Chicago Booth Review Podcast Why Is Some News More Newsworthy?

Hal Weitzman: What makes some stories newsworthy and others, well, a bit boring? Can you measure newsworthiness and are people writing thinking that the media tend to focus on bad news? Welcome to the Chicago Booth Review Podcast, where we bring you groundbreaking academic research in a clear and straightforward way. I'm Hal Weitzman, and today I'm talking with Chicago Booth's, Emir Kamenica about his research on media bias. Kamenica argues that accounting for newsworthiness changes how we view the media's apparent focus on negative news. How do you measure newsworthiness and is the news not always as negative as it may first appear? Emir Kamenica, welcome to the Chicago Booth Review Podcast.

Emir Kamenica: Thank you for having me.

Hal Weitzman: Now, Emir, we've got you here because you researched the news and having worked in the news industry all my life, I'm puzzled by the question that you tried to answer, which is what makes news newsworthy and other news, I guess, boring or less newsworthy? First I'm interested in how you came to this research topic in the first place.

Emir Kamenica: Well, what I spend most of my time thinking about on a daily basis is information in general. Most of my work is in information economics and obviously one very important aspect of news is that it provides information. There are obviously other reasons to read the news. Sometimes you read it for non-instrumental reasons, but this particular project starts with the view that one of the primary goals that people have when they tune into a news channel of whatever format is to find out something they do not know prior to turning the news on. If one takes that perspective, we can build on the machinery of information economics and turn a question of whether a particular piece of news is newsworthy into a formal one, into one that can be quantified through the lens of rational decision-making perspective. The general idea is that on any given day you have some belief about some variable.

Maybe it's you have a belief about what's going to happen to the stock market today. Most days the average daily return of the stock market is obviously positive, but very close to zero. You have some belief and then something happens. The stock market goes up, it goes down, goes up by little, goes down by a lot, something happened. The question is, we could ask, given your belief prior to learning the news, how surprised would you be? To what extent would your belief change if you actually learned what actually happened? And sticking with this example, this doesn't have to be the same day by day. There are, for example, periods where volatility is greater. What you should expect rationally in certain periods of time is that you still think the average return is going to be close to zero, but you won't be as surprised if it's large relative to some other period. Even a single seemingly identical piece of news, the market moved up 1.2% can be differentially newsworthy on days when volatility is low versus when volatility is high.

Hal Weitzman: The next question I was going to ask you is what makes news newsworthy? And you've started to answer it there. Basically, there's an element of surprise. If the market then is going up 1% a day for a consistent period of time, we'd all be very wealthy, which would be great.

Emir Kamenica: Yes.

Hal Weitzman: But anyway, let's imagine that that were to happen and one day it went up only half a percent, it would be less newsworthy then?

Emir Kamenica: No, quite the contrary. Suppose the market would be more if it moves.

Hal Weitzman: It changes the pattern.

Emir Kamenica: Yes. Precisely.

Hal Weitzman: Let's imagine, if it goes up 1% one day and then for five days in a row and then the following day it still goes up 1%, that's not newsworthy?

Emir Kamenica: Correct. And this is related. There's been a large literature on what might be termed the sociology of journalism, which does a more anecdotal approach, more case study approach to looking at what get covered by the news. And this literature and sociology of journalism has introduced terms such as the fatigue bias. Fatigue bias, according to this literature, is that something which used to be newsworthy stops being newsworthy because people quote unquote, as the term fatigue suggests, get tired of that particular piece of news. But there's a very different perspective, very different interpretation of the same phenomenon if one takes an information theoretic perspective to it, something used to be news because it was uncommon in something we didn't expect.

But as it happens day after day after day, it's not that we get tired of it, is it genuinely becomes less newsworthy. Something that used to surprise us now no longer should surprise because it has become the new norm. Something like a fatigue bias is not something one would think of as a bias coming from the kind of perspective that we're taking. It's just that a particular piece of news is less newsworthy once it is what people have come to expect. If it's something I would believe in expectation, even if I didn't turn the news on, well then it's not particularly newsworthy.

Hal Weitzman: But how do you think about interest? We care about the stock market partly because we have the skin in the game. Maybe we don't have skin in the game in a war that's happening in Africa because it really doesn't affect us very much. And I always think of when I was in the days when I was a daily reporter, I worked in South America. To a British readership, South America is not that exciting. The bar has to be much higher for it to be newsworthy versus something that happens in the UK.

Emir Kamenica: Absolutely, and there's certainly nothing irrational about people having more interest in one topic versus another. But if you wanted to measure from supposing that taking as given, which may not be correct, that the news media reports that which people would want to hear if it's a benevolent model of the news media, which is the one we have in mind in our empirical analysis-

Hal Weitzman: What do you mean benevolent?

Emir Kamenica: By benevolent meaning, take a perspective that the news media are trying to do what is best for their viewership. That's what I mean by benevolent.

Hal Weitzman: If they're trying to tell the truth, not manipulate.

Emir Kamenica: Well, they're going to tell the truth and moreover, when they report a number, it'll be correct number. But it's more than that. We're largely focusing on settings where what the truth is is unambiguous. The market went up as much as it did. The decision we study is not, what are you going to say? Well, if you talk about the stock market, you will tell the truth. The real question is do you want to talk about the stock market today? Do you want to talk about the weather today? Do you want to talk about the unemployment news today? Do you want to talk about deaths of US soldiers in Afghanistan today? Those are the questions we examined. To return to your comparison about countries, we for example look at the news media reporting on deaths of US servicemen in Iraq and deaths of US servicemen in Afghanistan.

You could take a particular seemingly identical piece of news like three soldiers were killed today. How newsworthy is that? It obviously depends with what's happening at that point as of that morning. But when you don't know what's going to happen that day, there are days when three soldiers being killed is absolutely dramatic, unexpected news. And there are days during certain periods of war where sad though it may be you are not that shocked to learn that three soldiers might've been killed. We, for example, compare seeming presence of bias of the kind that you described, by comparing Iraq and Afghanistan. And if you just look at the data naively, you might think that there is actually a difference in the likelihood that the media report on deaths in one country versus another. But if you look at the data more carefully through the lens of a model, it takes into account what the expectations were every morning and then whether what was reported in the evening that seeming difference in the way Iraq and Afghanistan are treated, disappears. What might look like to a naive observer-

Hal Weitzman: Just to clarify, these are horrible things to talk about, but just in theory, imagine that people are killed consistently and a consistent number of people are killed every day on a day when no one loses their life, that's more newsworthy than people being killed.

Emir Kamenica: Absolutely. Suppose-

Hal Weitzman: It's interesting because you think of the stock market, for example. If nothing happened, if there was no movement at all in prices, you would think that objectively, that doesn't sound that interesting, but it might be very interesting if this time of volatility.

Emir Kamenica: Absolutely. And honestly, you gave your example as a macabre one, but it's not suppose, which of course is not going to happen, but suppose that today nobody dies in the US, that would be an absolutely dramatic, unexpected, wild, crazy thing that if a news agency learned that zero people died today, you'd think that would be probably one of the biggest front-line news. We have no idea how this happened. It's completely implausible. No one dying of course can be a bigger piece of news.

Hal Weitzman: It sounds like the premise of a movie. We should write that one.

Emir Kamenica: That does actually sound like a pretty good premise.

Hal Weitzman: Tell us how you measure this. You're obviously an economist, you're using measurement. How do you measure this newsworthiness then?

Emir Kamenica: We start from a perspective that there is a decision-maker that is in the audience and has some belief about an unknown state of the world. And then we look at various examples of what the state of the world would be. One application, the state of the world is the movement of the stock market. Another application, it's the unemployment rate that gets released by Bureau of Labor of Statistics. In another application, it's the extent of precipitation. In another one, it's death of US soldiers in Iraq and Afghanistan. You start with some prior, and what this prior is you can learn various ways, you can motivate it theoretically. For instance, if you're interested in the prior for the stock market movement, you can actually learn something about volatility from price of certain financial instruments. Or you could just take a purely statistical approach and saying, "Look, given the pattern in past data, what would be a reasonable thing to expect will happen today?"

You start with some prior and then, this is intentionally abstract, some utility function that we are agnostic about as measuring newsworthiness. That depends on you adjusting your action to the true state of the world, which you do not know unless you tune into the nightly news and learn it. From the news media, there's a trade-off because you cannot report everything. That's simply infeasible. Selective benevolent reporting is central to our framework. The news media takes as given what your prior is, they observe the truth and then they have to choose whether to report that or report on something else. There's some outside option of reporting on a particular piece of news. If you take this perspective, you can prove a theorem that reporting data... And what is reporting data? Reporting data takes as input for every day and every say nightly news... What the reporting data looks like is, "On this day, here's what a reasonable prior would've been. Here's what actually happened, and here is whether the thing that happened was reported or not."

It turns out that the model I described of benevolent selective reporting to a rational audience puts very strong restrictions on what this reporting data might look like. And one of the contributions of this paper is to derive the mathematical properties that the reporting data set has to satisfy. And we could get into the weeds of that if you'd like. There's a neat connection to a seemingly unrelated topic on something called scoring rules and statistics, which have largely been used to think about how do you elicit someone's private beliefs. Here we're not eliciting anybody's private beliefs, but it turns out mathematically the kind of functions called scoring rules, which are often used to elicit beliefs, are also the kind of functions that the rational benevolent reporting requires discipline, the kind of reporting data that you see.

Hal Weitzman: You could score a news story and say, "This is how newsworthy it is?"

Emir Kamenica: Yes. There's a bit of a choice of which scoring rule you use depending on what you think the underlying function is. We use a particular scoring rule that is an off-the-shelf one. And once we do that, then on any given day for any given realization, we can assign a number of how newsworthy that finding was. And then our model predicts that the more newsworthy a particular realization is, the more likely it's to be reported. And then we actually look at the data and it turns out that the model matches the data extremely well. What we, using the formula, would say was a newsworthy event is in fact highly correlated whether that event gets talked about in the evening news.

Hal Weitzman: If you're enjoying this podcast, there's another University of Chicago podcast network show that you should check out. It's called The Pie. Economists are always talking about the pie, how it grows and shrinks, how it's sliced, and who gets the biggest share. Join host Tess Figland as she talks with leading economists about their cutting-edge research and key events of the day. Hear how the economic pie is at the heart of issues like the aftermath of a global pandemic, jobs, energy policy and much more. Okay, ready? Emir, in the first half we talked about what makes something newsworthy? And you talked about your scoring system and you can actually put a score on any... Do you ever watch the news and say, "Oh, that's a 73, that's only a 45?"

Emir Kamenica: No, I don't do that.

Hal Weitzman: You don't? It might be a fun game to play. Many people have the impression, and this is a common criticism of the news, that they only care about negative stuff, they never report anything good. They only report stuff that's bad. Obviously in the stock market, which we talked about in the first half, that isn't the case. The stock market pretty much gets reported every day by certain outlets or ignored and only reported when it's really big.

Emir Kamenica: But that's not quite. If in fact an outlet always reported in a stock market or never did, then our theory would be useless. We in fact rely very much on the fact that if you pick a typical nightly news, they will talk about the stock market on some, but not other. That's central to us. If they always did it or never did it, we don't have a theory at all. We don't have anything to explain. In fact, you could ask a question about negativity. Is it the case that a negative stock market movement that is just as big as a positive one and big, you have to be a little careful because the distribution of returns is not exactly symmetric, but equally newsworthy, are they equally likely to be talked about?

Hal Weitzman: As big as a positive.

Emir Kamenica: And in the context of the stock market, the answer is absolutely yes. There is no detected, neither apparent nor real negativity bias in the decision whether to talk about the stock market. A positive movement is just as likely to be talked about as an equally newsworthy negative movement. By contrast, we look at whether unemployment news is reported, and if you look at that data naively, you might detect something that looks like negativity bias. You might think that when there is a weekly negative movement, meaning the unemployment rate goes up, that's negative news, unemployment rate has gotten worse.

That does get talked about more. It's more likely to be reported than when the news is positive, the unemployment rate has come down. If you just did a simple... Let's look at the negative news and a positive news, it looks like negative news gets talked about more. If, however, you take our perspective and think about the genuine newsworthiness of these movements in the unemployment rate, you see that a large, if not the entire reason for the apparent discrepancy is that unemployment does not move in equally large steps when it goes up and when it goes down.

If I tell you that the unemployment rate has moved by a large amount, but I'm not telling you whether with up or down and you are familiar with the data, you should take this to be negative news. It turns out if there is a large movement in the unemployment rate, the unemployment rate is much more likely to have gone up than gone down. The way the unemployment rate typically works is it jumps up and then slowly trickles down. Now those jumps are more newsworthy than the trickling, which is why if you actually just count the number of times that the negative news reported, you'd find something that appears to be a negativity bias. But it's not really negativity bias, it's just that it turns out in the context of unemployment rates, if I tell you there's big news, you should expect it to be negative.

Hal Weitzman: And how does that apply to when you were tracking wars then? US involvement in foreign wars. If allied troops make progress, is that as newsworthy as people being killed?

Emir Kamenica: Well, in that specific context of casualties, we are constrained of what we can measure. We really want to know what the true state of the world is and whether it got reported. We don't have the true state of the world being like, "Oh, there was progress made during a particular day."

Hal Weitzman: There you're really only capturing bad news.

Emir Kamenica: Well, bad news in a sense, but it's all relative to the expectation. Yes, it's always bad news that someone die, but you could imagine that, as we discussed earlier, there are periods where you, in fact, that few people die, is good news because it's fewer than you might expect given the circumstances.

Hal Weitzman: That's fair enough. In general, if I say to you, does negativity bias exist? Your answer is maybe, but it's not as strong as the impression that we have.

Emir Kamenica: Because it gets talked about so much, we certainly looked for it and we did not find it in the context where we looked. Does that mean there's never any negativity bias anywhere? No, but it does give us pause and make us think that before claiming negativity bias, we want to give one a framework for what it would take to establish it. Doing something very simple and seemingly straightforward, like counting the number of stories that are negative versus positive is simply not the right way to go about answering the question, is there negativity bias?

Hal Weitzman: Because there's this other dimension which is, is it really newsworthy?

Emir Kamenica: Correct.

Hal Weitzman: I know you're not going to like this question, but I'm going to ask you anyway, which is about what's happened since your sample ended, which is in 2013, over a decade ago, and of course, the world's changed so fast. The number of people who watch TV news, which is what you were measuring, has dropped dramatically. I think there's a significant number of proportion of population that never watches TV news. They get their news from social media, there's no objectivity, there's no benign, what did you call it? Benign news editor, whatever, choosing what's newsworthy. There's no balance, there's no fact checking, there's no accuracy. Nobody cares. It's the whole point of social media is to rile people up and get them energized.

Emir Kamenica: You made a lot of claims there, which I think could be explored by collecting the data on what exactly is being reported.

Hal Weitzman: Fair enough.

Emir Kamenica: And we don't find in any way a shortcoming of the paper that we don't analyze all of those other news outlets or all the possible periods of history, including more recent and more distant in the past. What we really want to do this paper is provide a framework for how a researcher who wants to ask the kind of questions you're asking, like for instance, is there more negativity bias or is there more coverage of one topic versus another? What is the right way to go about answering that question? What kind of data do you need and what are the ways in which you want to analyze that data? We really think of our paper as providing a tool that can be applied to any period of time in any particular form of news. I think what you're asking here is you're advertising asking people to work on those papers, and I'd love it if they would do it.

Hal Weitzman: I'm just wondering if the algorithm now takes the place of the editor where the algorithm is choosing what you see.

Emir Kamenica: But the algorithm may or may not be well approximated by what the readership wants, and the readership very well may want to be told only things that they did not already expect. It is not clear if I were designing an algorithm that I would want to do something other than report on a piece of news when it is newsworthy.

Hal Weitzman: Well, we'll leave that as open for other researchers then to come and tell us what they're finding out. But there is an extension of your work here because maybe the algorithm is the editor, maybe it's not, but the user, the reader, the viewer is still choosing, making choices about what they are going to look at. And there they are time constrained, they're having to decide what is more interesting to them.

Emir Kamenica: Absolutely. And my co-authors on this paper are some of the world experts on precisely depth dimension. In their previous work, Matthew Genska and Jesse Shapiro have done absolutely foundational work studying the decision by the readership, which news source to consume. This is certainly an important topic and one that has been looked at.

Hal Weitzman: I'm just wondering, do you think we can conclude from your research that for those who are still watching TV news at least it's less biased than we used to think or than most people think?

Emir Kamenica: Well, I don't know what most people think.

Hal Weitzman: Again, we need to take you on it.

Emir Kamenica: I do certainly think that one gets a sense that there's a commonly held view that may not be as correct as one might think once you actually look at the data through the correct lens.

Hal Weitzman: And in that sense, I guess it relates to some of your other work, which is to do with just the sense that we have a feeling that, for example, and work of Genska and Shapiro as well, partisanship is more extreme than ever and perhaps when you look at the data, it's not. There's a parallel there I'm just thinking with your other work.

Emir Kamenica: Well, absolutely. I think what research is quite often about is looking at important topics where a lot of people think that something important is true, but establishing whether it is, requires looking at the data carefully and not just saying that it is obvious that something holds or not. And that's certainly what a lot of empirical research in economics and related fields is about.

Hal Weitzman: Well, Emir, this was certainly newsworthy to me, so I thank you very much for coming and talking to us about your research on the Chicago Booth Review Podcast.

Emir Kamenica: Thank you.

Hal Weitzman: That's it for this episode of the Chicago Booth Review Podcast, part of the University of Chicago Podcast Network. For more research, analysis, and insights, visit our website at chicagobooth.edu/review. When you're there, sign up for our weekly newsletter so you never miss the latest in business-focused academic research.

This episode was produced by Josh Stunkel. If you enjoyed it, please subscribe, and please do leave us a five-star review. Until next time, I'm Hal Weitzman. Thanks for listening.