Mainstream media and election results

Recent election results such as the election of Donald Trump in 2016 or the Brexit vote that year have surprised many of us. Many saw these results as ‘wrong’ in the sense that the data had not been properly combined. In particular, factually incorrect beliefs persist about many aspects of these choices and determine these overall preferences. The suspected culprit for the misinformation that has led to such choices is the new media, often accused of bias and bias, a debate that continues.

The media landscape has changed drastically in the last few years. Mainstream cable news used to be the dominant news provider, but in today’s environment, people can receive news from a plethora of possible media sources – from print media to YouTube channels. The prosperity of new media has probably benefited from allowing them to tailor their media preferences to their needs. However, the incredible diversity of viewpoints on offer, combined with new technology, has made it easier for citizens to distance themselves from the potentially conflicting viewpoints offered by traditional media outlets. Several papers have looked at the effect of echo chambers and the selective consequences of unawareness of being in echo chambers (e.g. Levy and Razin 2015, Ortoleva and Snowberg 2015).

Matejka and Tabellini (2020, 2021) also show that the diversity of available media sources allows people to shift attention from important but non-controversial issues to more extreme policy dimensions, thus increasing their influence.

Although the media landscape has become much richer, trust in traditional media has declined significantly over the past two decades. In the last five years in particular, this media distrust has followed radically different paths on either side of the US political spectrum. As we can see below, the gap in media trust between Republicans and Democrats has been staggering and widened during Trump’s presidency – perhaps due to the former president’s well-documented mainstream media.

Figure 1 Asymmetry of trust in mass media

Indeed, as noted by the Pew Foundation in Jurkowitz et al. (2020), “One of the clearest differences between Americans on the opposite side of the political aisle is that a larger share of Democrats trust a greater number of news sources.”

The impact of the aforementioned phenomenon on collective electoral outcomes is complicated, particularly in the United States, by the presence of a highly polarized landscape in which traditional ideological, religious, and ethnic identities are being replaced almost entirely by overlapping meta-identities. Democratic and Republican political beliefs. Citizens have become less responsive to new information or real national issues, as if political affiliation determines what information people absorb, rather than the other way around (e.g. Mason 2018). Furthermore, Kahan (2017) finds that people display motivated reasoning to defend political identity, while Kanders et al. (2022) note that individuals search for information to preserve their beliefs.

Relatedly, Guriv et al. (2019, 2021) show how the Internet has increased support for both left-wing and right-wing populist movements by reducing the cost of reaching voters.

These three phenomena—the emergence of a dense array of media outlets, partisan mistrust of the media, and media choice by citizens—reflect how political beliefs are formed and updated and thus the response to public decisions on election day. But can this new information environment create enough biased aggregate beliefs for an election? Further, could this provide a rationale behind the burgeoning distrust in mainstream media observed during Trump’s presidency?

Identity-based choice

In view of the above, our recent paper (Herrera and Sethi 2022) explores the link between political identity, media preferences and electoral outcomes.

We assume that every citizen identifies with a party on the left or the right and aims to protect their political identity. They choose media to maximize the probability that, after their beliefs are rationally updated, they believe their team is the best match for the world state. Such beliefs are based on only two cues: an ‘internal’ cue and an ‘outside’ cue. The internal signal is constructed from a structure that represents the collection of media outlets that an agent chooses. Signals coming from external outlets an agent is inadvertently exposed to – the nature of this signal will largely depend on the nature of the mainstream media. In other words, citizens choose media in an attempt to protect themselves from potentially unfavorable (from the perspective of their political affiliation) outside news to which they have some exposure.

Equally, one can think of agents as having two natures – a heart and a mind. The heart selects internal media while the mind votes sincerely. The purpose of the heart is to convince the mind to vote for the party of the heart’s choice. In both interpretations, citizens are fully rational in the way they process all the information they receive and update their beliefs based on the two signals and vote according to their hindsight for the better candidate.

The election aggregates all votes, each conditional on two independent signals of which two candidates are preferable. The core behavioral assumption of our model is not information processing but preferences that drive each citizen’s choice of ‘in-media’, i.e. suitable media outlets, each of which we view as a specific known signal structure.

Inequality in exposure to external news

Citizens who distrust the mainstream media may choose to avoid exposure to it. Thus, an asymmetric distrust of mainstream media may imply an asymmetry in exposure. The nature of media that agents choose to use (‘in-media’) is influenced by their beliefs about mainstream (‘outside’) media. Differences in media preferences have surprising implications for electoral outcomes.

In the example below, we study the size of the electoral advantage (margin of victory) of the party less exposed to the mainstream media and assume that each citizen votes for the party he or she believes to be rationally superior. We assume that both types of citizens (left- or right-affiliated) are symmetric other than exposure. In particular, ‘L’ and ‘R’, have a large and equal number of biases of each type. Also, there are two equally likely states of the world, ω = L, R, indicating which of the two candidates is better.

To illustrate the impact of the richer media landscape made possible by the proliferation of new media, we first consider electoral outcomes in the absence of inside media and compare them to outcomes where citizens choose their inside media.

Type-L citizens receive iid symmetric binary signals from mainstream news with an accuracy of 0.75, while type-R citizens receive more noisier iid mainstream signals with a lower accuracy of 0.51. The winning margin and winning probability for the R side depends entirely on the perception of outside cues and is summarized in the table below:

Thus, asymmetric exposure to mainstream media produces symmetric electoral outcomes. In this baseline case, the ideal/correct candidate is always selected, i.e. the information is perfectly integrated. There are no individual media preferences by citizens, and thus political beliefs, or citizen type, play no role.

Now, suppose instead that citizens can also optimally curate internal sources of media. Here, their voting decisions are made after reasonably updating on not one, but two signals. If each agent chooses the media to maximize the chance of preserving political beliefs, then election outcomes are no longer symmetric. In fact, it can be severely skewed. In this example, R’s margin of victory and probability of winning are:

In this case, the R party has the advantage of the former winning margin in the election. Surprisingly, it also has ex-post facilities. Namely, party R wins the election in any state in the world in this case and the agents vote based only on the information they receive (logical updates), but the information is not aggregated.


The surprising result above is driven by the fact that citizens of both political parties choose qualitatively different media. Low-exposed citizens (rights in our example) struggle with much less informative outside cues, implying that they maximize their chances of maintaining their political beliefs by choosing a one-sided cue structure, for which news favorable to a candidate is too frequent and Thus not so informative, while adverse news is rare and therefore harmful. It’s probably reminiscent of partisan outlets like Fox News or Breitbart News for Republicans. On the other hand, to contend with more informative outside signals, more open citizens (the left in our example) optimally choose more balanced news.

Importantly, in a world where there is not a rich set of signal structures (in-media) available for agents to choose from, even if there is partisan bias that drives media choice, we would not see a complete bias in the total electoral outcome in the example above. .

Additionally, we find that an asymmetry in distrust of the mainstream media by partisan citizens can lead to a significant electoral advantage for the less trusting party. The way this advantage is created is that distrustful citizens essentially ignore a hostile outside signal, prompting them to choose more biased domestic media, which is more likely to preserve their political beliefs even in the wrong state of the world. This may be one of the reasons we have observed that party elites disparage the mainstream media.

If citizens of both parties also have different antecedents biased toward their side, the scope for failure of information integration increases. The results remain qualitatively unchanged if we assume that each citizen votes for his culturally endorsed party only if he believes it is good, and abstains otherwise, as we assume that voting/position margins determine electoral outcomes. In this case, all winning margins will be only half. The results are also robust to agents placing utility in holding the posterior more favorable to them or to small amounts of utility in voting for the correct party.


Benkler, Y, R Faris, H Roberts and E Zuckerman (2017), “Study: Right-Wing Media Ecosystem Led by Breitbart Shifts Broader Media Agenda”, Columbia Journalism Review.

DeLavigna, S. and E. Kaplan (2007), “The Fox News Effect: Media Bias and Voting”, Quarterly Journal of Economics 122(3): 1187–1234.

Guriev, S., N. Melnikov and E. Zhurovskaya (2019), “Knowledge is power: mobile internet, government trust and populism”,, 31 October.

Guriev, S, N Melnikov and E Zhuravskaya (2021), “3g Internet and Trust in Government”, Quarterly Journal of Economics 136(4): 2533-2613.

Herrera, H and R Sethi (2022), “Identity-Based Selection”, CEPR Discussion Paper 17203.

Jurkowitz, M, A Mitchell, E Shearer and M Walker (2020), “US Media Polarization and the 2020 Election: A Nation Divided”, Pew Research Center.

Kaanders, P, P Sepulveda, T Folke, P Ortoleva and B De Martino (2022), “Humans actively support exemplar evidence to support prior beliefs”. Alife 11: e71768.

Kahan, Dan M. (2017), “Misconceptions, Misinformation, and the Argument of Identity-Protective Knowledge”, Yale Law and Economics Research Paper No. 575.

Levy, G and R Razin (2015), “Correlation Neglect, Voting Behavior, and Information Aggregation”, American Economic Review 105(4): 1634-45.

Mason, L. (2018). The Uncivilized Pact: How Politics Became Our IdentityUniversity of Chicago Press.

Matějka, F and G Tabellini (2020), “Political information in the age of the Internet”,, 10 January.

Matějka, F and G Tabellini (2021), “Electoral Competition with Rationally Indifferent Voters”, Journal of the European Economic Association 19(3): 1899-1935.

Ortoleva, P and E Snowberg (2015), “Overconfidence in Political Behavior”, American Economic Review 105(2): 504–35.

Leave a Reply

Your email address will not be published.