Does The Media Really Affect Elections?

The 2020 election will soon be upon us. As usual, news outlets will play a crucial role informing the public about the candidates. But could their decisions actual swing elections?

That’s the argument put forward by Prof. Gregory Martin from Stanford University in a recent paper. The data he’s collected shows that the decisions made by reporters and editors may have surprising effects on who voters support.

Listen on Apple Podcasts or wherever you enjoy podcasts.

Transcript:

William Howell:

I'm Will Howell.

Wioletta Dziuda:

I'm Wioletta Dziuda.

Anthony Fowler:

I'm Anthony Fowler, and this is Not Another Politics Podcast.

William Howell:

So the economy is in free-fall.

Tape:

All right. The breaking news, the new unemployment numbers just in. Weekly jobless claims, they paint a full grim picture of the last two months.

William Howell:

And then, it's a lot of talk about what impact the changes in the economy is going to have on Trump's chances of holding office and how the media is covering the economy, whether or not he can really find refuge in Fox News or whether, or not the media will facilitate his political resurrection in the face of a declining economy, or whether or not they're going to make life even more difficult for him.

Wioletta Dziuda:

So, one question you could ask is, if I'm listening to Fox News and they decide not to focus too much on unemployment and they decide to talk instead about Obamagate, or something similar, is this going to affect how voters actually respond to unemployment in November?

Anthony Fowler:

This is of course a really difficult question to answer, right? It's really hard to know, how much does media coverage or media emphasis affect elections? Because suppose this mayor got lots of positive news coverage, and this mayor got lots of negative news coverage, and let's see how that corresponds with their electoral fortunes, you'd find a huge correlation, but you'd have no way of knowing how much of it was the media and how much of it was because, well, maybe the media was positively covering the mayor that was actually doing a good job. And so, you can see why it's very difficult to answer this question empirically, but we'd really like to know if the media changed the way they were covering things, or they changed their priorities, would it actually affect the reelection chances of politicians?

William Howell:

So I think you talked to somebody who has written a paper that we're going to talk about today that developed a strategy for distinguishing these two kinds of effects.

Anthony Fowler:

I did. I talked to Greg Martin at Stanford who wrote a paper about this exact question, and he takes advantage of media coverage, newspaper coverage, of unemployment statistics to try to get a sense of what the effect of, conditional on the state of the world, what's the effect of more or less positive news coverage about a governor on a governor's reelection chances?

Greg Martin:

So I think the underlying idea of the paper, the thing that we're trying to study is the idea of agenda setting by the media. So, this is the idea, if you talk to reporters, what they'll tell you is the media doesn't tell people what to think, but it can tell people what to think about, right? So the choice of what to cover shapes the debate and you can imagine how that's influential for election outcomes.

And then yeah, if it happens to be the case that media coverage focuses on one issue where one candidate is favored, that's going to be good for that candidate compared to the others. Of course, the problem with actually measuring what the influence of the media is, is that there's not some cabal of editors that are sitting together in a room and deciding what to cover. They're reacting to what they think the public wants.

And so, it's hard to separate changes in coverage from changes at underlying demand for those topics. So what we did, and the way that we separate those two things, is we're focusing on the coverage of the economy. So, coverage of unemployment news and the thing that separates changes of underlying conditions from changes in coverage, is the occurrence of what we call milestone events, right?

So, this is when the unemployment rate crosses some round number, right? That tends to produce a discontinuous change in the amount of coverage of unemployment. So, you can imagine two different states, imagine initially the unemployment rate is five and a half in one state, and it's 5.6 in another state, both states, the unemployment rate goes up by 0.4. In one state, you hit 6% and the other, you just miss it you, you stay at 5.9.

In the state that hits six, you get more coverage of unemployment than you would in the state that's 5.9. So even though those states are really similar, both in the amount of unemployment and the amount of change, one gets more coverage of the economy than the other.

Anthony Fowler:

Okay. This is really interesting. So the Bureau of Labor Statistics is putting out these regular monthly reports, is that right? They're putting out monthly reports on state unemployment.

Greg Martin:

Yeah. So every month the BLS releases a report with estimates and that's an important part of the paper, is that these are their estimates. They're rounded to the nearest decimal place, but yeah, every month they release, for every state, this is our estimate of the fraction of people who are looking for work, who can't find it.

Anthony Fowler:

And so, presumably there are lots of reporters who cover state politics, who they look at that report every month. If your story is right, the reporter believes that it's a more newsworthy story, if unemployment just hit 6%, as opposed to 5.9%, that's a more newsworthy story. And so, they're more likely to go write an article about, "Hey guys, unemployment is on the rise, look, we're at 6%. And we haven't been at 6% in a long time."

Greg Martin:

Yeah, I think that's right. So that is the basic story is that crossing one of these round numbers changes the reporter's perception of the newsworthiness of the story.

Anthony Fowler:

So maybe, do you want to just give us the punchline first? So what did you actually find? So you're going to use this to then confirm that indeed news coverage about unemployment is greater when one of these round numbers is hit, and then you're going to use that to estimate the effect of a good or a bad, a positive or negative, news story on gubernatorial elections.

Greg Martin:

Yeah, that's right. So there's basically two results. One is coverage of the economy increases discontinuously around these round numbers. So it's substantially more coverage, it's 10% or so more coverage of the economy when we just cross that round number, compared to just missing it.

And that translates to, in gubernatorial elections, in the release prior to the election, we happened to hit a milestone and you get changes in governor vote share by fairly large amounts. So if you get a bad milestone, which is unemployment is rising and just crosses a threshold number, you get something like a 10 point decrease in incumbent vote share.

If you had a good milestone, which is unemployment is falling and just crosses a round number, you get at something like a five-point increase in the gubernatorial share, relative to a similar case where unemployment fell by a similar amount, but did not cross a round number.

Anthony Fowler:

So obviously then, these new signals matter, but the bad news signal seems to hurt the governor more than a good new signal helps.

Greg Martin:

Yeah. There are two different mechanisms that you could think of for that that would cause that. One is, it's just about the response of the media. And so, we do find that the media response is also asymmetric. So that's a well-known, I think that people have studied before is negativity bias in the media, that media do tend to favor stories about things getting worse compared to things getting better.

Again, this has to do with perceptions about newsworthiness and what people want. So I think that's part of it, is just that there's more coverage. And then, it's also possible that there's just some asymmetry in voters' reactions to information directly. And we don't have a good theory for this, but there are psychological stories about loss aversion, where people react more strongly to the negative changes relative to a reference point than to positive changes relative to that same point.

Anthony Fowler:

Even without invoking a behavioral story, I could imagine you could tell a somewhat rational story, which is that the governors are positively selected for themselves. So they won the last election, so most of the voters like them. And so, in the absence of any new information, most people are probably going to reelect. And then, even if every, each voter individually, their beliefs are responding in a somewhat symmetrical way, you're going to have a lot more voters who like the governor who fall away then the other way around.

Greg Martin:

Yeah. Sure. That's definitely possible. I think that's, for incumbents, no news is good news, in some sense. Yeah.

Anthony Fowler:

Right, right. That's interesting. Do you think your results have anything to say about the health of democracy? I know this is a topic that lots of people have strong opinions about now. They say that the voters are too hyper-partisan, they say that they're not paying attention. They're uninformed. Is it somewhat reassuring that voters are responding in reasonable ways to the information they get in their local newspaper about whether the governor's doing a good job or not?

Greg Martin:

Yeah, no, I think it is reassuring in some sense in that there really is a margin of voters who are swing voters and who are willing to change their votes in response to information. You do sometimes hear people say things like, "We're just a bunch of tribal partisans, and everyone's just voting for their party, and information can't change anyone's minds."

I think this paper does show that these external swings in information and actually can move people's votes. And I think that is a positive finding for the health of democracy. Yeah.

Anthony Fowler:

Is there anything else that you want to say to our listeners or something, a big takeaway that you want them to draw from the study?

Greg Martin:

I think the big takeaway is just, this is providing evidence that the things that the media cover are influential. So it is the case that the choice of the set of information that people have access to when they vote does matter independently. And that, when we observed media effects, it's not purely based, just reflecting underlying tastes of the population. There is some independent influence. And so, that's something we should think about.

Wioletta Dziuda:

So basically, what they find is that whenever unemployment crosses a certain milestone, let's say 5% - 6% before a gubernatorial election, then media cover unemployment much more. So that's already very interesting. And the second finding that they have is that, in those states where the milestone was reached, people actually penalize or reward the governor at the election time. And the numbers that they are getting at actually to pretty large. So for example, what they find is that, if unemployment goes up to, let's say 5% in a state, the governor receives 10 percentage points fewer votes, than a governor in a state where the unemployment just went up to 4.9.

And, if the surprise is positive, if actually the milestone's positive, the unemployment drops to 3% let's say, then there's also a larger effect, but it's much smaller. So on average governors get five percentage points more than they would have otherwise. But still I find these effects to be extremely large, I wouldn't have expected this in the first place.

William Howell:

So they're exploiting this discontinuity in media coverage that, we can understand a part from the state of the world. It's that, there's something about round numbers that then stimulates heightened media attention to unemployment rates.

Wioletta Dziuda:

So, can we talk a little bit about this round numbers, this milestone? So, I really liked the results, I think there's a lot of interesting information there. But, I just want to understand whether the interpretation that the authors, I think try to put forward of these numbers is the right one. So, why is it the case that media indeed report more on unemployment when they see this round number? Is it because journalists have this bias, or is it because voters have this bias where they don't focus so much on 4.9, but when they see five, they freak out because I think it's a high unemployment? And journalists just adapt to that, and they provide this information at a higher rate when it's this exciting 5% information.

William Howell:

And because they could be understood. We can't know on the basis of these findings, right? Do we want to associate this with the media presenting stories that they know will actually resonate with the public? Or, is it about independently you can imagine lots of stories being written at the 4.9 threshold, just because the media, journalists think that it's the responsible thing to do, is to also talk about the 4.9, but all those fall flat because nobody knows how to understand that?

Wioletta Dziuda:

Yeah, I wonder and of course it's very hard to do this kind of an experiment but, we have a lot of research in sociology, economics, political science, that shows that people indeed have this weird perception of numbers. 4.9 is not as high as five. If something costs 4.9, then I'm more likely to buy than if something goes five, even though I'm equally likely to buy it if something costs 5.1. So, if people indeed have this weird perception of numbers, perhaps it would be the case that, if the media report the same amount on unemployment every time unemployment changes, and the Bureau of Labor Statistics releases its numbers, people just ignore 4.9, and then their attention picks up when it's five.

We would probably observe exactly the same results, and then we would be telling a story about people are behavioral and then maybe there will be some room for cooking up the numbers and so on and so forth. Or maybe for the media to sort of round things up and down to explore this behavioral response of the voters. But, if that's the story, and I don't know whether it is, but if that's the story, I don't know whether we learn more about, to what extent media affects voters' behavior.

William Howell:

As opposed to our curious relationship with numbers.

Wioletta Dziuda:

Exactly. That our behavior has this weird feature that we would reward a governor if unemployment fell down to three, as opposed to 3.1, and we would do that independent of whether media reports to the same extent, less or more, and the media just is responding to the demand for information from us, from us, irrational human beings.

William Howell:

Okay, that cues, right there, you've just queued our friend and colleague Anthony to come in, and rescue the voter.

Anthony Fowler:

I see, are you going to get me ... okay, I'm going to defend voters. I mean, yes, I guess it's certainly possible, what you're saying is certainly consistent with their paper, it could be the case. Suppose there was no discontinuity in reporting, there could still be an electoral effect because the voters just pay more attention, or care more because, 6.0 sounds like a bigger deal than 5.9. I am personally pretty skeptical that that could be the story. Notice these milestones are not just any round numbers, they're round numbers that also have not been hit in a long time.

For a state that hasn't had 6% unemployment in a long time, and then all of a sudden gets to 6% employment, that's a big deal. And you can imagine that the voters would be upset about that, to learn about that, and obviously they're learning from the newspaper, whether or not, is this a high priority thing or not? And if the news story said 5.9, I would imagine that most voters would say, "That's a big number." Most voters know that 5.9 is close to six.

Wioletta Dziuda:

You would think so, you would think so, but we do have a lot of research showing that, indeed there's this huge discontinuity in our reactions. So, I'm not saying I'm discarding their story, and I'm not saying that there is more, but my prior goes more in the direction of my story. But I think it's impossible to distinguish between those two here, and I think we have a lot of evidence that this other story might be going on too.

William Howell:

I'm not altogether clear on what's at stake here though, Wioletta. Like, we see the kind of milestone coverage that they're talking about vis-a-vis unemployment, in all kinds of domains. We're seeing it right now on the casualty rate of COVID right? When we hit 50,000 that was a big story that went up everywhere. Now, is that because the media for their own reasons feel like 50,000 is worth of story, as opposed to 45, 49,000 does not? Or is it because they anticipate the 50,000 resonating with the larger public in ways that 40,000 would not? And for that reason, they're willing to pay more coverage to it? Either way what Greg is doing, is leveraging this discontinuity that shows heightened media coverage, that then leads to punishments.

Wioletta Dziuda:

So I think there are two stories here. So the first story that I believe in is that, if the death number from COVID in the US before the election reaches certain milestone, suppose we are, instead of 197, we have 200,000, this paper seeks to suggest that there will be an effect on the election. Maybe it will be teeny tiny, but there will be an effect on the election, okay? And I think the data that comes from this paper is, it's pretty clear that the effects seem to be large; when the variable that hits this milestone is indeed important to the voters.

So here I've learned a lot. The second question is, what's the role of the media in generating this effect? Is it that if we somehow forced media to spend exactly the same amount of time on 197,000 deaths, as on 200,000 deaths, would we get very minor effect? Or would we get actually a huge effect? So is it that, could we change the outcomes? Could we make people respond to 197,000 exactly the same way as to 200,000, because those are the same numbers, if we just change how media report? Or is it just there's something in our psychology?

And I think this is not an unimportant question, and especially if you're an editor of a newspaper, or if you are the programming director of CNN and Fox News. Because if you realize that you indeed can affect the outcome, or at least the share of the votes, by selectively deciding how much time to spend on 200,000 versus 147, 97, that's a huge thing.

William Howell:

So if you're a media outlet and you're rooting against Trump, and you want to present some negative news, and you're up to 198,000 the day before election day, going out and issuing that story isn't going to matter, it might not matter at all. You have to hold out until you actually pass the threshold. Or maybe it does that's what's at stake.

Wioletta Dziuda:

And that's what I'm saying, I don't think the paper convinced me in either direction. The paper only told me that somehow hitting 200,000 will matter.

William Howell:

Yeah because they have to take advantage of these discontinuities as they exist in the world, they don't get to randomize the news coverage. If they get to randomize the news coverage, then they could speak to the question that you have and distinguish between coverage at 198 versus 200,000. Instead they're just leveraging this, right?

Wioletta Dziuda:

Yeah, yeah.

William Howell:

So, if we take the findings in this paper seriously and apply them to the upcoming election it seems to me, what we can expect to be Trump's efforts to say, "Hey, before COVID, the economy was booming. It was the best ever," won't resonate, because what really matters is the state of the economy immediately before the election, that how you talk about or reference distant events. At least, there isn't any evidence in this paper that suggests that that's going to go to buy you much.

Wioletta Dziuda:

Yeah, I agree. I was actually puzzled by that. So they also looked at what happens if unemployment reach a certain milestone let's say six months, a year before their election, and they don't see any difference in coverage and they don't see any difference in response. It's not only the effect of that on the election, but also the effect of that on the approval polls that are just concurrent with the unemployment, which is just sort of ... Yeah, I don't know how to take it, but I agree with what you said that if you were a president that you would really worry right now, because you would think, "Whatever I've done before, this is all going to be forgotten."

Anthony Fowler:

Well, that's interesting. There's a few things to say about that. I mean, of course what you do now does influence what unemployment will be right before the election, so there's still an incentive for the president or the governor to work very hard and care about things that voters care about. They do show in the paper that newspapers are much more likely to cover state unemployment right before a gubernatorial election. Similarly, there's just a bigger first stage effect in the sense that there's a bigger discontinuity in coverage across these milestones when the milestone is crossed right before the election than if the milestone has crossed many months before the election. The paper doesn't necessarily show that voters care more about recent unemployment, it just shows that voters are using recent unemployment to learn about the performance of their governor, and they're much more likely to get coverage of unemployment when it's happening right before the election.

William Howell:

We've been talking about what we can do with these findings, to what we can generalize them. I wanted to raise a slightly different issue, which is that the kinds of media effects that they've identified have to do with the media, for the most part, faithfully representing what the actual unemployment rate is. But much of our concern about media effects have to do with fake news and spin and how we talk about things, and do we then make things up? What if I talk about this scandal versus that scandal? Wioletta, you and I, we have a paper.

Wioletta Dziuda:

Yes, we do.

William Howell:

That tries to identify the politics of scandals, and essential piece of that paper is that many scandals are fake, right? That they actually aren't in fact, true, and yet they generate lots of media coverage and accusations that surround behavior that may not have actually occurred.

Wioletta Dziuda:

And affect voters' voting in the end.

William Howell:

Exactly. This is just a particular kind of media effect. It doesn't obviously generalize to all the various other kinds of editorial decisions that editors at newspapers around the country presumably are making.

Anthony Fowler:

I'd like to think, and you can tell me if you think I'm wrong about this. I would like to think that for most respectable news outlets, they do of course have their various partisan slants. Their owners and their readers have their own partisan orientations and so on. Nevertheless, especially in the modern era, it would be extremely controversial and rare for a newspaper to just make up a lie. They might choose which things to cover. There might be a flimsy allegation about some potential scandal and they cover that, but it would be extremely rare for them to just completely fabricate evidence or data of some kind. Generally speaking, we're not worried that people are getting outright lies and fabrications. What we are worried about when we worry about media slant is whether or not the owner who has ... Rupert Murdoch says, "Let's cover this story a lot, and let's downplay that story," and that is what this paper is about, so I would potentially argue that this paper does actually have a lot to say about that kind of media slant.

William Howell:

So, okay, it is about heightened salience of an issue about which people broadly recognize as being vitally important to our politics, so if Rupert Murdoch were to simply say, "I'm going to now raise the salience of something tangential," I'm not sure I'd want to take these findings and say, "You see? He then can alter the electoral returns of an incumbent office just by raising the salience of something." This is about unemployment. Could he do the same thing about, I don't know, pick your economic oddity? My guess is no.

Wioletta Dziuda:

Yeah, and I think there's another effect that might occur. Of course there are people who only listen to or read certain types of newspapers, but those are probably voters that would not be affected by a lot of information. But then when you think about independent voters, presumably they look at the internet and they see new stories from different outlets, and then if Murdoch decided to manipulate voters by just focusing on a particular issue, and you don't see other newspapers or other media outlets focusing on that issue, that might not be as effective as Murdoch might like. So I think extrapolating from that for to what extent media are able to manipulate us by just selectively deciding what to talk about, I think that's still a big open question at this ... I mean, I have an answer in my head, okay? I think they are able to manipulate us, but I think, yeah, there are countervailing forces.

Anthony Fowler:

Yeah. I guess, I think I'm willing to say that I learned more from the paper than you guys are, because in some sense, slant, so there's how much you cover something and then there's also how you cover it, but that's also kind of included in their estimates, right? So I would imagine if we did a deep dive and looked at all these articles, the ones that are covering the new threshold are putting a little bit of a spin on it. They're saying, "We've never had unemployment this high. Not since 2003. Isn't this terrible?" What have you, and so in some ways, that spin is included here in the estimates. The estimates are being generated in a world in which most readers have a general sense of the slant of their paper, right?

Most readers might have a general sense that, "I like the Washington Post, and it has a left-leaning slant." If Rupert Murdoch bought the Washington Post, but I didn't find out about it, and then all of a sudden they start reporting news a completely different way, what would the effect of that be? I think it's hard for us to say, extrapolate from this study and say what that would be, but more reasonable changes, the Washington Post deciding what we are going to start covering. This thing that we don't like about Nancy Pelosi, let's say, I think probably this paper does have something to tell us about that.

William Howell:

So take as an example, Trump's summit with Kim Jong Un, yeah? A day into it or hours into it, this is talking about old news. Here's one. Decides, "I'm going to get up and I'm going to leave. This is going nowhere," and he gets up and he leaves. You can imagine media outlets talking about that in very different ways, a sign of his naiveté for having agreed to go in the first place, or in terms of his strength for not putting up with unreasonable demands being put before him and that he puts the American people first. I guess my suggestion here is that that's a decision that's available to the editor that could go in very different directions. It's one event. There was a summit, and that thing happened, and they could talk about it in very different ways, and editors do this in markedly different ways without lying. When we worry about the influence that the media is having on people's vote choices and how they understand politics, it seems to me they worry a lot about those kinds of decisions and that this paper doesn't speak to it.

To be clear, this is not a rap against the paper at all in my mind. I really liked the paper. I just think that it is important for us to be clear in our minds about the type and the class of media effects that they're documenting, because there are many others that have to do with how you talk about something or that even if it's just about salience, it's the salience about some things are going to vary more. I mean, not just in terms of what the media does, but should they independently decide to give lots of coverage to some things? I suspect it wouldn't have any impact on public opinion, but on some other things they would, just because the media randomly decide to cover increases in exports to Angola, and every time we met a threshold that, boom, they did a whole bunch of additional coverage, I suspect it wouldn't have any impact on political outcomes.

Wioletta Dziuda:

I mean, I have a counterexample to your example.

William Howell:

Oh, let's hear it.

Wioletta Dziuda:

So during Obama's presidency, suddenly it became extremely important for every Republican, especially for Republicans who really don't follow the economy so closely, to care about budget deficits. And this was a huge thing. You would interview people and they're like, "Yeah, budget deficits are the most important thing." Budget deficits were not important during Bush's administration and they stopped being important during Trump's administration extremely quickly, so I think he'll have a very good example where just the talking point is that putting emphasis on something, even if this thing is perhaps the quantity that you don't necessarily understand or you don't care about, makes you care about it. And I can see how this kind of exactly the same channel flows through the media.

William Howell:

How far do you want to take that? I agree with that. It's a good example. That same thing you could talk about thresholds and balance of trade between the US and China. Suddenly we're talking about that. We didn't before. But how far do you want to take that? Is a milestone on every conceivable economic indicator going to matter equally, as long as it receives a lot of coverage in voter behavior? My guess is not, that you can't just sort of generate outrage or adulation about anything. It matters what that underlying phenomenon is, that underlying dimension is.

Anthony Fowler:

I agree with that. I think that actually cuts against one of the things you were talking about earlier, Will, which is, of course it does matter how the media puts a spin on things, but they're constrained in how much they ... You can't go so far. Nobody can write that story that says, "Wow, our state just hit 20% unemployment for the first time ever. Isn't that fantastic? Our governor is doing such a great job." Nobody is doing that and if they did, nobody would take them seriously. And so, yes—

William Howell:

No. I agree. When it comes to the milestones, there's a clear valance associated with them that constraints how you can talk about it. But we have a lot of evidence that shows that people get their news from sources outside of traditional news sites. But we worry about media effects and slant. Your point is, "Well, there's not all that much slant when it comes to certain kinds of news outlets." And that may or may not be true, but there certainly are domains where there's a tremendous amount of spin to the extent that people get their news from those outlets, you'd want to know what effect that spin has on people's views.

We just don't get that from this paper. We don't know. And it may be that when they used to listen to Bill O'Reilly, it was just entertainment. People said, "I know what I'm getting from him, and I'm not going to update it all because I know I'm getting a bunch of spin." That could be, in which case there aren't big effects there at all. But I don't know. I just think the class of media effects is big and vast and we've got one piece of them here in this paper.

Anthony Fowler:

So do you find this is a topic we come to a lot? How good do we feel about democracy or not? One version of the story is, look at how manipulable these voters are. They just see one headline that says the governor's doing a good or bad job and then they change their votes. The optimistic interpretation is unemployment is a very important thing that voters care a lot about and unemployment gives them a lot of information about how well their governor is performing and they are responsive to information. They would like to get more information. They would like to be reading.

In a perfect world, they'd be getting a ton of information and these thresholds wouldn't matter, but when you do give them the information, they respond and they say, "Oh, look. My governor, I thought he was doing a good job, but he's not. Look, unemployment's really high." And they change their vote. So they're not so hyper-partisan. They're not so diluted. They're not so bamboozled that they just vote whichever way their party tells them to vote or whatever it is. They actually respond to the information and it seems like a good thing.

William Howell:

Well, some do. We don't know. It could be that plenty of people are hearing that we're at a bad milestone and they're like, "Forget it. I'm going to either ... Either these other things compensate for that, or I'm going to pretend like I never heard that." I think, to your point, I agree with you to the extent that there are effects, they are effects that are tied to something that plainly matters and they're not perverse. They're not like, "I hear that something bad happened and now I'm therefore going to double down on the incumbent."

And so, to the extent that the finding speak to the health of the democracy, you'd have to say that they are a reason to be optimistic. If you had no effects, then what you'd say is, "Well, people aren't updating at all and that's a problem." Or maybe they're right, that they're getting new news and we have lots of background studies that show that it isn't the lack of updating, it's because they're fully informed about the state of the world. And alternatively, if you saw effects that were perverse, of opposite sign, then you'd say, "What the hell is going on?" Here, you've got updating in line with the valence of the news. And so I don't know. That seems good.

Wioletta Dziuda:

Yeah. And I think I wouldn't even worry too much about this behavioral response that voters have, that they don't pay attention to 4.9, but pay attention to 5. After all, we are human beings. We have limited attention. We can only pay attention to certain things. So I think it's reasonable to believe the evolutionarily somehow we decided we are going to pay attention to round numbers, but as long as we do pay attention to some numbers, I think on average, we are still responding to the economic circumstances when we vote. And I think that's good news for us.

William Howell:

So, is that your bottom line, Wioletta? Is that where you come out?

Wioletta Dziuda:

Yes.

William Howell:

I really liked the paper and learned a bunch from it. I didn't have strong priors about ... I didn't expect there to be such notably asymmetric effects and that it does speak to really important ways in which the media can shape the voters' understanding of incumbent politicians and their willingness to continue to back them. Where do you come out, Anthony?

Anthony Fowler:

I come out thinking I learned a lot from this paper, actually. I remember the first time I read the paper and I was actually kind of skeptical for a lot of the reasons that we talk about being skeptical of social science, in general, on this podcast. You think like, "Is this really plausible? Only so many people read the paper? And is it really plausible that there's that much of a discontinuity at these round numbers?" Just even on the question of whether or not there would be such a big discontinuity in news coverage. I wasn't sure that that would even be true.

And I was gradually convinced as I read the paper and I even had some exchanges with the authors, and I learned more about the paper and saw more of the evidence. And I gradually became convinced that no, there really is something here. That the news reporters do really respond in this way and it does have an electoral effect. And so I thought we're learning something about the media. We're learning that the media signals affect voters. We're learning something about the voters, too. We're learning that they are trying to incorporate this economic information into their vote choices and that when you give it to them, they do, in fact, change their positions.

William Howell:

Thanks for listening to Not Another Politics Podcast.

Wioletta Dziuda:

Our show is a podcast from the Harris School of Public Policy and is produced by Matt Holdup. Thanks for listening.