases out the differences in how high-traffic stories get treated in terms of longer-term coverage.
An edited transcript of the conversation appears below.
Searching for a Bias toward Profit
In this study, what we are interested in is documenting any sort of bias that might be happening in newspapers related to the availability of digital data and big data. Consumers often wonder what goes through an editor’s mind when they look at newspapers: How do news stories make it to the newspaper that they read or the TV that they watch? Are they selected based on some sort of particular agenda of the newspaper, or are they sorted based on their importance?
Now, if you ask a journalist, up until recently, at least, they would give you the answer that, “Yes, news stories are selected based on their newsworthiness, their importance, and the novelty — how important they are, how new they are, to the public.” However, if this is the case, consumers may also wonder, “Why is it that we see so many cat pictures?” or “Why is it that we see so many quizzes on the news media?”
“When we looked at whether traffic that’s received by articles influenced the coverage of stories, we found that the effect is present only for the hard news.”
We had an alternate hypothesis: Perhaps there are other concerns in an editor’s mind, especially these days, when the editors are fed so much information about how the individual stories on online news websites are doing.
In particular, we were thinking that there might be a relationship between the traffic that individual articles receive and how long stories and particular topics are covered in a newspaper Twitter . We tested this hypothesis using data from a large online version of a reputable Indian newspaper. We wanted to see if those stories whose first articles received a high number of clicks for various reasons were somehow covered for a longer period of time in the newspaper. We also tested various additional measures of resources that an editor might allocate to the story.
What we found was interesting. We indeed found that there is a relationship between the number of clicks that are received by the article and the amount of time that a story is essentially covered by follow-up articles that are repeating on multiple days. Those two are positively correlated.
There are a couple key takeaways from our research. First, there is indeed a relationship that whenever a news article receives a higher number of clicks, there is a higher level of editorial resources allocated to the particular story. This is something that makes sense at a correlational level. But you would wonder, “To what extent is this relationship also causal?” What I mean by that is, to what extent is it the case that, just because a news story is receiving a higher number of clicks, you see a longer period of time that is allocated to it?
Establishing this was the main effort that we exerted in our research. What we did was to find some instruments that would affect the amount of clicks that are received by the article, but not necessarily the editorial decision of how much time to allocate to it. And the two instruments that we used were … days with rain, and … ordinary power shortages. Our idea is that, if there is a rainy day, people would be staying indoors, so they would be more likely to go online. That is going to create some sort of a shock in the number of clicks.
Similarly, on the days when there are ordinary power shortages, it’s likely that the chance that people will go online is going to go down. That’s going to reduce the number of clicks. When we look at the relationship between clicks and these two instruments, we find a significant relationship. And whenever we looked at the editorial policy related to these two instruments, we are also seeing that there is a relationship in terms of how long the coverage of a particular story is, or how many follow-up articles appear that relate to a particular first story that was published. What would that mean? That might mean that, if it happens to be a rainy day, or if it happens to be that, in India, if there was an ordinary power shortage on the day that a Kim Kardashian article first broke, that story actually may run for a couple more days in the newspaper, as opposed to another story.
“What we are telling the news editors is, in some sense, ‘Yes, it might help to use digital data, as long as you’re not using this in order to dilute the quality of information that the consumers are receiving.’”
The conclusion that surprised us the most was probably what we found once we started separating news into further classifications and made a comparison among them. What we did was analyze news as hard versus soft news. What I mean by “hard news” is the type of news that tends to provide information to the consumers — for example, news related to politics or business. And by “soft news,” I mean news such as entertainment or sports. When we looked at whether traffic that is received by articles influenced the coverage of stories, we found that the effect is present only for the hard news; it doesn’t exist for the soft news. Or to put it in more academic terms, the effect is not significant.
What this might imply is that consumers should not necessarily worry that, just because traffic information is provided to news editors, and news editors tend to pay more attention to it, it doesn’t necessarily mean they’ll start seeing newspapers full of cat pictures or more entertainment stories. In addition, we also quantified the effects of the traffic. So what we can say is, for example, increasing news story clicks by about 700 clicks is going to result in about three additional days of coverage on the story. And that might also mean three additional articles on average about … that same topic.
The Right Way to Deal with Traffic Data
In the news industry, right now, there is a lack of consensus in terms of what newspapers should be doing with all of this digital data. When we look at some newspapers, they tend to recommend their journalists to stay away from digital data. This is done in order to encourage them to come up with better stories, newsworthy stories, stories that matter to society. When we look at other examples — Gawker.com is mentioned to be an example of this — these media outlets tend to provide some sort of an incentive for their journalists to create stories that can bring a higher number of clicks. Usually these incentives are in the form of financial payments.
It’s not clear exactly what policy the news industry should embrace. Our study, I think, shows what can happen, or at least shows what’s happening currently in one large, reputable portion of the news industry. What we are telling the news editors is, in some sense, “Yes, it might help to use digital data, as long as you’re not using this in order to dilute the quality of information that the consumers are receiving.” And for the consumers, I think we are telling them that they don’t have to necessarily worry about the quality of information that they receive from newspapers. Digital data might still actually help them. It can increase the number of stories that are informative, in some sense, or the coverage of the stories that are informative.
“The other side of, essentially, letting clicks drive your news is, what happens to representations of minority views?”
For policymakers, this brings up other questions. Because the other side of, essentially, letting clicks drive your news is, what happens to representations of minority views? When the majority of clicks are coming from the populations or the sub-groups that are also in larger portions in the society, it might bring up some questions about viewpoint diversity, which is important. And they might need to start asking some questions about what this policy does to the viewpoint of some minority groups….
Treading New Ground
Technically, our research is the first that’s documenting the existence of bias that’s coming from traffic data. So in that sense, we can’t really make a comparison to previous research. But we can say that we are providing an evidence of a bias that is in a different format. And that is essentially very new, something that we are just starting to see in newspapers.
While we are still conducting and carrying out further analysis to see if we can make additional statements about this type of behavior, we are getting lots of feedback from scholars, and also peers in the industry. As far as following up our research, with the same research co-author, I’m going to be following now on social media, not just the traditional news media. We are going to look at other impacts, other factors, that might be influencing people’s lives. One particular study we are focusing on that is relevant, but a very different topic, is looking at politicians’ entries on social media and how that influences their chances of raising campaign donations.
This post originally appeared on the Knowledge@Wharton website.