Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic...

41
Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich * Matthew A. Baum Adam J. Berinsky Justin de Benedictis-Kessner § Teppei Yamamoto January 22, 2020 Preliminary version prepared for the 2020 Southern Political Science Association Conference Abstract The single shot nature of experiments on the effects of partisan media on public opinion may limit the relevance of estimates that such studies produce for politics and policy. For example, there might be cumulative effects from multiple doses of partisan me- dia such that the combined effect of repeated exposures on political attitudes is much greater than that of a single dose. Similarly, the persuasive effect of partisan media might be temporary and decay quickly after a single exposure. We implement a novel multi-wave experiment that allows us to examine these concerns. We find that the per- suasive effects demonstrate substantial durability, decaying only mildly over the course of a week following treatment. Additionally, we find no evidence of cumulative effects of repeated exposure to partisan media, and instead slight moderation. Together, these results suggest that partisan media’s influence on public opinion is persistent, but the additive effects of “filter bubbles” are limited. We appreciate the research assistance of Grace Chao, Henry Feinstein, and Kaitlin Tucci, and funding from the National Science Foundation (SES-1528487) and the Political Experiments Research Lab (PERL) at MIT. * PhD student, Department of Political Science, Massachusetts Institute of Technology, [email protected] Kalb Professor, John F. Kennedy School of Government, Harvard University, Matthew [email protected] Mitsui Professor, Department of Political Science, Massachusetts Institute of Technology, [email protected] § Assistant Professor, Department of Political Science, Boston University, [email protected] Associate Professor, Department of Political Science, Massachusetts Institute of Technology, [email protected]

Transcript of Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic...

Page 1: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Dynamic Persuasion: Decay and Accumulation ofPartisan Media Persuasion

Zachary Markovich∗ Matthew A. Baum† Adam J. Berinsky‡

Justin de Benedictis-Kessner§ Teppei Yamamoto¶

January 22, 2020

Preliminary version prepared for the2020 Southern Political Science Association Conference

Abstract

The single shot nature of experiments on the effects of partisan media on public opinionmay limit the relevance of estimates that such studies produce for politics and policy.For example, there might be cumulative effects from multiple doses of partisan me-dia such that the combined effect of repeated exposures on political attitudes is muchgreater than that of a single dose. Similarly, the persuasive effect of partisan mediamight be temporary and decay quickly after a single exposure. We implement a novelmulti-wave experiment that allows us to examine these concerns. We find that the per-suasive effects demonstrate substantial durability, decaying only mildly over the courseof a week following treatment. Additionally, we find no evidence of cumulative effectsof repeated exposure to partisan media, and instead slight moderation. Together, theseresults suggest that partisan media’s influence on public opinion is persistent, but theadditive effects of “filter bubbles” are limited.

We appreciate the research assistance of Grace Chao, Henry Feinstein, and Kaitlin Tucci, and fundingfrom the National Science Foundation (SES-1528487) and the Political Experiments Research Lab (PERL)at MIT.∗PhD student, Department of Political Science, Massachusetts Institute of Technology, [email protected]†Kalb Professor, John F. Kennedy School of Government, Harvard University,

Matthew [email protected]‡Mitsui Professor, Department of Political Science, Massachusetts Institute of Technology,

[email protected]§Assistant Professor, Department of Political Science, Boston University, [email protected]¶Associate Professor, Department of Political Science, Massachusetts Institute of Technology,

[email protected]

Page 2: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Introduction

Statements by pundits, politicians, and citizens alike lament the rise of political polarization

and the role that partisan news media may have played in this process. The proliferation of

choices in both television and online media allows for individuals to select into ideologically

consonant news sources – selective exposure (e.g. Lazarsfeld, Berelson, and Gaudet, 1948;

Sears and Freedman, 1967) – and potentially may lead to consumption of exclusively one-

sided news content (Groeling, 2013). Over time, this might lead to people believing this

one-sided version of issues without attention to the multifaceted nature of reality.

Recent research has shown that such ideological self-selection does occur (Arceneaux,

Johnson, and Murphy, 2012; Gaines and Kuklinski, 2011; Iyengar and Hahn, 2009; Stroud,

2011), though it may not be as extensive as previously assumed (Bakshy, Messing, and

Adamic, 2015; Brundidge, 2010; Guess, 2018; Messing and Westwood, 2014; Mummolo,

2016). Other research has suggested that, in addition to polarization resulting from selec-

tive exposure, partisan media itself may further polarize people’s attitudes (Bullock, 2011;

de Benedictis-Kessner et al., 2019; Dilliplane, 2014; Feldman, 2011; Levendusky, 2013b).

This may crucially depend on the degree to which the people exposed to such partisan me-

dia might self-select into consuming that media on their own (de Benedictis-Kessner et al.,

2019; Leeper, 2017).

Zaller’s (1992) theory of opinion change also suggests that experimental observations of

persuasion by partisan media may occur because such experiments make certain considera-

tions more accessible to survey respondents. Experimental results indicating the moderation

of persuasion effects by media preferences may be a consequence of those considerations be-

ing differentially accessible to different types of media consumers. This would suggest that

such persuasion effects are artificial and would quickly decay, in contrast with more long-

lasting opinion changes from new information (Coppock, 2017; Coppock et al., 2018). Such

competing theories of opinion change present contrasting normative implications: if survey

experimental treatment effects quickly decay, they are less of a problem for democracy than

1

Page 3: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

if they are persistent (Gaines, Kuklinski, and Quirk, 2007).

Single-exposure experiments are inherently limited in their ability to detect such the-

oretical processes as they ordinarily cannot assess the long-term impacts of experimental

treatments. In this paper, we avoid this problem by conducting a large-scale multiwave

experiment that tests for both the accumulation of media persuasion and the decay of such

persuasive effects. We find that partisan media has limited cumulative effect when people

are experimentally exposed to such treatments multiple times relative to a single time over

the course of multiple weeks. In addition, we find that the persuasion effects of partisan

media only decay slightly over the course of a week. These results indicate, in the framework

of Zaller (1992), that partisan media may provide additional considerations to consumers

rather than simply making certain considerations more accessible. These results support

a theory of opinion change that explains previous findings on the moderation of partisan

media persuasion and extend theory to longer time horizons. Consequently, our work recon-

ciles previous experimental findings with both real-world patterns of polarization and public

opinion theory. Moreover, our results suggest that partisan media’s persuasive impact may

be especially potent and long-lasting among the people who have fewer previous considera-

tions. This may consequently lead to more severe polarization and problems for democratic

accountability.

Accumulation of Partisan Media’s Effects

Most people who watch or read partisan media do so repeatedly. Studies examining the

effects of a single exposure to partisan news therefore are limited in their ability to predict

real-world effects of such media (Slater, 2004). Limited studies, however, have used ex-

perimental designs incorporating multiple exposures to partisan news and assessed whether

repetition of the arguments contained in partisan news has an additive or cumulative ef-

fect. This research has pointed to limited accumulation of persuasive effects after repeated

2

Page 4: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

exposure to media dependent on the content, circumstances, and individuals receiving infor-

mation.

In one example of such research, Levendusky (2013a) shows survey respondents either

like-minded partisan news editorials or apolitical articles in a two-wave experiment. His

effects demonstrate that multiple exposures to like-minded partisan media can polarize peo-

ple’s attitudes more than just a single exposure, but that this effect is smaller than the initial

effects of (novel) political news. Thus we might expect that persuasion effects compound

after multiple exposures relative to single exposure to partisan media. Similar research test-

ing the effects of repeated exposure to political media indicates limits to this accumulation,

and suggests that repeated exposure may have no effect beyond that of a single exposure

(Lecheler and de Vreese, 2013). This research is in line with communication theory predicting

that repeated framing of information may be redundant (Baden and Lecheler, 2012).

Limited accumulation of persuasion effects may be due to qualities of both the indi-

vidual and the information being presented. For instance, “pretreatment” effects of the

prior exposure to partisan media that limit the degree of attitude change that can occur

subsequently (Druckman and Leeper, 2012). Similarly, if individuals are exposed to cross-

cutting arguments that counter one another with competing messages, repetitive exposure

may not have an additional impact on attitudes or beliefs (Chong and Druckman, 2010).

Thus individuals’ prior exposure may moderate the cumulative nature of persuasion effects

(de Benedictis-Kessner et al., 2019; Huber and Arceneaux, 2007; Zaller, 1992). This would

lead us to expect large initial effects and large effects among those with limited previous ex-

posure, but little additional impact of repeated treatment by partisan media. The message

itself may also influence the accumulation of attitude change. Negatively framed information

may be more effective when repeated, while positively framed information is not (Lecheler

et al., 2015). Similarly, if the topic of information is relevant to individuals (Cacioppo and

Petty, 1979; Liu et al., 2019) or the argument contained in that information is stronger (Ca-

cioppo and Petty, 1989), multiple exposures to a certain framing of that information may

3

Page 5: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

be more persuasive than a single exposure.

Persistence and Decay

Partisan media’s persuasive effects over time may accumulate, but they may also decay.

If persuasion effects are short-lived they may not be as critical a problem for democratic

deliberation than if they are persistent (Gaines, Kuklinski, and Quirk, 2007). Identifying

the long-term potentially pernicious persistence of persuasion effects is a crucial step to

understanding the role of the media in reality.

Previous work on informational framing in news articles indicates that experimental

estimates of persuasion may be quite persistent.1 Partisan media’s polarizing impact, in

particular, may endure for at least two days (Levendusky, 2013a, p. 85-86). Some persuasion

may only decay in size by approximately half over ten days, and persist at that level for up to

a month (Coppock et al., 2018). This may be due to the content of the information presented

to people. Persuasion may be especially persistent if it provides new information to people

rather than simply framing existing information (Coppock, 2017), or if it negatively frames

information (Lecheler and de Vreese, 2011). Partisan media’s content often involves the

provision of new information as well as persuasive arguments, and so its persuasive effects

may be persistent.

Some examinations of this question suggest that persuasion effects may decay, but only

in certain segments of the population — such as those who are high in “need for cogni-

tion” (Chong and Druckman, 2010), or moderate in levels of political knowledge (Baden

and Lecheler, 2012). Such individual-level characteristics may moderate the persistence of

partisan media’s influence.

On the other hand, research on campaigns and highly polarized policy issues suggests a

1One related theory in research on media effects that we do not discuss at length here argues that the per-suasive impact of political communication will not manifest immediately, but will have a delayed “sleepereffect” (e.g. Hovland, Lumsdaine, and Sheffield, 1949; Hovland and Weiss, 1951). More recent examina-tions of the sleeper effect, however, have struggled to show any evidence of its existence except in limitedcircumstances (Capon and Hulbert, 1973; Coppock, 2017; Gillig and Greenwald, 1974; Jensen et al., 2011).

4

Page 6: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

different conclusion on the persistence of persuasion. Hill et al. (2013) show that presidential

campaign advertisements have short-lived effects on attitudes that decay by 50% after 4

days, but that still persist even 6 weeks later, albeit with small effect sizes. Thus, in highly

salient campaigns, political persuasion by advertisements may be minimally persistent. Hill

et al. (2013) also show that the persistence of political advertising’s effects is even smaller in

lower-level (Senate, House, and gubernatorial) elections. Similarly, Dowling, Henderson, and

Miller (2019) show that while informational treatments about the Affordable Care Act can

durably affect people’s levels of knowledge about policy-related facts, the effects on policy

attitudes do not persist even one week later. Likewise, the effects of news information on

evaluations of government may also dissipate within a single week (Kalmoe et al., 2019).

Thus in both low-salience elections and on policy issues, attitudinal persuasion may be

short-lived. Partisan media’s effects on policy attitudes such as those we examine here may

therefore decay quickly.

Research Design and Data

To assess these questions, we conduct a large-scale experiment embedded within a three-wave

panel survey using a version of the PICA design (de Benedictis-Kessner et al., 2019; Knox

et al., 2019) adapted for a multi-wave context. A visual diagram of this adapted design is

presented in Figure 1, showing the first wave at the top, the second wave in the middle, and

the final wave at the bottom.

In this design, we first measure the stated preference of all respondents over three media

options early in the survey. Specifically, we asked: “If you were given the choice of news

articles from the following three sources to read, which of the three would you choose?” We

presented each choice with an accompanying logo of the network, while randomizing the

order in which the options appeared on the screen. We then move to a “washout” period,

where we asked participants to answer demographic questions not directly related to the

5

Page 7: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Figure 1: Experimental Design

Wave 1 W1 “forced exposure” condition

randomize and assigntreatment

MSNBC

Entertainment

Fox

measureW1

outcomes

measurestated

preference“washout”

periodrandomize

designs

MSNBC

Entertainment

Fox

W1 “free choice” condition

measure and assignrevealed preference

MSNBC

Entertainment

Fox

Wave 2 W2 “forced exposure” condition

randomize and assigntreatment

MSNBC

Entertainment

Fox

measureW2

outcomes

measurestated

preference“washout”

periodrandomize

designs

MSNBC

Entertainment

Fox

W2 “free choice” condition

measure and assignrevealed preference

MSNBC

Entertainment

Fox

Wave 3

measureW3

outcomes

media choice, and to complete distraction tasks.2 We then randomly assign participants

to either a forced exposure or a free choice treatment condition in each of the first two

waves. Those in the forced exposure condition were assigned to read one of three news

reports, while those in the free choice condition were allowed to choose one of these media

options.3 Participants in both conditions then read their assigned or chosen news reports

from either Fox News, MSNBC, or an entertainment network (the Food Network) that we

created from real online news stories and then edited to equalize their length and framing.

2The goal was to minimize the possibility that measuring stated preferences might contaminate respondents’voluntary choices later in the survey. These distraction tasks enhance external validity by allowing for theinstability of preferences over time.

3For those in the free choice arm, we asked, “Which of these three articles would you like to read now?”

6

Page 8: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

For the partisan media treatments, the articles discussed either the economic, social, safety,

or public health effects of legalizing marijuana. On each topic, the Fox News and MSNBC

articles were nearly identical except for text that was either supportive of (MSNBC) or

opposed to (Fox) marijuana legalization. The Food Network articles discussed how to save

money while grocery shopping, tips for buying meat, and how grocery stores might change

in the future.

Finally, we asked respondents a number of questions measuring their opinions about mar-

ijuana policy and anticipated behavior regarding the stories they had read. Specifically, in all

three waves we asked respondents ten questions about their policy preferences on marijuana

and drugs. We asked them if they agreed or disagreed with the following statements, “The

legalization of marijuana leads to fewer people using more serious drugs, such as heroin

and cocaine” and “Marijuana use increases violent crime,” corresponding to the articles

about public safety and health consequences of legalization. Similarly, we asked respondents

whether legalization would make the economy better or worse, which corresponded directly

to the interventions about the tax and economic implications of legalization. Other ques-

tions addressed feelings regarding marijuana more broadly. Respondents placed themselves

on 7-point scales between strongly agreeing or strongly disagreeing with the phrases: “Gov-

ernment efforts to enforce marijuana laws cost more than they are worth,” “Using marijuana

is morally wrong,” “Marijuana should be legal for medical use,” “Marijuana use is a serious

problem today,” and “Marijuana should be legal for recreational use.” Finally, respondents

placed themselves along another seven-point scale between, at one end, whether habitual

drug use should be thought of as a criminal offense or, at the other, a medical problem.

They then placed several substances, including marijuana, on scales from very dangerous

to very safe. We recode all opinion variables to 0-1 intervals, with the most liberal or per-

missive of legalization at 0 and the most conservative or opposing of legalization at 1. To

capture variation in underlying latent attitudes and beliefs about marijuana, we form an

internally-consistent additive index of these ten opinion questions (α = 0.89). We use this

7

Page 9: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

index as the primary dependent variable on which we assess polarization and persuasion.

In the first two waves, we also asked respondents to indicate their likelihood of forwarding,

discussing, posting to social media, or seeking out additional information on the story they

had just read. We combined these four measures into an additive index of sharing behavior,

which ranges from 0 (least likely to share) to 1 (most likely to share).

We administered this experiment via an online survey to a national sample of respondents

recruited through Survey Sampling International (SSI).4 In the first wave, this consisted of

7,393 survey respondents in the first wave, 4,926 in the second wave, and 4,526 in the third

wave. Respondents to the first two waves were invited to participate in the next wave one

week after they had completed the previous survey wave. This yielded a final retention rate

in the third wave of 61.2%.

Using these data, we assess the persuasive treatment effects of partisan media, as well

as the decay and the cumulative nature of these treatment effects with a series of models.

First, to assess immediate persuasion effects, we fit the following model:

Marijuana indexij = β0 + β1Treatij (1)

where i designates respondents and j designates experimental waves. This specification thus

examines the causal effect of a given media treatment on attitudes measured immediately

afterwards, collapsing across the first two waves in our experiment.

We also consider whether partisan media’s persuasive effects varied by respondents’ stated

media preferences. To do this, we interacted their stated media preference with the treatment

received using the following model specification:

4SSI recruits participants through various online communities, social networks, and website ads. Whendeploying a particular survey, SSI randomly selects participants for survey invitations. We asked SSI torecruit a target population that matched the (18 and over) census population on education, gender, age,geography, and income. The result is a diverse national sample, albeit not a probability sample.

8

Page 10: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Marijuana indexij = β0 + β1Media Preferenceij + β2Treatij + β3Treatij ×Media Preferenceij

(2)

Next, we assess the decay of these treatment effects by regressing the outcome measured

in wave j + 1 on the treatment in wave j. These models take the following form:

Marijuana indexi,j+1 = β0 + β1Treatij (3)

such that we measure the effect of treatment in wave 1 on wave 2 outcomes and the effect

of treatment in wave 2 on wave 3 outcomes.

We also assess the cumulative impact of partisan media using our multiwave panel. We

examine whether treatment with the same type of media twice results in a greater change in

respondent opinions than after treatment only once, or if there are no such additive effects.

To examine this type of accumulation, we fit the following model:

Marijuana Index Wave 2i = β0+β1Treatment Wave 1i+β2Treatment Wave 2×Treatment Wave 1i

(4)

This approach allows for there to be significant heterogeneity in the effect of treatment

with different kinds of media, but limits our statistical power for certain analyses as some

subgroups of the analysis are relatively rare.5

5For example, people who prefer MSNBC but were exposed to Fox News twice comprise under one twenty-seventh of our full forced choice arm.

9

Page 11: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Results

Persuasion Effects

We begin our analysis by replicating previous findings about the short term persuasive

effects of partisan media in a single round experiment. We present estimates for the effect

of partisan media on attitudes expressed in the same wave in Table 1. Our results reveal a

clear persuasive effect of partisan media. The forced choice results indicate that respondents

who read Fox News or the entertainment article rather than MSNBC reported attitudes that

were significantly more conservative. In particular, the effect of reading Fox News rather

than MSNBC is approximately the same size as the attitudinal difference between responses

from Democrats and independents in the free choice arm of our first wave.

Table 1: Media Persuasion Effects, Waves 1 and 2 pooled

Dependent variable:

Attitudinal Index

Entertainment 0.013∗

(0.008)

Fox 0.031∗∗∗

(0.008)

Constant 0.357∗∗∗

(0.005)

Observations 5,173

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Effect Heterogeneity By Media Preference

We next present our forced choice group results disaggregated by respondents’ stated pref-

erences. Figure 2 shows the persuasive effects of Fox relative to MSNBC in our three stated

10

Page 12: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

preference groups for the first and second waves pooled. In both survey waves, those re-

spondents who preferred Fox and MSNBC who were assigned the Fox article reported more

conservative opinions than respondents that read the MSNBC article. In contrast, among

those respondents who preferred entertainment, there was a more muted persuasion effect.6

● ●

Attitudinal Index Sharing Index

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.15

(Moreliberal)

−0.10

−0.05

0.00

0.05

(Moreconservative)

0.10

−0.15

(Lesswillingnessto share)

−0.10

−0.05

0.00

0.05

(Greaterwillingnessto share)

0.10

Trea

tmen

t effe

ct o

f rea

ding

Fox

rat

her

than

MS

NB

C, W

1/W

2 co

llaps

ed

Figure 2: Persuasion Results, Wave 1 and 2 pooled

We present these results pooling treatment in both wave 1 and wave 2 on the same-wave

attitudes in tabular form in Table 2. The treatment effect of exposure to Fox News rather

than MSNBC, shown in the second line of the table, is positive and statistically significant

in the groups of respondents who preferred Fox or MSNBC, but not those who stated a

preference for entertainment. 7

6If the two waves are pooled together (as we do here and in Appendix Table A1), the results suggestsignificant persuasive effects among Fox and MSNBC prefers but not among those who prefer entertainment.However, when we disaggregate the two waves in Appendix ??, we find a statistically significant effect amongentertainment preferrers in wave 1.

7In addition, in Appendix A we present analyses examining the heterogeneity in the effect sizes amongstated preference subgroups. The small interaction terms in the models suggest little heterogeneity in ourtreatment effects by respondents’ stated media preferences.

11

Page 13: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Table 2: Immediate Persuasion Effects, Stated Preference Subgroups

Prefer:Entertainment Fox MSNBC

(1) (2) (3)

Entertainment 0.006 0.010 0.020(0.013) (0.012) (0.013)

Fox 0.017 0.039∗∗∗ 0.034∗∗∗

(0.013) (0.012) (0.013)

Constant 0.345∗∗∗ 0.420∗∗∗ 0.285∗∗∗

(0.009) (0.009) (0.009)

Observations 1,576 2,083 1,516

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Decay of Persuasive Effects

We next examined the durability of these persuasive effects. In Figure 3 we show these

results using the treatment in one wave on outcomes in the following wave. 8 Comparing

these figures to Figure 2, we can see that the treatment effects of Fox relative to MSNBC

are similar when measured with same-wave attitudes as when measuring attitudes in the

following wave. These results suggest that partisan media’s persuasive effects decay only

slightly over the course of one week.

We also present these results in Table 3, pooling across waves of the survey and respondent

preference subgroups, in order to look at the holistic effect of treatment in wave j on outcomes

in wave j + 1.

These results further demonstrate the mild decay in the size of persuasion effects between

waves. The treatment effect of reading Fox News rather than MSNBC from the forced choice

8We also present results not collapsed across the two waves of treatment in the Appendix, where we examinethe effects of treatment in wave 1 on outcomes in wave 2, using only the free choice arm of wave 2. as wellas the effect of treatment in wave 1 on outcomes in wave 3, and treatment in wave 2 on outcomes in wave3.

12

Page 14: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

● ●●

Attitudinal Index Sharing Index

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.15

(Moreliberal)

−0.10

−0.05

0.00

0.05

(Moreconservative)

0.10

−0.15

(Lesswillingnessto share)

−0.10

−0.05

0.00

0.05

(Greaterwillingnessto share)

0.10

Trea

tmen

t effe

ct o

f rea

ding

Fox

rat

her

than

MS

NB

C o

n ne

xt w

ave,

W1/

W2

colla

psed

Figure 3: Effects of Partisan Media on Next Wave Outcomes

arm of the experiment declines by only 24 % relative to the estimate of treatment on attitudes

measured immediately following media exposure from Table 1. These small decreases in the

size of our treatment effects over the course of a week suggest that even a single dose of

partisan media may have more long-lasting effects than has been frequently assumed.9

Table 4 shows these results disaggregated by respondent’s media preferences. These

results uncover some heterogeneity in the decay of partisan media’s treatment effects. In

particular, the treatment effect of partisan media among people who prefer entertainment

appears to decay more than among other groups of media consumers. The effect of treatment

with Fox rather than MSNBC on respondents’ attitudes in the next wave is statistically

indistinguishable from zero among the group of respondents who stated a preference for

9One concern with this modeling approach is that it might overstate the effect of treatment in wave 1 onopinion in wave 2 if treatment in wave 1 also impacts the media choices of the free choice arm respondentsin wave 2. For example, if respondents treated with Fox News in wave 1 were more likely to choose Fox newsagain in wave 2, this modeling approach would overstate the persuasive effect of treatment with Fox Newsin wave 1. To avoid this concern, we can assess the decay of the effects of treatment in wave 2 on wave 3outcomes given that the third wave of our survey included no media choice or exposure. We assess whetherthe size of these lagged effects differs between waves in Appendix C, and our results indicate no significantheterogeneity in the size of the effect decay between waves, suggesting that this concern is unlikely to beproblematic.

13

Page 15: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Table 3: Decay of Media Persuasion Effects, All Respondents

Dependent variable:

Next Wave Attitudinal Index

Entertainment 0.002(0.010)

Fox 0.023∗∗

(0.010)

Constant 0.370∗∗∗

(0.007)

Observations 3,199

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

entertainment. In comparison, the effects of treatment on respondents’ attitudes in the

next wave among those who prefer MSNBC and those who prefer Fox are more durable.

Respondents preferred either partisan news option and who read Fox rather than MSNBC

reported attitudes that were still more conservative in the following wave. In both cases

these effects are statistically significant and their size is on par with the immediate effects

of partisan media.

14

Page 16: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Table 4: Decay of Media Persuasion Effects, Media Preference Subgroups

Prefer:Entertainment Fox MSNBC

(1) (2) (3)

Entertainment −0.014 −0.001 0.023(0.017) (0.015) (0.017)

Fox −0.00004 0.032∗∗ 0.040∗∗

(0.017) (0.016) (0.017)

Constant 0.362∗∗∗ 0.436∗∗∗ 0.285∗∗∗

(0.012) (0.011) (0.012)

Observations 931 1,316 953

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

Cumulative Persuasion Effects

We next assess whether the impact of partisan media accumulates after multiple treatments

one week apart. We first examine the treatment effect of one extreme relative to another:

treatment with Fox in both waves rather than MSNBC in both waves. We plot these effects

of double treatment in Figure 4 along the vertical axis, disaggregated by respondents’ stated

preferences along the horizontal axis, using only respondents who were in the forced choice

arm of the experiment in both waves.

These results indicate little added impact of a second treatment on attitudes, and in fact

perhaps some regression to the mean after two treatments. Across all preference subgroups of

respondents, there are no statistically significant effects of treatment with Fox twice relative

to MSNBC twice. This contrasts with the results from the previous section, indicating that

the effects from one wave of treatment on attitudes are durable from one wave to the next.

The results in this section indicate that, following a second (forced) treatment with partisan

media, opinions may return to previously held viewpoints. The lack of differences between

groups treated twice with oppositional partisan media is evidence of a limit on attitudinal

15

Page 17: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

●●

Attitudinal index W2 Sharing index W2

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.30

(Moreliberal)

−0.20

−0.10

0.00

0.10

(Moreconservative)

0.20

−0.30

(Lesswillingnessto share)

−0.20

−0.10

0.00

0.10

(Greaterwillingnessto share)

0.20

Trea

tmen

t effe

ct o

f rea

ding

Fox

twic

e ra

ther

than

MS

NB

C tw

ice

Figure 4: Effect of Wave 1 and Wave 2 Treatment on Wave 2 Outcomes

polarization from repeated exposure.

How does this compare to people who received only one “dose” of partisan media across

the two waves? In Figure 5 we plot the treatment effect on attitudes, comparing the group

of people assigned to read Fox once (and entertainment media in the other wave) to the

group assigned to read MSNBC once (and entertainment media in the other wave). These

effects, in contrast, indicate a small persuasive effect on all subgroups of respondents, though

these effects are not statistically significant. This further corroborates the interpretation of

the previous effects as evidence that one dose of partisan media may polarize attitudes, but

additional exposure does not further polarize opinions.

To examine the accumulation of partisan media’s persuasive effects more systematically,

we assess whether our treatment effects of one exposure to partisan media differ from our

treatment effects from multiple exposures in a pooled format. In Appendix D, we present

the results of pooled analyses of cumulative effects. These analyses suggest little evidence

of cumulative persuasion effects. This suggests that there is little additional effect from

16

Page 18: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Attitudinal index W2 Sharing index W2

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.10

(Moreliberal)

−0.05

0.00

0.05

(Moreconservative)

0.10

−0.10

(Lesswillingnessto share)

−0.05

0.00

0.05

(Greaterwillingnessto share)

0.10

Trea

tmen

t effe

ct o

f rea

ding

Ent

erta

inm

ent +

Fox

once

rat

her

than

Ent

erta

inm

ent +

MS

NB

C o

nce

Figure 5: Effect of Wave 1 and Wave 2 Treatment on Wave 2 Outcomes

multiple treatments with the same partisan media.10

10Interestingly, respondents who were treated with MSNBC twice reported more conservative opinions rel-ative to the baseline of respondents treated with MSNBC just once, although this trend falls short ofstatistical significance.

17

Page 19: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Polarization of Future Media Choices

Finally, we assess how partisan media affects future media choices. We find evidence suggest-

ing an impact of treatment with partisan media on future media consumption choices, but

it is limited to the segment of people who would not ordinarily consume partisan political

news. Tables 5 and 6 show the distribution of wave 2 media choices by the article read in

wave 1. Table 5 limits the results to respondents in the forced choice condition in wave 1

while Table 6 shows the results for respondents in the free choice condition in wave 1. Both

tables are limited to the respondents in the free choice condition in wave 2 and the outcome

is the actual article chosen and read by respondents rather than stated preferences. Because

wave 1 treatment was randomized for the group in Table 5, the values in the table repre-

sent unbiased estimates of the effect of partisan media consumption on future media choice.

In contrast, the results in Table 6 are more likely to reflect preexisting preferences among

the consumers of each kind of partisan media (in other words, how enduring the preference

people have for Fox News/MSNBC/Entertainment media might be one week later).

Wave2 ChoiceMSNBC Entertainment Fox

MSNBC 0.32 0.27 0.41Wave 1 Forced Article Entertainment 0.29 0.33 0.38

Fox 0.30 0.32 0.37

Table 5: Fraction of Wave 2 Free Choice Respondents Choosing Each Media Type by RandomWave 1 Treatment

The results in Table 5 suggest a small effect of partisan media exposure on future media

choices. Among respondents given their choice of media in wave 2, Fox News is the most

popular choice. Curiously, a larger fraction of respondents assigned MSNBC in wave 1

consume Fox News in wave 2 (41%) compared to those assigned Fox in wave 1 do (37%).

Overall, respondents only choose to read the same media in wave 2 they were assigned in

wave 1 32% of the time, suggesting no tendency to consume the same media in both waves.

The chi-squared test fails to reject the null hypotheses that the article assigned in wave 1 is

18

Page 20: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

independent from the article read in wave 2 (p ≈ 0.4). This provides evidence that across

the entire sample, partisan media exposure did not influence future choices.

We present these results graphically in Figure 6, with respondents’ assigned media in

the first wave on the left and their chosen media in the second wave on the right, using

respondents who were in the forced choice arm in wave 1 and the free choice arm in wave

2. The color of each alluvium in the plot corresponds to their chosen media in the second

wave of the survey, and comparing the size of the alluvia of the same color between different

forced choice groups in wave 1 on the left represents the treatment effect of media exposure.

MS

NB

CF

ox

En

tert

ain

me

nt

MS

NB

CF

ox

En

tert

ain

me

nt

0

300

600

900

1200

W1 Media Assigned W2 Choice

Media Read, Forced Choice W1/Free Choice W2

Fre

q

Figure 6: Forced Choice W1 Respondents’ Future Choices

Wave2 ChoiceMSNBC Entertainment Fox

MSNBC 0.73 0.16 0.11Article Chosen Wave 1 Entertainment 0.10 0.73 0.16

Fox 0.08 0.14 0.78

Table 6: Fraction of Wave 2 Free Choice Respondents Choosing Each Media Type by Wave1 Article Free Choice

In contrast, respondents in the free choice arm of our experiment in both waves indicated

an overwhelmingly tendency to consume the same media in both waves, which we show in

19

Page 21: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Table 6. This trend was strongest among respondents who consumed Fox News in wave 1.

Of this group, 78% of respondents chose to read Fox News in wave 2 again. Similarly, 73%

of respondents who chose to read MSNBC in wave 1 again chose to read MSNBC in wave

2 and 73% of respondents who chose to read entertainment in wave 1 also chose to read

entertainment in wave 2. We show these patterns in visual form in Figure 7, using only

respondents who were in the free choice arm in both wave 1 and wave 2, with the color of

the alluvia corresponding to the choice of media in the first wave. The small defection rates

are demonstrated by the lack of large alluvia moving from any wave 1 choice group to a

different wave 2 choice group. This gives some evidence that media choices remain relatively

stable over the period of a week in a free choice scenario.

MS

NB

CF

ox

En

tert

ain

me

nt

MS

NB

CF

ox

En

tert

ain

me

nt

0

300

600

900

1200

W1 Choice W2 Choice

Revealed Media Preference, Free Choice Arm

Fre

q

Figure 7: Free Choice Respondents’ Choices Over Time

Although consumers of partisan media are generally consistent in their media choices,

there is still a non-negligible rate of defection. 11% of respondents that chose MSNBC in

wave 1 chose to read Fox News in wave 2 while 8% of respondents that read Fox News in wave

1 chose to read MSNBC in wave 2. This suggests that people choose to expose themselves

to some amount of counter-attitudinal messaging, even if it is less common than continuing

to choose their preferred media. Together, these findings suggest that while consumers of

20

Page 22: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

partisan media have a consistent preference for their preferred media, this tendency is not

driven by the persuasive effect of partisan media.

We examine the persuasive impact on future choices in a more systematic fashion in

Table 7. We use a multinomial logit model to assess the impact of wave 1 media treatment

on wave 2 media choice, using MSNBC as the baseline choice and showing the effect on

choosing Entertainment rather than MSNBC in column 1 and Fox rather than MSNBC in

column 2. This analysis is limited to respondents who were in the forced choice group in

wave 1, but the free choice group in wave 2. This table reveals no significant effects of wave

1 media treatment on wave 2 media choice, corroborating the trends revealed in the earlier

tables. The positive coefficients of treatment with either entertainment or Fox in the first

column indicate that they are slightly more likely to choose entertainment in wave 2, and

the small negative coefficients in the second column indicate they are less likely to choose

Fox. All of these effects on future media choice across the entire sample are statistically

indistinguishable from zero, however.

Table 7: Impact of W1 Media Assigned on W2 Media Choice

Dependent variable:

Entertainment Fox

(1) (2)

Entertainment 0.287 −0.005(0.189) (0.177)

Fox 0.223 −0.067(0.186) (0.174)

Constant −0.160 0.266∗∗

(0.137) (0.124)

Akaike Inf. Crit. 2,515.860 2,515.860

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01

However, these overall null effects across the entire sample belie some group-specific

21

Page 23: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

effects on future choices. Figure 8 shows the effects of treatment with partisan media on

choice in the next wave of the survey, disaggregated by respondents’ stated media preferences.

Each point represents the treatment effect of reading Fox rather than MSNBC in wave 1

on the probability of choosing a given media option in the free choice arm of wave 2, for

entertainment (in the left panel), Fox (in the middle), or MSNBC (on the right), divided up

by stated preference groups along the horizontal axis.

ChooseEntertainment, W2

ChooseFox, W2

ChooseMSNBC, W2

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

-0.20

(Lesslikely to choose)

-0.10

0.00

0.10

(Morelikely to choose)

0.20

Tre

atm

en

t e

ffe

ct o

f re

ad

ing

Fo

x r

ath

er

tha

n M

SN

BC

in W

1

Figure 8: Effect of Wave 1 Treatment on Wave 2 Choices

These plots show that, for the most part, there is little effect of consuming partisan media

on future behavior, as measured by respondents’ actual choices of media when they were

allowed to select from these options. The only group of media consumers on which partisan

media seems to influence future choices is those respondents who stated a preference for

entertainment. Among these respondents, those who were assigned to read Fox in wave 1

were 12 percentage points less likely to choose Fox in wave 2 when given the choice than

those respondents who were assigned to read MSNBC in wave 1. In other words, this group

of people who would not ordinarily expose themselves to partisan media actively chose to re-

balance their media diets after a single exposure to partisan media. Among these consumers,

exposure to Fox caused them to avoid this choice in the future, relative to those who were

22

Page 24: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

exposed to MSNBC, who were more likely to choose Fox in the future.

These effects on actual choices did not seem to extend to stated preferences in either

wave 2 or wave 3, however. Figures 9 and 10 show the effects of treatment with Fox relative

to MSNBC in wave 1 or wave 2 on respondents’ stated preferences in the following wave.

Across all subgroups in both waves, treatment with partisan media did not influence future

stated preferences.

PreferEntertainment, W2

PreferFox, W2

PreferMSNBC, W2

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

-0.15

(Lesslikely to prefer)

-0.10

-0.05

0.00

0.05

(Morelikely to prefer)

0.10

Tre

atm

en

t e

ffe

ct o

f re

ad

ing

Fo

x r

ath

er

tha

n M

SN

BC

in W

1

Figure 9: Effect of Wave 1 Treatment on Wave 2 Stated Preferences

23

Page 25: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

PreferEntertainment, W3

PreferFox, W3

PreferMSNBC, W3

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

W1 PreferEntertainment

W1 PreferFox

W1 PreferMSNBC

-0.15

(Lesslikely to prefer)

-0.10

-0.05

0.00

0.05

(Morelikely to prefer)

0.10

Tre

atm

en

t e

ffe

ct o

f re

ad

ing

Fo

x r

ath

er

tha

n M

SN

BC

in W

2

Figure 10: Effect of Wave 2 Treatment on Wave 3 Stated Preferences

Conclusion

The widespread proliferation of partisan political news media has led to worries about its

pernicious effects on polarization and the resulting problems for democracy. Recent research

has indicated that partisan media can potentially polarize consumers of the information that

they broadcast. Yet we have had little evidence of the dynamic long-term impact of partisan

news on political attitudes and choices. Understanding this long-term impact is crucial

for diagnosing potential consequences for both public opinion and democratic deliberation

(Gaines, Kuklinski, and Quirk, 2007).

In this paper, we provide the first examination of the long-term persuasive impact of

partisan media on policy attitudes that evaluates these effects within the context of individ-

uals’ choices over media options. We use a PICA experimental design (de Benedictis-Kessner

et al., 2019; Knox et al., 2019) adapted for a three-wave opinion survey that allows us to

disaggregate both our short-term and long-term estimates of persuasion by the preferences

of media consumers among different news options. Our results indicate both the extent and

limits of partisan media’s effects over time. While repeated exposure to partisan media has

24

Page 26: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

limited cumulative effects on all types of consumers, the persistence of persuasion – even

from just one instance of experimental exposure to partisan media – suggests a long-term

impact of partisan media on public opinion.

Our results showing little accumulation of partisan media’s effects on political attitudes

support theories of political communication’s long-term influences on public opinion. The

fact that there is little additive effect from multiple instances of exposure to arguments

contained in partisan news stories suggests that there are limits to the political polarization

that can be attributed to partisan news. Pundits and politicians have decried the pernicious

effects of “filter bubbles” or “information silos” resulting from consumers’ ability to self-

select into news diets that only exposure them to slanted news from one side (Pariser, 2012;

Sunstein, 2001). The normative argument underlying these claims is that such one-sided

exposure to partisan news will have an impact on attitudes due to repeated reinforcement

of those arguments that subsequently drives polarization. In contrast, our results suggest

that the additive impact of repeated exposure to such news sources is no more extreme

than a single exposure to partisan news, and therefore likely is not the cause of additional

polarization among those who consume it repeatedly.

The effects of partisan news on public opinion over time are not, however, merely tem-

porary. Our findings support a theory of attitude change that is persistent after a single

exposure to partisan news media. After a week, the decay in the size of the persuasion effects

we identify is only minimal. Partisan media’s effects are not simply a short-term artifact of

the survey experimental environment, but instead have important implications for politics

and policy. The fact that the duration of persuasion is at least a full week after one instance

of experimental exposure to partisan news suggests that the media can have durable effects

on public opinion that subsequently may influence policy.

While our results contribute to a broader understanding of the relevance of media ef-

fects for democracy, they also point to several areas open for future research. For instance,

researchers should examine not just whether or not media effects decay over time, but deter-

25

Page 27: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

mine the conditions under which these effects are more or less durable (e.g. Coppock, 2017).

In addition, the dynamic impact of partisan media on people’s behavior in the real-world

deserves greater attention. Understanding both the long-term effects of media on people’s

attitudes and their behavior can help produce a more comprehensive view of partisan media’s

role in individuals’ lives and democracy.

26

Page 28: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

References

Arceneaux, Kevin, Martin Johnson, and Chad Murphy. 2012. “Polarized Political Commu-

nication, Oppositional Media Hostility, and Selective Exposure.” Journal of Politics 74(1):

174–186.

Baden, Christian, and Sophie Lecheler. 2012. “Fleeting, Fading, or Far-Reaching? A

Knowledge-Based Model of the Persistence of Framing Effects.” Communication Theory

22(4): 359–382.

Bakshy, Eytan, Solomon Messing, and Lada A. Adamic. 2015. “Exposure to ideologically

diverse news and opinion on Facebook.” Science 348(6239): 1130–1132.

Brundidge, Jennifer. 2010. “Encountering “Difference” in the Contemporary Public Sphere:

The Contribution of the Internet to the Heterogeneity of Political Discussion Networks.”

Journal of Communication 60(4): 680–700.

Bullock, John G. 2011. “Elite Influence on Public Opinion in an Informed Electorate.”

American Political Science Review 105(3): 496–515.

Cacioppo, John T, and Richard E Petty. 1979. “Effects of Message Repetition and Position on

Cognitive Response, Recall, and Persuasion.” Journal of Personality and Social Psychology

37(1): 97–109.

Cacioppo, John T, and Richard E Petty. 1989. “Effects of Message Repetition on Argument

Processing, Recall, and Persuasion.” Basic and Applied Social Psychology 10(1): 3–12.

Capon, Noel, and James Hulbert. 1973. “The Sleeper Effect—an Awakening.” Public Opinion

Quarterly 37(3): 333–358.

Chong, Dennis, and James N Druckman. 2010. “Dynamic Public Opinion: Communication

Effects over Time.” American Political Science Review 104(4): 663–680.

27

Page 29: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Coppock, Alexander. 2017. “The Persistence of Survey Experimental Treatment Ef-

fects.” Working paper. Online: https://alexandercoppock.com/papers/Coppock_

persistence.pdf.

Coppock, Alexander, Emily Ekins, David Kirby et al. 2018. “The Long-Lasting Effects

of Newspaper Op-Eds on Public Opinion.” Quarterly Journal of Political Science 13(1):

59–87.

de Benedictis-Kessner, Justin, Matthew A. Baum, Adam J. Berinsky, and Teppei Yamamoto.

2019. “Persuading the Enemy: Estimating the Persuasive Effects of Partisan Media with

the Preference-Incorporating Choice and Assignment Design.” American Political Science

Review 113(4): 902–916.

Dilliplane, Susanna. 2014. “Activation, Conversion, or Reinforcement? The Impact of Parti-

san News Exposure on Vote Choice.” American Journal of Political Science 58(1): 79–94.

Dowling, Conor M, Michael Henderson, and Michael G Miller. 2019. “Knowledge Persists,

Opinions Drift: Learning and Opinion Change in a Three-Wave Panel Experiment.” Amer-

ican Politics Research (forthcoming).

Druckman, James N, and Thomas J Leeper. 2012. “Learning More from Political Communi-

cation Experiments: Pretreatment and Its Effects.” American Journal of Political Science

56(4): 875–896.

Feldman, Lauren. 2011. “The Opinion Factor: The Effects of Opinionated News on Infor-

mation Processing and Attitude Change.” Political Communication 28(2): 163–181.

Gaines, Brian J., and James H. Kuklinski. 2011. “Experimental Estimation of Heterogeneous

Treatment Effects Related to Self-Selection.” American Journal of Political Science 55(3):

724–736.

28

Page 30: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Gaines, Brian J., James H. Kuklinski, and Paul J. Quirk. 2007. “The Logic of the Survey

Experiment Reexamined.” Political Analysis 15: 1–20.

Gillig, Paulette M, and Anthony G Greenwald. 1974. “Is it Time to Lay the Sleeper Effect

to Rest?” Journal of Personality and Social Psychology 29(1): 132.

Groeling, Tim. 2013. “Media Bias by the Numbers: Challenges and Opportunities in the

Empirical Study of Partisan News.” Annual Review of Political Science 16: 129–151.

Guess, Andy. 2018. “(Almost) Everything in Moderation: New Evidence on Americans’ On-

line Media Diets.” Working paper. Online: https://webspace.princeton.edu/users/

aguess/Guess_OnlineMediaDiets.pdf.

Hill, Seth J, James Lo, Lynn Vavreck, and John Zaller. 2013. “How Quickly We Forget:

The Duration of Persuasion Effects from Mass Communication.” Political Communication

30(4): 521–547.

Hovland, Carl I, and Walter Weiss. 1951. “The Influence of Source Credibility on Commu-

nication Effectiveness.” Public Opinion Quarterly 15(4): 635–650.

Hovland, Carl I, Arthur A Lumsdaine, and Fred D Sheffield. 1949. Experiments on Mass

Communication. Princeton, NJ: Princeton University Press.

Huber, Gregory A, and Kevin Arceneaux. 2007. “Identifying the Persuasive Effects of Pres-

idential Advertising.” American Journal of Political Science 51(4): 957–977.

Iyengar, Shanto, and Kyu S. Hahn. 2009. “Red Media, Blue Media: Evidence of Ideological

Selectivity in Media Use.” Journal of Communication 59(1): 19–39.

Jensen, Jakob D, Jennifer K Bernat, Kari M Wilson, and Julie Goonewardene. 2011. “The

Delay Hypothesis: The Manifestation of Media Effects over Time.” Human Communica-

tion Research 37(4): 509–528.

29

Page 31: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Kalmoe, Nathan P, Raymond J Pingree, Brian Watson, Mingxiao Sui, Joshua Darr, and

Kathleen Searles. 2019. “Crime News Effects and Democratic Accountability: Experi-

mental Evidence From Repeated Exposure in a Multiweek Online Panel.” International

Journal of Public Opinion Research 31(3): 506–527.

Knox, Dean, Teppei Yamamoto, Matthew A Baum, and Adam J Berinsky. 2019. “De-

sign, Identification, and Sensitivity Analysis for Patient Preference Trials.” Journal of the

American Statistical Association 114(528): 1532–1546.

Lazarsfeld, Paul, Bernard Berelson, and Hazel Gaudet. 1948. The People’s Choice: How the

Voter Makes up His Mind in a Presidential Campaign. New York: Columbia University

Press.

Lecheler, Sophie, and Claes H de Vreese. 2011. “Getting Real: The Duration of Framing

Effects.” Journal of Communication 61(5): 959–983.

Lecheler, Sophie, and Claes H de Vreese. 2013. “What a Difference a Day Makes? The Effects

of Repetitive and Competitive News Framing over Time.” Communication Research 40(2):

147–175.

Lecheler, Sophie, Mario Keer, Andreas RT Schuck, and Regula Hanggli. 2015. “The Effects of

Repetitive News Framing on Political Opinions over Time.” Communication Monographs

82(3): 339–358.

Leeper, Thomas J. 2017. “How Does Treatment Self-Selection Affect Inferences about Po-

litical Communication?” Journal of Experimental Political Science 4(1): 21–33.

Levendusky, Matthew. 2013a. How Partisan Media Polarize America. Chicago: University

of Chicago Press.

Levendusky, Matthew S. 2013b. “Why Do Partisan Media Polarize Viewers?” American

Journal of Political Science 57(3): 611–623.

30

Page 32: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Liu, Jiawei, ByungGu Lee, Douglas M McLeod, and Hyesun Choung. 2019. “Effects of Frame

Repetition through Cues in the Online Environment.” Mass Communication and Society

pp. 1–19.

Messing, Solomon, and Sean J Westwood. 2014. “Selective Exposure in the Age of Social

Media: Endorsements Trump Partisan Source Affiliation When Selecting News Online.”

Communication Research 41(8): 1042–1063.

Mummolo, Jonathan. 2016. “News from the Other Side: How Topic Relevance Limits the

Prevalence of Partisan Selective Exposure.” Journal of Politics 78(3): 763–773.

Pariser, Eli. 2012. Filter Bubble: Wie wir im Internet entmundigt werden. Munich: Hanser.

Sears, David O, and Jonathan L Freedman. 1967. “Selective Exposure to Information: A

Critical Review.” Public Opinion Quarterly 31(2): 194–213.

Slater, Michael D. 2004. “Operationalizing and Analyzing Exposure: The Foundation of

Media Effects Research.” Journalism & Mass Communication Quarterly 81(1): 168–183.

Stroud, Natalie J. 2011. Niche News: The Politics of News Choice. New York: Oxford

University Press.

Sunstein, Cass R. 2001. Republic.com. Princeton University Press.

Zaller, John. 1992. The Nature and Origins of Mass Opinion. Cambridge University Press.

31

Page 33: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Supplementary Appendix for

“Dynamic Persuasion: Decay and Accumulation of the

Persuasive Effects of Partisan Media”

Page 34: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

A Main Persuasion Effects by Stated Preferences

A.1 Immediate Effects, Pooled

We first pool across the forced choice groups in waves 1 and 2 to assess the immediate effectsof partisan media. These effects are described in the main text, but also presented here inTable A1.

Table A1: Media Persuasion Effects by Respondent Stated Preference

Dependent variable:

attitudinal IndexForced Free No Controls Free Demographics

(1) (2) (3)

Entertainment 0.017 0.021 0.051∗

(0.013) (0.017) (0.029)

Fox 0.032∗∗∗ 0.067∗∗∗ 0.153∗∗∗

(0.012) (0.020) (0.046)

Prefer Entertainment 0.054∗∗∗ 0.094∗∗∗ 0.108∗∗

(0.012) (0.019) (0.045)

Prefer Fox 0.132∗∗∗ 0.139∗∗∗ 0.140∗∗∗

(0.012) (0.019) (0.028)

Entertainment × Prefer Entertainment −0.009 −0.048∗ −0.078(0.018) (0.026) (0.055)

Fox × Prefer Entertainment −0.015 −0.063∗∗ −0.121∗

(0.018) (0.032) (0.067)

Entertainment × Prefer Fox −0.005 −0.027 −0.051(0.018) (0.029) (0.045)

Fox × Prefer Fox 0.007 −0.048∗ −0.125∗∗

(0.017) (0.027) (0.054)

Constant 0.288∗∗∗ 0.285∗∗∗ 0.277∗∗∗

(0.009) (0.006) (0.010)

Observations 5,169 5,220 5,120R2 0.068 0.082 0.088Adjusted R2 0.066 0.081 0.087Residual Std. Error 0.216 (df = 5160) 0.212 (df = 5211) 0.562 (df = 5111)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01Standard Errors Clustered by Respondent

A.2 Immediate Effects, Disaggregated into Wave 1 and Wave 2

When we disaggregate the immediate effects of partisan media into the two different waves,we see slightly different effects among entertainment-preferrers in the first wave than in thesecond wave.

A-1

Page 35: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Attitudinal Index Sharing index

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

-0.10

(Moreliberal)

-0.05

0.00

(Moreconservative)

0.05

-0.10

(Lesswillingnessto share)

-0.05

0.00

(Greaterwillingnessto share)

0.05

Tre

atm

ent

effe

ct o

f w

atch

ing

Fox

rat

her

than

MSN

BC

Figure A1: Persuasion Results, Wave 1

Attitudinal index W2 Sharing index W2

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.15

(Moreliberal)

−0.10

−0.05

0.00

0.05

(Moreconservative)

0.10

−0.15

(Lesswillingnessto share)

−0.10

−0.05

0.00

0.05

(Greaterwillingnessto share)

0.10

Trea

tmen

t effe

ct o

f rea

ding

Fox

rat

her

than

MS

NB

C in

W2

Figure A2: Persuasion Results, Wave 2

A-2

Page 36: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

B Persuasion Results Continuous Treatment

To examine the effects of persuasion in another way in order to gain statistical power relativeto models that were fully disaggregated, in this section we also fit a model that makes anassumption about the symmetrical effects of MSNBC and Fox News – that is, Fox Newshas an effect on attitudes that is the opposite of the effect due to MSNBC. Formally, thisassumption defines the causal effect of partisan media τ as:

τ =E(Attitudinal Index|Fox)− E(Attitudinal Index|Read Entertainment) = (5)

− (E(Attitudinal Index|MSNBC)− E(Attitudinal Index|Read Entertainment)) (6)

τ can therefore be estimated by defining a continuous treatment variable, D, that is 1 fortreatment with Fox News, −1 for treatment with MSNBC, and 0 for Entertainment andfitting the model:

Attitudinal Index = α + τD + ε (7)

By pooling in this manner and assuming symmetry in the effects of the two news outlets,we gain additional power to detect the accumulation of persuasion effects.

Table A2 presents results for the persuasive effects of partisan media using this continuoustreatment indicator. The columns of Table A2 correspond to the wave of data used for themodel. Columns 1 and 2 correspond to models run on data from waves 1 and 2, respectively,while column 3 shows the results from a model pooling over both waves.

These results suggest that partisan media has an effect on respondents’ attitudes ofapproximately 0.015 along the unit scale — in the more conservative direction for Fox, andin the more liberal direction for MSNBC. Compared with the previous results that did notassume symmetry of persuasion effects, the pooled model estimate of .015 is substantivelysimilar to the previous estimates of -.013 for treatment with MSNBC and .018 for treatmentwith Fox (both relative to entertainment). The F test suggests the model which assumessymmetric persuasive effects for Fox and MSNBC actually significantly improves model fitrelative to the model which does not. Similarly, the F-test also suggests that a model whichallows treatment effects to vary between waves does not improve model fit. This demonstratesthat the assumption of symmetric persuasive effects for Fox and MSNBC is plausible.

A-3

Page 37: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Table A2: Main Results with Continuous Treatment

Dependent variable:

Attitudinal IndexWave 1 Wave 2 Waves 1 and 2 Pooled

(1) (2) (3)

Media Treatment 0.017∗ 0.013∗ 0.015∗

(0.005) (0.006) (0.004)

Constant 0.369∗ 0.376∗ 0.372∗

(0.004) (0.005) (0.004)

Observations 3,073 2,096 5,169R2 0.004 0.002 0.003Adjusted R2 0.004 0.002 0.003Residual Std. Error 0.220 (df = 3071) 0.228 (df = 2094) 0.223 (df = 5167)

Note: ∗p<0.05Standard Errors Clustered by Respondent In Pooled Model

A-4

Page 38: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

C Heterogeneity of Decay Effects by Wave

● ●

●●

Attitudinal index W2 Sharing index W2

PreferEntertainment

PreferFox

PreferMSNBC

PreferEntertainment

PreferFox

PreferMSNBC

−0.15

(Moreliberal)

−0.10

−0.05

0.00

0.05

(Moreconservative)

0.10

−0.15

(Lesswillingnessto share)

−0.10

−0.05

0.00

0.05

(Greaterwillingnessto share)

0.10

Trea

tmen

t effe

ct o

f rea

ding

Fox

rat

her

than

MS

NB

C in

W1

Figure A3: Persuasion of Wave 1 Treatment on Wave 2 Outcomes (Free Choice W2 only)

Attitudinal index W3

PreferEntertainment

PreferFox

PreferMSNBC

-0.15

(Moreliberal)

-0.10

-0.05

0.00

0.05

(Moreconservative)

0.10

Tre

atm

ent

effe

ct o

f rea

ding

Fox

rat

her

than

MSN

BC

in W

1

Figure A4: Persuasion of Wave 1 Treatment on Wave 3 Outcomes (Free Choice W2 only)

A-5

Page 39: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Attitudinal index W3

PreferEntertainment

PreferFox

PreferMSNBC

-0.15

(Moreliberal)

-0.10

-0.05

0.00

0.05

(Moreconservative)

0.10

Tre

atm

en

t e

ffe

ct

of

rea

din

gF

ox r

ath

er

tha

n M

SN

BC

in W

2

Figure A5: Persuasion of Wave 2 Treatment on Wave 3 Outcomes

A-6

Page 40: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

Dependent variable:

Next Wave attitudinal IndexForced Free No Controls Free Demographics

(1) (2) (3)

Entertainment 0.015 0.053∗∗∗ 0.006(0.012) (0.012) (0.019)

Fox 0.034∗∗∗ 0.129∗∗∗ 0.022(0.012) (0.011) (0.018)

Wave 3 0.017 0.010 0.014(0.011) (0.010) (0.032)

Entertainment × Wave 3 −0.012 −0.014 −0.046(0.017) (0.015) (0.043)

Fox × Wave 3 −0.015 −0.005 0.015(0.017) (0.014) (0.043)

Constant 0.359∗∗∗ 0.309∗∗∗ 0.366∗∗∗

(0.008) (0.008) (0.014)

Observations 4,156 4,092 4,066R2 0.003 0.057 0.018Adjusted R2 0.002 0.056 0.017Residual Std. Error 0.228 (df = 4150) 0.219 (df = 4086) 0.617 (df = 4060)

Note: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01Standard Errors Clustered by Respondent

A-7

Page 41: Dynamic Persuasion: Decay and Accumulation of Partisan Media … · 2020. 8. 3. · Dynamic Persuasion: Decay and Accumulation of Partisan Media Persuasion Zachary Markovich Matthew

D Cumulative Effects

Table A3: Cumulative Effects of Media Exposure

Dependent variable:

W2 Marijuana IndexForced Choice No Controls Free Choice Free Demographics

(1) (2) (3)

Fox At Least Once 0.008 0.071∗ 0.065∗

(0.026) (0.030) (0.023)

Entertainment At Least Once −0.013 0.055∗ 0.052∗

(0.026) (0.028) (0.024)

MSNBC Once −0.037 −0.028 −0.040(0.034) (0.043) (0.030)

Fox Twice −0.026 0.085∗ 0.085∗

(0.038) (0.036) (0.031)

Entertainment Twice −0.008 −0.006 −0.019(0.036) (0.039) (0.033)

Constant 0.408∗ 0.311∗ 0.321∗

(0.049) (0.050) (0.045)

Observations 1,044 1,065 1,061R2 0.005 0.100 0.086Adjusted R2 −0.0002 0.096 0.082Residual Std. Error 0.233 (df = 1038) 0.214 (df = 1059) 0.295 (df = 1055)F Statistic 0.968 (df = 5; 1038) 23.619∗ (df = 5; 1059) 19.904∗ (df = 5; 1055)

Note: ∗p<0.05

A-8