You are on page 1of 23

804500

research-article2018
JMQXXX10.1177/1077699018804500Journalism & Mass Communication QuarterlyLuo et al.

Meta-Analysis of Agenda-Setting and Parental Effects


Journalism & Mass Communication Quarterly
2019, Vol. 96(1) 150­–172
A Meta-Analysis of News © 2018 AEJMC
Article reuse guidelines:
Media’s Public Agenda-Setting sagepub.com/journals-permissions
DOI: 10.1177/1077699018804500
https://doi.org/10.1177/1077699018804500
Effects, 1972-2015 http://journals.sagepub.com/home/jmq

Yunjuan Luo1, Hansel Burley2, Alexander Moe3,


and Mingxiao Sui4

Abstract
This study used rigorous meta-analytic approaches to analyze empirical agenda-
setting studies published from 1972 to 2015, and 67 studies that met the inclusion
criteria for analysis produced a moderate grand mean effect size of 0.487. A multiple
regression analysis revealed one most notable predictor that classified the basis for
the study correlation as either content categories or survey participants. A multiple
regression of a subgroup using content categories produced homogeneity. The mean
effect size for these studies was 0.51. This is an indication of consistency in findings
across agenda-setting studies and the presence of strong news media’s public agenda-
setting effects.

Keywords
agenda setting, meta-analysis, media agenda, public agenda, news media

Since the Chapel Hill study by McCombs and Shaw (1972) set the research tradition
examining the influence of patterns of media coverage on the public’s perceived impor-
tance of issues, agenda-setting theory has become one of the most popular communica-
tion theories. In addition, the interest in agenda-setting research has continued to be
strong among scholars. Y. Kim, Kim, and Zhou (2017) found that the number of schol-
arly articles published in this field has been increasing over time, with a noticeable rise

1South China University of Technology, Guangzhou, China


2Texas Tech University, Lubbock, USA
3Coastal Carolina University, Conway, SC, USA
4Ferrum College, VA, USA

Corresponding Author:
Yunjuan Luo, School of Journalism and Communication, South China University of Technology, Higher
Education Mega Center, Panyu District, Guangzhou 510006, Guangdong, China.
Email: yunjuan.luo@yahoo.com
Luo et al. 151

since 2000. The vast majority of those studies provide extensive evidence supporting
the idea that audiences perceive issues emphasized by the news media as important.
That is, issue salience can transfer from the media agenda to the public agenda, which
is often referred to as “public agenda setting” effects (Dearing & Rogers, 1996).
During the past four decades, the exploration of agenda-setting effects has expanded
from the first level of agenda setting (the transmission of issue salience) to the second
level of agenda setting (the transmission of attribute salience). More recently, the
research approach of agenda-setting studies has been upgraded to the third level that
looks at the transfer of issues and attributes in bundles from the media network agenda
to the public network agenda (e.g., Guo, 2012, 2013; Guo & McCombs, 2011). The
agenda-setting hypothesis has been tested across various media outlets at different
levels using a wide variety of methodologies (e.g., Conway & Patterson, 2008;
Roberts, Wanta, & Dzwo, 2002; Wanta, Golan, & Lee, 2004), and agenda-setting
effects have been observed inside and outside the United States in political and nonpo-
litical settings (e.g., Dursun-Ozkanca, 2011; Ghorpade, 1986; McCombs, Lopez-
Escobar, & Llamas, 2000).
Given the large volume of agenda-setting studies published in the past few decades,
it would be valuable to summarize across a body of combined empirical evidence to
arrive at a reliable conclusion that explains the relationship between the media agenda
and the public agenda. Although meta-analysis is a unique and powerful tool for sta-
tistically synthesizing existing research outcomes, few meta-analyses have been done
in this field. A meta-analysis of empirical agenda-setting studies conducted by Wanta
and Ghanem (2007) nearly 10 years ago found an overall mean correlation of .53
between the media agenda and the public agenda. None of the methodological factors
investigated in their study were found to significantly increase or decrease agenda-
setting effects.
Built upon Wanta and Ghanem’s (2007) work, this study examines the longitudinal
evolution of news media’s public agenda-setting effects using a statistical meta-
analysis of empirical studies published in this area from 1972 to 2015. It not only looks
at the magnitude of agenda-setting effects by estimating the weighted grand mean
effect size over a combined sample of studies, but also explores the impact of some
important factors on the magnitude of effects found in agenda-setting research.

Literature Review
News Media as Agenda Setters
The news media are the primary message mediators that shape the public’s perceptions
of the world (B. C. Cohen, 1963; Lippmann, 1946). Despite various alternative
sources, news media are still the major avenue for most individuals to acquire infor-
mation about the outside environment (e.g., Prior, 2013; Protess & McCombs, 2016).
Specifically, by selecting and presenting news, the news media play an important role
in shaping public opinion (McCombs & Shaw, 1972). This conventional wisdom on
news selection sets the basis for the original agenda-setting hypothesis (Price &
152 Journalism & Mass Communication Quarterly 96(1)

Tewksbury, 1997). That is, the public’s perceptions of issue salience are largely a
result of intensive news coverage of these issues (e.g., Iyengar, 1979; Iyengar &
Kinder, 1987; McCombs & Shaw, 1972; McCombs, Shaw, & Weaver, 1997).
McCombs and Shaw (1972) conducted the initial investigation of news media’s
agenda-setting effects during the 1968 presidential election. They did a cross-sectional
survey among undecided registered voters in Chapel Hill, North Carolina, asking them
to outline the key issues of the presidential campaign. The answers from the respon-
dents were coded into five issue categories. Then they rank-ordered these categories to
construct the public agenda based on the percentage of respondents who named those
issues as important. At the same time, they collected content from news media serving
the local voters and coded it into the same five issue categories. The amount of media
coverage for those issues was then rank-ordered to compose the media agenda. A
strong correlation of .97 was found between those two issue agendas using Spearman’s
rank correlation.
Most of the later agenda-setting studies followed McCombs and Shaw’s initial
approach to correlate issue salience in media content with issue concerns of respon-
dents in a survey to explore the relationship between the media agenda and the public
agenda, with audiences’ exposure to the news media as a given (McCombs, 1981).
However, exposure to media messages had some effects on the magnitude of agenda-
setting effects displayed by individuals as well (Wanta & Ghanem, 2007). For exam-
ple, Wanta and Wu (1992) included exposure to the news media measured by the
frequency of media use as an independent variable in their study. They analyzed the
content of four news outlets (e.g., Southern Illinoisan) and designated five issues
receiving the most coverage as media issues. They also conducted a telephone survey
of 341 residents in Jackson County, Illinois, where they asked the respondents how
much they were concerned with each media issue. The scores on those five issues were
then summed to form a media issue concern index for each individual, which became
a measure of the perceived importance of media issues. The result of Pearson’s prod-
uct–moment correlation test indicated a significant and positive relationship (r = .31,
p < .001) between exposure to the news media and the media issue concern index,
suggesting that higher exposure to media messages could lead to strong agenda-setting
effects.

Moderators of agenda-setting effects. The initial 1968 Chapel Hill study (McCombs &
Shaw, 1972) and other earlier work concentrated on a direct relationship between the
media agenda and the public agenda. Increasingly, agenda-setting effects are found to
be conditional on many factors such as the obtrusiveness of issue (e.g., Demers, Craff,
Choi, & Pessin, 1989; Hester & Gibson, 2007), frequency of media use (e.g., Iyengar,
1979), need for orientation (e.g., Weaver, 1980), and political leanings (e.g., Muddi-
man, Stroud, & McCombs, 2014). Given a wide range of methodologies used in
agenda-setting research, Wanta and Ghanem (2007) examined the impact of some
methodological factors on the magnitude of agenda-setting effects, including the num-
ber of issues examined (single issue vs. issue sets), the unit of analysis under investiga-
tion (individual vs. aggregate data), the variable under investigation (media content vs.
Luo et al. 153

media exposure), and the time frame under investigation (cross-sectional survey vs.
longitudinal examination). They found none of those methodological artifacts matter
in agenda-setting research. This study will reexamine selected factors and explore
other moderators (i.e., characteristics of studies), including research approach, media
type, time lag, and place of study, which could potentially influence the magnitude of
effects found in agenda-setting research.

Issue versus attribute agenda setting. The original research approach of agenda-set-
ting theory emphasizes the news media’s capacity to transfer issue salience to the
public (McCombs & Shaw, 1972). While news coverage of issues matters, an equally
important question is whether and how differentiated presentations of an issue affect
how the public thinks about this issue (Wanta et al., 2004). The second level of agenda
setting thus focuses on attribute salience rather than issue salience, with the assump-
tion that “the attributes linked to the object in the news media are mentally linked to
the object by the public” (Wanta et al., 2004, p. 367).
Benton and Frazier’s (1976) study first extended agenda-setting research to the
second level, revealing that the news media did affect the public’s awareness of pro-
posed solutions and specific knowledge of policy proposals. This is consistent with
Becker and McCombs’s (1978) finding that in addition to setting the public agenda,
the news media also affected certain aspects of the public’s opinion of presidential
candidates by emphasizing some cognitive components of candidate image. Weaver,
Graber, McCombs and Eyal (1981) then introduced these issue-related objects (i.e.,
personal concerns or stereotyping) as an additional layer of agenda setting. Following
this framework, a majority of empirical studies have found support for this attribute
agenda-setting effect (e.g., Craft & Wanta, 2004; Wanta et al., 2004), where the attri-
butes individuals link to an object (e.g., candidate) are largely a function of news
coverage (e.g., Golan & Wanta, 2001; King, 1997).
A more notable finding is that attribute agenda setting may have a stronger effect
on the public than issue agenda setting. Wu and Coleman (2009) grounded their specu-
lation on the hierarchy of effects theory, summarized as feelings about attributes (sec-
ond level) have a stronger impact than cognition about issues (first level). In their
study of the 2004 U.S. presidential election, they indeed found the correlation between
the media agenda and the voters’ agenda was smaller at the first level (r = .516) than
at the second level (r = .668). Even though this difference only held for Kerry and not
for Bush, it still suggests a possibility that second-level attributes exert a stronger
agenda-setting effect on the public than does the salience of issues (Wu & Coleman,
2009). This is consistent with previous findings that attribute agenda-setting studies
often revealed higher correlations than the issue agenda-setting studies did (e.g.,
Lopez-Escobar, Llamas, & McCombs, 1998).

Agenda setting across news media. News media are a key component of agenda-
setting effect. Now spanning more than four decades, empirical studies have examined
across a variety of news media, including newspapers, magazines (e.g., McCombs &
Shaw, 1972), television (e.g., Iyengar, 1979; Iyengar & Kinder, 1987), online media
154 Journalism & Mass Communication Quarterly 96(1)

(Roberts et al., 2002), and social media (Russell Neuman, Guggenheim, Mo Jang,
& Bae, 2014). A primary reason one would speculate agenda-setting effects to vary
across different media platforms is their distinctive capacity in carrying important
problems in society (Chaffee & Metzger, 2001). Of all news disseminators, news-
papers often provide more space to cover substantive news issues than television or
magazines (McCombs & Shaw, 1972). While the online media (i.e., electronic bulletin
boards or newspapers’ online versions) are more advantageous in offering intensive
coverage of various issues, they also allow for a high flexibility for selective exposure
(Bennett & Iyengar, 2008), which may inhibit their capacity in drawing the public’s
attention to important issues (Althaus & Tewksbury, 2002).
Empirical examination reveals mixed findings, however. In general, newspapers
are found to exert stronger agenda-setting effects than television news (e.g., McClure
& Patterson, 1976; McCombs & Shaw, 1972), as newspapers often have more space
for timely issues (McCombs & Shaw, 1972) and perform better in drawing public
attention (Strömbäck & Kiousis, 2010). However, this conditional influence of media
channels on agenda-setting effects may not hold in nations other than the United
States. In their examination of agenda-setting effects during the 2006 Swedish parlia-
mentary election, Strömbäck and Kiousis (2010) found no consistent evidence of dif-
ferences related to media channels (television, print, and radio). Hence, it is still an
open question whether and how the effect of media agendas on public agendas is
conditional on media channels—newspaper, television, radio, and online media.

Time lag in agenda-setting research. Another essential consideration of agenda-


setting function is time lag, also called time frame in many agenda-setting studies.
In general, “time-lag selection is important because it demonstrates the time-varying
causal effects” (Roberts et al., 2002, p. 454). A conventional consensus is that a long-
term frame is required for the media coverage to have the most impact on the public
agenda (Pralle, 2006). This is largely because repeated exposure to the media agenda
reinforces the salience of both issues and attributes for the public (Iyengar & Kinder,
1987). However, scholars divide greatly in defining the “most optimal effect span.”
Some suggest an optimal time frame of 1 week to 2 weeks for network newscasts, 3
to 4 weeks for newspapers, and 8 weeks for national news magazines (Wanta & Hu,
1994). Other studies suggest it may take 2 to 6 months (Stone & McCombs, 1981) or
3 to 4 months (Shoemaker, Wanta, & Leggett, 1989) for the media agenda to transfer
to the public agenda.
Overall, existing empirical examinations have employed a wide range of time
frames from several days (e.g., Roberts et al., 2002) to months (e.g., Wanta et al.,
2004), with the effect of agenda setting varying depending on each study’s time-lag
selection. Roberts and her colleagues (2002) examined agenda-setting effects by five
media outlets on four issues, revealing that a time lag of 7 days produced the largest
number of significant correlations relative to time lags of 1 day to 6 days. Wanta et al.
(2004) also suggest a slightly larger agenda-setting effect with a time lag of 6 months
than that of 3 months. However, it is notable that a moderate length of time lag may be
preferable in some conditions. While long-term media coverage often reinforces the
Luo et al. 155

public agenda, the agenda-setting effects of some issues are actually determined by
“the status of chronic accessibility in the minds of the public, whereby temporary ups
and downs in media salience have little impact on perceived importance” (Shehata &
Strömbäck, 2013, p. 250). As such, while it often takes a long time for the news media
to set agendas for the public successfully, this phenomenon does not function neces-
sarily as “the longer, the better.”

Agenda-setting research inside and outside the United States. Since the initial 1968
Chapel Hill study (McCombs & Shaw, 1972), the news media’s effects in establishing
issue and attribute salience for the public have been observed in a variety of nations
such as the United States (e.g., Iyengar, 1979; Iyengar & Kinder, 1987; Tan & Weaver,
2007, 2013; Wanta et al., 2004), Britain (e.g., Dursun-Ozkanca, 2011), Spain (e.g.,
McCombs et al., 2000), and China (e.g. Luo, 2014). If country characteristics serve
as contingent conditions of agenda-setting effect (Peter, 2003), it is rational to suspect
variations in the news media’s agenda-setting function across nations.
The above assumption that the agenda-setting function varies across nations rests
on several well-established conventions. First, media systems greatly vary across
nations. This is because the development of news media is rooted in a country’s
broader political, cultural, and economic structure (Hallin & Mancini, 2004).
Specifically, some well-examined political factors include the type of political
regime, the state–citizen relationship, the media’s connection to government inter-
ests, and the state’s intervention in newsroom. Cultural factors often relate to a
nation’s cultural values, such as individualism versus collectivism. Regarding eco-
nomic factors, they often include a nation’s structure of media markets and the
development of advertising industry (e.g., Albæk, van Dalen, Jebril, & de Vreese,
2014; Hallin & Mancini, 2004; Hanitzsch et al., 2011; Mellado & Lagos, 2013).
Because of such factors, journalists across nations often diverge in their journalistic
perceptions, some concentrating on news while others focusing more on entertain-
ment (Albæk et al., 2014). Presumably, news media with emphasis on substantive
news are more likely to establish the media agenda and thus influence the public
agenda than those entertainment-oriented media. This variation not only exists
between individual media outlets, but can also be found among nations and coun-
tries in which the media systems greatly differ (Iyengar et al., 2010). Second, news
media across nations also differ in their press freedom and media credibility
(Freedom House, 2015). As media credibility often mediates media effects (e.g.,
Buturoiu & Corbu, 2015), it is likely that media credibility will influence the news
media’s function in transferring the media agenda to the public agenda.

Research Questions
To explore whether and how the news media’s agenda-setting function has evolved
over time, this study uses a meta-analysis approach to estimate a statistical average of
existing empirical studies in the past four decades to get an updated estimate of the
relationship between the media agenda and the public agenda. It will also examine
156 Journalism & Mass Communication Quarterly 96(1)

what factors could potentially influence the magnitude of agenda-setting effects.


Therefore, the following research questions are proposed:

RQ1: What is the average effect size of agenda-setting effects found in existing
studies?
RQ2: Which factors are related to the variability in effect size?

Method
This meta-analysis study used techniques suggested by Hedges and Olkin (1985),
Lipsey and Wilson (2001), and Borenstein, Hedges, Higgins, and Rothstein (2009).
SPSS (v.22) script was employed for production of homogeneity (Z and Q) and other
statistics. Wilson’s (2005) SPSS script was used for weighted linear regression.
According to Lipsey and Wilson (2001), a meta-analysis has six steps: (a) identify a
reason for the study, (b) locate all possible studies that fit reason, (c) select those stud-
ies that fit within the preestablished criteria, (d) code these studies based upon a code
sheet into a database, (e) calculate values that can be used in meta analyses (e.g.,
Fisher’s Z, inverse variance weight), and (f) conduct the statistical analyses testing for
homogeneity, significance, and subgroup effects.

Literature Search
Two databases were used for literature search, including Academic Research Complete
and Communication Source. Although those two databases may not provide an exhaus-
tive list of agenda-setting studies, they are two major databases that are mostly con-
sulted in communication research. Peer-reviewed full-text English journal articles
published from 1972 to 2015 with the keywords “agenda setting” in the subject terms
or in the title were retrieved from the databases. Dissertations, conference papers, and
manuscripts published in other languages were not included in the sample.
Initial search generated a total of 436 agenda-setting articles, but for an article to be
included in this study, it also had to meet the following characteristics based on the
focus of the study and statistical requirements for meta-analysis:

1. The study must include an empirical test of agenda-setting hypothesis and


present a relationship between the media agenda and the public agenda because
the focus of this study is the news media’s agenda-setting effects on the public
agenda, aka “public agenda setting” (Dearing & Rogers, 1996). Therefore,
those studies focusing on who set up the media agenda (e.g., intermedia agenda
setting) and who set up the policy agenda were excluded in this meta-analysis.
In this way, the sample size was significantly reduced mainly because “agenda
setting” used as the keywords for the initial search was very general.
2. The study should employ correlation statistics, either Pearson’s r or Spearman’s
rho, two most commonly used data analysis strategies in agenda-setting
research. The correlation coefficient r was recommended as the most useful
Luo et al. 157

measure of effect size in meta-analysis (Hunter & Schmidt, 1990). Consistent


with most other meta-analysis literature (e.g., Benoit, Hansen, & Verser, 2003;
Hunter & Schmidt, 1990; Wanta & Ghanem, 2007), studies that employed
regression were not included in the sample unless they also provided zero-
order correlations for the media–public relationship. We excluded a total of 36
studies that used regression analysis without providing zero-order correlations
but met all the other inclusion criteria. While researchers could have converted
beta coefficient correlations using an imputation approach, the range is limited
to beta values ± .50 (Peterson & Brown, 2005). Many agenda-setting correla-
tions exceeded this threshold. In addition, Peterson and Brown (2005) encour-
aged researchers to select original correlation coefficients before using imputed
values, whenever possible. Finally, our methods used study sample size as a
weight for effect size. A spurious imputation with a large sample size could
dramatically affect results. Therefore, we chose to follow a conservative route
in this report and not include studies that reported results from regression
analyses.
3. The study should test the first-level or/and second-level agenda-setting effects.
Recent studies that only tested the third-level agenda setting were excluded
because they used network analysis, a different methodological and analytical
approach compared with that used in traditional agenda-setting research.
4. The study must report enough information to compute an index of effect size
such as sample size in correlation statistics (N of analysis), Basis for N, and
correlation coefficients (r values). Basis for N is a determination of the source
of the sample size used in a study’s calculation of the correlation statistic,
either on survey participants (Basis for N = 1) or on text analysis of content
categories (Basis for N = 2). Therefore, N of analysis is determined by the
sample size of the survey or the number of categories derived from content
analysis of news articles in the study. For example, Wanta and Wu’s (1992)
study mentioned above falls into the first category (Basis for N was coded as 1,
N = 341 participants), while McCombs and Shaw’s original study is a typical
example of the second category (Basis for N was coded as 2, N = 15 content
categories).

To examine the relationship between the media agenda and the public agenda, a
total of 67 studies that remained after the above screenings were coded (Table 1 sum-
marizes the studies retained for inclusion in this meta-analysis). For comparison pur-
poses at this stage, Wanta and Ghanem (2007) used their search methods to find 90
studies; however, after their inclusion and exclusion criteria were considered, they
kept 45 studies in their analysis. These 45 remaining studies produced 90 effect sizes,
meaning that several studies reported multiple effect sizes. Borenstein et al. (2009)
discussed several ways of combining effect sizes within a study, with independence of
the studies as the defining feature of the effect sizes drawn from a study. Wanta and
Ghanem (2007) reported multiple effect sizes in studies but did not directly address
how they were treated. Boulianne (2009, 2015) treated multiple effect sizes within a
158 Journalism & Mass Communication Quarterly 96(1)

Table 1. Studies Included in the Meta-Analysis (k = 82).

Author Year Basis for N N of analysis r value Country of study


McCombs and Shaw 1972 2 15 .967 United States
McLeod et al. 1974 1 157 .051 United States
McLeod et al. 1974 1 136 .162 United States
Tipton et al. 1975 2 7 .238 United States
Weaver et al. 1975 1 421 .13 United States
Benton and Frazier 1976 2 13 .65 United States
Kaid et al. 1977 2 7 .64 United States
Palmgreen and Clarke 1977 2 55 .53 United States
Palmgreen and Clarke 1977 2 33 .82 United States
Williams and Larsen 1977 2 7 .643 United States
Williams and Larsen 1977 2 9 .652 United States
Sohn 1978 2 41 .191 United States
Swanson and Swanson 1978 2 25 .35 United States
Iyengar 1997 2 28 .456 United States
Winter and Eyal 1981 2 27 .384 United States
Culbertson and Stempel 1984 2 8 .06 United States
Sohn 1984 2 9 .68 United States
Atwater et al. 1985 2 6 .52 United States
Ghorpade 1986 2 19 .51 United States
Salwen 1987 2 7 .87 United States
Salwen 1988 2 7 .555 United States
Eaton 1989 2 13 .4 United States
Yagade and Dozier 1990 2 13 .79 United States
Brosius and Kepplinger 1992 2 53 .137 Germany
Shaw and Martin 1992 2 6 .03 United States
Wanta and Wu 1992 1 341 .31 United States
Iyengar and Simon 1993 2 8 .85 United States
Wanta and Hu 1993 2 41 .213 United States
Wanta and Hu 1994 2 11 .74 United States
Ader 1995 2 21 .169 United States
McCombs et al. 1997 2 15 .391 Spain
Dalton et al. 1998 2 13 .88 United States
Lopez-Escobar et al. 1998 2 6 .73 United States
McCombs et al. 2000 2 15 .633 Spain
Kim et al. 2002 2 6 .532 United States
Tsfati 2003 2 18 .71 United States
Tsfati 2003 2 18 .58 United States
Tsfati 2003 2 18 .13 United States
Yang and Stone 2003 2 7 .36 United States
Craft and Wanta 2004 2 8 .643 United States
Craft and Wanta 2004 1 345 .096 United States
Kiousis and McCombs 2004 2 11 .52 United States
(continued)
Luo et al. 159

Table 1. (continued)
Author Year Basis for N N of analysis r value Country of study
Wanta et al. 2004 2 26 .585 United States
Wanta et al. 2004 2 21 .372 United States
Kiousis 2005 2 5 .801 United States
Kiousis 2005 2 5 .588 United States
Kiousis 2005 2 5 .783 United States
Coleman and Banning 2006 1 1,807 .1235 United States
Kiousis et al. 2006 2 10 .723 United States
Coleman and McCombs 2007 2 5 1 United States
Coleman and McCombs 2007 2 5 1 United States
Golan et al. 2007 2 10 .81 United States
Kim and McCombs 2007 2 22 .58 United States
Tan and Weaver 2007 2 19 .278 United States
Conway and Patterson 2008 2 10 .488 United States
Conway and Patterson 2008 2 7 −.071 United States
Kiousis and Wu 2008 2 25 .352 United States
Kiousis and Wu 2008 2 25 .464 United States
Besova and Cooley 2009 2 9 .723 United States
Besova and Cooley 2009 2 9 .733 United States
Mohamed and Gunter 2009 2 10 .74 Egypt
Ragas and Roberts 2009 2 8 −.26 United States
Tan and Weaver 2009 2 9 .432 United States
Wu and Coleman 2009 2 30 .516 United States
Wu and Coleman 2009 2 9 .298 United States
Balmas and Sheafer 2010 2 8 .675 Israel
Tan and Weaver 2010 2 41 .26 United States
White and Park 2010 1 455 .025 United States
Wirth et al. 2010 2 7 .195 Switzerland
Dursun-Ozkanca 2011 2 28 .61 Britain
Dursun-Ozkanca 2011 2 9 −.242 Britain
Kiousis 2011 2 6 .556 United States
Valenzuela 2011 2 15 .275 Canada
Du 2012 2 99 .573 Multiple
countries
Kim et al. 2012 2 5 .667 South Korea
Zhang et al. 2012 2 7 .665 China
Bowe et al. 2013 2 6 −.588 United States
Camaj and Weaver 2013 2 7 .751 United States
Lee and Coleman 2014 2 8 .616 United States
Luo 2014 2 10 .845 China
Ragas et al. 2014 2 38 .803 United States
Vu et al. 2014 2 10 .846 United States
160 Journalism & Mass Communication Quarterly 96(1)

study as independent. This appears to have had the effect of raising the number effect
sizes in these studies, while potentially confounding results by including samples
more than once. Our default stance was that independent subgroups within a study
contribute independent information to the grand mean effect size, so we treated these
subgroups as independent studies. However, many studies presented effects sizes that
were not independent, so we combined the summary information from these to pro-
duce one effect size for the study. This approach is consistent with all weighted meta-
analysis procedures since Hedges and Olkin (1985) introduced the approach (Card,
2012). As we included effect sizes with only independent samples, the number of
effect sizes is reduced, but those included should be unconfounded estimates of study
results.

Coding Procedure
A codebook including categories and variables was designed based on literature
review of agenda-setting research and previous meta-analysis studies. The unit of
analysis was each study, which was coded for its descriptive and critical features,
including year of publication (publication year of the study), journal title, number of
author, place of study, N of analysis, Basis for N (1 = survey participants, 2 = content
categories), r value, correlation analysis (1 = Spearman’s rho, 2 = Pearson’s r),
research approach (1 = first level, 2 = second level), media type,1 time lag, type of
content (1 = national media content, 2 = regional/local media content, 3 = both
national and regional/local media contents), type of issue (1 = issue sets, 2 = single
issue), type of survey (1 = national survey, 2 = regional/local survey), size of survey,
and study design (1 = cross-sectional, 2 = longitudinal). These studies produced 82
effect sizes.
Coding was done by two of the authors, who are very familiar with agenda-setting
theory and have done agenda-setting research before. The coders met several times to
discuss the coding process and to familiarize themselves with the codebook. Several
pretests were conducted to refine the coding process and content categories. The inter-
coder reliability test was run on 15% of the coded articles, and intercoder reliability
calculated using Krippendorff’s alpha for all the variables was above .85, except for “r
value” (.826) and “time lag” (.835).2 Both coders coded the rest of the articles and
discussed the coding afterward. In case there was a disagreement, it was resolved by
agreement so that the final coding of all the articles was agreed by both coders.

Effect Size Calculations and Statistical Analyses


The researchers used SPSS v.23 to clean the data and to run preliminary analyses and
graphics. Missing data emerged because of unreported study characteristics in articles.
Approximately, 1.5% of the studies had incomplete data, so SPSS multiple imputation
was used to replace missing data. No correlational data were missing in the final anal-
ysis. Therefore, prior to analysis, there were no missing data in the final data set.
Because of the differences in sample sizes of studies, those studies with large sample
Luo et al. 161

sizes were not treated as equal to studies with smaller sample sizes. The weight in all
analyses is defined as the inverse of the variance or w = 1 / SE 2 , where the variance,
SE2, is strongly related to sample size (Card, 2012).
The common effect size statistic for this analysis is Pearson’s product–moment cor-
relation or Spearman’s rank correlation. As the unit of analysis was each study, the
default was one effect size per study. However, as noted above, many studies pre-
sented more than one effect size, with the various effect sizes representing multiple
groups. In general, if the groups produced independent effect sizes, they were treated
as separate studies (Borenstein et al., 2009) because independent effect sizes contrib-
uted independent information to the grand mean effect size. If the groups were not
independent (e.g., the study reported an overall correlation for an entire group and then
reported correlations for subgroups based on gender), only the overall correlation was
used. If effect sizes of subgroups were presented with no overall correlation statistic,
the subgroup effect sizes were averaged (e.g., the study reported correlations for only
females and males with no overall correlation for the entire study), thus producing the
overall statistic. If the subgroups had different sample sizes, a weighted average was
used. If a study examined both the first-level and second-level agenda setting, effect
sizes were calculated separately for each level. Both significant and nonsignificant
correlation coefficients were included to avoid bias toward significant findings.
The researchers first calculated a grand mean effect size for all studies and then ran
two significance tests. One tested the magnitude of the grand mean effect size, and the
other tested the heterogeneity of effect sizes by using Hedges’s Q test (Card, 2012).
Weighted meta-regression procedures using SPSS matrix script developed by Wilson
(2005) and Comprehensive Meta-Analysis Software (CMA; n.d.) v.2 were run for
final analysis purposes. By treating each of the moderators as a continuous predictor
(by dummy coding the categorical variables into dichotomous variables), this test can
help explain heterogeneity among the moderators (Card, 2012). This test involves
regression of the effect sizes onto a series of continuous moderators testing them
simultaneously rather than testing them one at a time. The moderators were entered
into the equation in order of strength of relationship with the effect sizes. CMA meta-
regression procedures indicated unacceptable multicollinearity among predictor vari-
ables. A follow-up weighted least squares regression verified the presence of
multicollinearity, and a review of a correlation matrix indicated significant correla-
tions for dummy variables related to whether study data were drawn from survey par-
ticipants (Basis for N = 1) or content categories from text analysis (Basis for N = 2).
Only those variables that had the largest or significant relationship with the grand
mean effect size were entered into the final regression analysis.

Results
Overall, the weighted grand mean correlation was .487 (confidence interval [CI] =
[.43, .56]). This value was significantly large with Z = 147.70, df = 81, p < .001. This
statistic is similar to a t test used to test the magnitude of a Pearson’s product–moment
correlation. Using a fixed-effects homogeneity analysis, the grand mean correlation
162 Journalism & Mass Communication Quarterly 96(1)

was also significantly heterogeneous with Q = 334.25, df = 81, p < .001. This signifi-
cant result of Q statistic means that the variability across studies is more than expected
(Lipsey & Wilson, 2001). As sample size plays such an important role in the calcula-
tion of the grand mean effect size, the researchers believed that differences across a
moderator that determined the sample size of correlation analysis, called “Basis for N”
in the study, were the most logical source for the observed heterogeneity. From his
variable, two major groups emerged from the data: those that used content categories
as Basis for N (k = 75) and those that used survey participants as Basis for N (k = 7).
Articles using content categories from text analysis produced a mean of .508, and
those effect sizes based on survey participants produced a mean of .129. We performed
follow-up analysis of variance (ANOVA) analog tests for each moderator, and we
found homogeneity for each, other than Basis for N.
To confirm the results for the above Q tests, a multiple regression analysis with all
the data from both groups was run to uncover the source of heterogeneity for the grand
mean effect size. In the analysis, the dependent variable was study effect sizes, and the
predictors were Basis for N, along with a group of coded study characteristics, includ-
ing year of publication, place of study, research approach, media type, time lag, cor-
relation analysis, type of content, type of issue, type of survey, size of survey, and
study design. As alluded to above, all the categorical variables were dummy coded as
0s and 1s. For example, place of study was recoded as 1 = United States and 0 =
international; research approach was recoded as 1 = first level and 0 = second level;
and media type was recoded as 1 = newspaper and 0 = not newspaper. Type of con-
tent was recoded into two dummy variables, national content (1 = national, 0 = not
national) and regional content (1 = regional, 0 = not regional), because some studies
reported content from both national and regional media and some studies reported
neither. Therefore, national and regional content were not mirror opposites. The
regression results show that R2 was significant (Q = 123.43, df = 14),3 accounting for
71% of the variance. Only one predictor, regional content, was significant at p < .05
(see Table 2). Two other variables, Basis for N and research approach, were significant
at p < .10, a standard recommended by meta-analysis methodologists (e.g., Lipsey
&Wilson, 2001; Rosenthal, 1984).
A second multiple regression was conducted using only the data from the larger
group of articles that used content categories from text analysis as Basis for N. All the
predictors mentioned above were included, except for Basis for N variable. The find-
ings show that R2 was not significant (Q = 10.61, df = 12), accounting for 18% of the
variance (see Table 3). This indicates overall homogeneity among the study effect
sizes. This also helps to confirm that the source of the sample size (N) in correlation
statistics used in a study is critical to understanding this sample of study findings, in
general. After excluding those studies using survey participants as Basis for N, the
regression results produced homogeneity. In addition, the lack of significance among
predictors as shown in Table 3 is an indication of small between-category differences.
This is strong evidence for the weighted grand mean effect size of approximately .51
being a stable representation of agenda-setting effects in existing studies using content
categories from text analysis as Basis for N.
Luo et al. 163

Table 2. Regression Analysis Results for All Studies (k = 82).

β p
Basis for N (1 = content categories) 0.409 .0968
Year of publication 0.2598 .1292
Place of study (1 = United States) 0.1089 .3318
Research approach (1 = first level) 0.5311 .0963
Media type (1 = newspaper) 0.057 .774
Time lag 0.0162 .8726
Correlational analysis (1 = Spearman’s rho) 0.0792 .6691
National content (1 = national) 0.1534 .6535
Regional content (1 = regional) 0.4387 .0304
Type of issue (1 = issue sets) 0.2637 .3055
Type of survey (1 = national) 0.3431 .256
Size of survey 0.0615 .5937
Study design (1 = cross-sectional) 0.058 .7022

Table 3. Regression Analysis Results for Studies Using Content Categories as Basis for
N (k = 75).

β p
Year of publication 0.239 .1918
Place of study (1 = United States) 0.2171 .2528
Research approach (1 = first level) −0.0765 .7176
Media type (1 = newspaper) 0.1162 .4918
Time lag −0.029 .8655
Correlation analysis (1 = Spearman’s rho) 0.04 .8343
National content (1 = national) 0.1197 .7239
Regional content (1 = regional) −0.1363 .4736
Type of issue (1 = issue sets) −0.0666 .7532
Type of survey (1 = national) 0.3515 .1512
Size of survey 0.0305 .8808
Study design (1 = cross-sectional) 0.2409 .2125

Discussion
In the past four decades, agenda setting has developed to be one of the most important
theories in media effects research. Hundreds of articles and publications have been
produced in this field. As agenda-setting theory has continuously evolved over the
years, many scholars have moved beyond the test of original agenda-setting hypothe-
sis to explore other questions such as who set the media agenda in the agenda-building
process. However, news media’s agenda-setting effects on the public agenda as the
core idea of this theory are still a primary interest in agenda-setting research. Especially
164 Journalism & Mass Communication Quarterly 96(1)

in a new media environment, the popular spread of the Internet and social media raises
the question as to whether agenda-setting effects of news media can survive in the
digital age. Pessimistic views even predict the end of agenda-setting influence based
on the assumption that media proliferation and audience segmentation could result in
every citizen building an individualized agenda. To examine the presumably diminish-
ing agenda-setting function of news media and get a more reliable indication of the
media–public relationship over time, this study adopted a meta-analysis approach for
synthesizing the findings from existing studies published from 1972 to 2015.
The results of this meta-analysis showed that the overall weighted grand mean cor-
relation was .487. According to Cohen’s (1988) conventions for interpreting effect
sizes, a correlation between .30 and .50 suggested a medium effect size. However, this
weighted grand mean effect size for all studies needs to be interpreted with caution
because of the large sample size in the cases using survey participants as Basis for N.
In those cases such as Wanta and Wu’s (1992) study mentioned above in the literature
review, the sample size of correlation analysis was the number of survey participants
(N = 341). The mean effect sizes for those cases was 0.129, much lower than the mean
effect size of 0.508 for the studies using content categories from text analysis such as
the original study by McCombs and Shaw (1972). One explanation is that studies
using survey participants analyzed the data at the individual level, while the studies
using content categories analyzed the data at the aggregate level. As agenda setting
was originally proposed as a societal influence (McCombs & Shaw, 1972), the aggre-
gate-level studies found much stronger agenda-setting effects than individual partici-
pant-level studies did.
Following J. Cohen’s (1988) guidelines, the mean effect size for studies using con-
tent categories as Basis for N was large, an indication of the presence of strong agenda-
setting effects. To underscore the importance of this statistic, using the Binomial Effect
Size Display developed by Rosenthal (1984), a correlation of .508 is equivalent to a
Cohen’s D (essentially a z score) of 1.2 (Randolph & Edmondson, 2005). This is far
from a negligible effect, which provides a strong support for the existence of news
media’s agenda-setting influence on the public agenda.
A test of homogeneity of the effect size revealed that the grand mean correlation was
significantly heterogeneous (Q = 334.25, p < .001), indicating that the differences in
effect sizes are larger than what would be expected from sampling error alone (Hunter
& Schmidt, 1990). The multiple regression involving all the studies found that Basis for
N was a significant factor. Given the fact that those studies with survey participants as
Basis for N had huge sample sizes, they drove the significance values and greatly con-
tributed to the variance among effect sizes. When only those studies using text analysis
were considered, significant heterogeneity for the grand mean effect size disappeared,
which also implies the need to treat and think about those two groups of studies sepa-
rately. Therefore, a second regression analysis was run to focus on the larger group of
cases (n = 75), with content categories from text analysis as Basis for N.
The results of the second regression analysis showed that no predictors were sig-
nificantly related to the mean effect size for studies based on text analysis. This non-
significant result provides evidence that means across the categories of moderator
Luo et al. 165

variables are similar to each other and similar to the mean for studies based on text
analysis. In other words, those studies form a homogeneous group and differences in
effect size were due to random error rather than any moderating variable (Benoit
et al., 2003). Even though scholars have concerns about the reduction of agenda-set-
ting effects in a changing media environment that provides consumers with a multi-
tude of media channels to choose from and allow them to construct individualized
agendas (e.g., Bennett & Iyengar, 2008; Chaffee & Metzger, 2001; McCombs, 2004),
year of publication as a nonsignificant factor implies that agenda-setting effects keep
stable over the years and suggests that the agenda-setting influence of news media
remains strong as it used to be. Although emergent new media technologies provide
more media choices for the public to choose, the news media can set each other’s
agendas, which is called “intermedia agenda setting,” and tend to converge on the
most important issues despite their different forms or preferences (Groshek &
Groshek, 2013). The generally homogenized media content leads to a convergence
rather than a divergence of public agendas, which reinforces the news media’s
agenda-setting influence.
Other critical factors documented in the literature, such as research approach, media
type, and time lag, did not show a significant impact on the magnitude of agenda-set-
ting effects, either. The news media could be equally powerful in influencing the per-
ceived importance of issues and attributes in the public agenda at both the first-level
and second-level agenda setting. Although many studies (e.g., K. Kim & McCombs,
2007; Kiousis & Wu, 2008; McCombs & Shaw, 1972) showed the predominant power
of newspapers, especially those prestigious national newspapers such as the New York
Times, in the agenda-setting process, the synthesis demonstrated that newspapers as
well as other news media are often equally sharing the role of setting the public agenda
(McCombs, 2004).
Time lag is another critical consideration in agenda-setting research. Although a
majority of studies followed the recommendation of setting the optimal time span
between 4 and 6 weeks (Winter & Eyal, 1981), the time lags employed in existing
research have varied widely, ranging from a few days or a few weeks, up to several
months and even a year. However, the variance in time lag was not found to signifi-
cantly increase or decrease agenda-setting effects. Another potential factor, place of
study, was not found to be a significant predictor. Even though the agenda-setting
hypothesis was first tested in the United States, strong agenda-setting effects have also
been observed in studies conducted in other countries (e.g., McCombs et al., 2000;
Takeshita & Mikami, 1995), where their political and media system could be similar
or different from that of the United States. To sum up, from study to study, inside and
outside of the United States, over several decades, and even with improved methods
and technologies, agenda-setting effects appear to be consistent and stable.
Finally, these findings illuminate a possible methodological artifact in the field
with the use of Spearman’s rank correlation as a focus. This artifact is illuminated by
the improved methods used in this study. According to Higgins and Green (2011), all
modern methods use a weighted average of effect sizes from individual studies. As
suggested by Hedges and Olkin (1985), we used study sample size as this weight.
166 Journalism & Mass Communication Quarterly 96(1)

This departs from the approach used by Wanta and Ghanem (2007). The weighting
method forced us to consider the calculation methods used in the original studies as
an essential study feature. For example, Wanta and Ghanem (2007) reported a sam-
ple size of 304 for a study by Atwater, Salwen, & Anderson (1985), also included in
this analysis. However, Spearman’s rank correlation calculations used by Atwater et
al. were actually much smaller (n = 6) than the reported sample size. In many cases,
researchers had no references to human study participants, so the textual analyses
were based solely on readings of texts. In some cases, where direct respondent data
were available, researchers converted interval level data to ordinal level, resulting in
loss of information. This loss of information may result in higher agenda-setting
relationships (and in some cases, perfect relationships) among content categories
than those that exist in data drawn directly from human participants. While the over-
all results support the findings from Wanta and Ghanem (2007), the weighted sig-
nificance tests suggest a bifurcated field, divided by how agenda-setting effects are
measured. More research and thought on this conclusion is needed, and that discus-
sion is beyond the scope of this meta-analysis.

Limitations and Future Research


Meta-analyses are unique in that they offer a systematic investigation of the literature
within a field. While this aspect of a meta-analysis may be its strength, it is also its
weakness. The first limitation is that grand mean effect size data may be confounded
because the Basis for N categories originated from vastly different sources. Analyses
using weights based on sample sizes will be biased in favor of those studies with larger
samples, and in this meta-analysis, it favored studies based on large samples of human
participants. The inclusion of seven cases using a large number of survey participants
with low correlation coefficients in the calculation diminished the overall effect size.
The analyses in this study also provide evidence for two different approaches to mea-
suring agenda-setting effects: the basis for the study correlation as either content cat-
egories or survey participants. The results from each are probably not directly
comparable because the underlying populations of the statistics are so different, and
therefore, each needs further investigation.
A second limitation of this study is that the researchers limited the selection of stud-
ies to peer-reviewed published studies in only two major databases. Dissertations and
conference presentations would have added to the variety of studies. In fact, missing
studies from other databases and of all types could change the results if they had been
added to the study. Meta-analysts call this the file drawer problem, where the studies
selected are not representative of the true population of studies (Lipsey & Wilson,
2001). To test this possibility, the researchers did run an Orwin’s Failsafe N procedure
that resulted in a value of 120 studies of no effect (i.e., r = 0), which would be needed
to bring the grand mean effect size of 0.49 down to a small mean effect size of 0.2.
That is, many weak effect sizes would be needed to negatively affect the results here.
This analysis provides evidence suggesting that the sampling bias presented by miss-
ing studies may not affect the results.
Luo et al. 167

A third limitation of this study is that one methodological feature (type of correla-
tion used in the analysis) dominated the findings. Lipsey and Wilson (2001) suggest
that methodological features are not neutral and their impact should be assessed. The
researchers found considerable inconsistencies in the reporting of methods. This may
mask methodological problems. A future study can focus more on determining and
coding good versus poor methodological and reporting practices. The researchers
hope to study this issue further because it may present an important source of variance
in effect sizes. Nonetheless, more and more complete reporting by agenda-setting
researchers on methodological practices will make meta-analyses easier to conduct, a
practice that could enrich the development of agenda-setting theory.
Future research should also consider the inclusion of regression results in the analy-
sis, even though it is not a common practice in meta-analyses and it is hard to find an
appropriate method of deriving the effect size from beta-weights (Benoit et al., 2003),
while others suggest caution in the use of these values (Peterson & Brown, 2005). It
would be interesting to see whether the news media’s effects in setting the public
agenda remain stable and strong after controlling for other factors in more advanced
data analyses. Another shortcoming of the meta-analysis is that it included correlation
statistics, only. These correlations do not represent causal relationships between the
media agenda and the public agenda. Although cross-lagged correlation tests are com-
monly used in agenda-setting research to test the causal relationship between a pair of
agendas at two points in time, scholars do not agree as to whether or not a directional
correlation over time can be used to make statements about causality (Rosenthal &
Rosnow, 2008). Therefore, more rigorous panel design and experimental methods
should be used in future studies to establish the chain of causality and provide a stron-
ger support for the agenda-setting hypothesis.

Declaration of Conflicting Interests


The author(s) declared no potential conflicts of interest with respect to the research, authorship,
and/or publication of this article.

Funding
The author(s) received no financial support for the research, authorship, and/or publication of
this article.

Notes
1. Media type variable was first coded for the presence of all the media examined in the study,
including newspaper, TV, radio, magazine, Internet website (e.g., online counterparts of
traditional news outlets, independent online news websites), and social and mobile media
(e.g., Facebook, Twitter). Because newspaper is the major media type under investigation
in the sample articles, this variable was later recoded as a dummy variable in the regression
analysis (1 = newspaper, 0 = not newspaper).
2. Krippendorff’s alpha of .8 or above indicates an adequate level of intercoder reliability
(Krippendorff, 2004).
168 Journalism & Mass Communication Quarterly 96(1)

3. Adjusted R2 values were not available with the software packages used for this type of
meta-regression, which is one limitation of the study.

References
Albæk, E., van Dalen, A., Jebril, N., & de Vreese, C. (2014). Political journalism in compara-
tive perspective. New York, NY: Cambridge University Press.
Althaus, S. L., & Tewksbury, D. (2002). Agenda setting and the “new” news: Patterns of
issue importance among readers of the paper and online versions of the New York Times.
Communication Research, 29, 180-207.
Atwater, T., Salwen, M. B., & Anderson, R. B. (1985). Interpersonal discussion as a potential
barrier to agenda-setting. Newspaper Research Journal, 6(4), 37-43.
Becker, L. B., & McCombs, M. E. (1978). The role of the press in determining voter reactions
to presidential primaries. Human Communication Research, 4, 301-307.
Bennett, W. L., & Iyengar, S. (2008). A new era of minimal effects? The changing foundations
of political communication. Journal of Communication, 58, 707-731.
Benoit, W. L., Hansen, G. J., & Verser, R. M. (2003). A meta-analysis of the effects of viewing
U.S. presidential debates. Communication Monograph, 70, 335-350.
Benton, M., & Frazier, P. J. (1976). The agenda setting function of the mass media at three
levels of “information holding.” Communication Research, 3, 261-274.
Borenstein, M., Hedges, L. V., Higgins, J. P. T., & Rothstein, H. R. (2009). Introduction to
meta-analysis. West Sussex, UK: John Wiley.
Boulianne, S. (2009). Does Internet use affect engagement? A meta-analysis of research.
Political Communication, 26, 193-211.
Boulianne, S. (2015). Social media use and participation: A meta-analysis of current research.
Information, Communication & Society, 18, 524-538.
Buturoiu, D. R., & Corbu, N. (2015). Moderators of framing effects on political attitudes: Is
source credibility worth investigating? Central European Political Studies Review, 17, 155-
177.
Card, N. (2012). Applied meta-analysis for social science research. New York, NY: Guilford
Press.
Chaffee, S. H., & Metzger, M. J. (2001). The end of mass communication? Mass Communication
and Society, 4, 365-379.
Cohen, B. C. (1963). Press and foreign policy. Princeton, NJ: Princeton University Press.
Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ:
Lawrence Erlbaum.
Comprehensive meta-analysis. (n.d.). [Computer software]. Englewood, NJ: Biostat.
Conway, M., & Patterson, J. R. (2008). Today’s top story? An agenda-setting and recall experi-
ment involving television and internet news. Southwestern Mass Communication Journal,
24, 31-48.
Craft, S., & Wanta, W. (2004). U.S. public concerns in the aftermath of 9-11: A test of second
level agenda-setting. International Journal of Public Opinion Research, 16, 456-463.
Dearing, J. W., & Rogers, E. M. (1996). Agenda-setting. Thousand Oaks, CA: SAGE.
Demers, D. P., Craff, D., Choi, Y. H., & Pessin, B. M. (1989). Issue obtrusiveness and the
agenda-setting effects of national network news. Communication Research, 16, 793-812.
Dursun-Ozkanca, O. (2011). European Union enlargement and British public opinion: The
agenda-setting power of the press. Perspectives on European Politics and Society, 12, 139-
160. doi:10.1080/15705854.2011.572642
Luo et al. 169

Freedom House. (2015). Freedom of the press 2015. Retrieved from https://freedomhouse.org/
report/freedom-press/freedom-press-2015#.V3Km0bgrLIU
Ghorpade, S. (1986). Agenda setting: A test of advertising’s neglected function. Journal of
Advertising Research, 26, 23-27.
Golan, G., & Wanta, W. (2001). Second-level agenda setting in the New Hampshire primary:
A comparison of coverage in three newspapers and public perceptions of candidates.
Journalism & Mass Communication Quarterly, 78, 247-259.
Groshek, J., & Groshek, M. C. (2013). Agenda trending: Reciprocity and the predictive capacity
of social networking sites in intermedia agenda setting across topics over time. Media and
Communication, 1, 15-27.
Guo, L. (2012). The application of social network analysis in agenda setting research: A meth-
odological exploration. Journal of Broadcasting & Electronic Media, 56, 616-631. doi:10.
1080/08838151.2012.732148
Guo, L. (2013). Toward the third level of agenda setting theory: A network agenda setting
model. In T. Johnson (Ed.), Agenda setting in a 2.0 world: New agendas in communication
(pp. 112-133). New York, NY: Routledge.
Guo, L., & McCombs, M. (2011, May). Network agenda setting: A third level of media effects.
Paper presented at the ICA, Boston, MA.
Hallin, D. C., & Mancini, P. (2004). Comparing media systems: Three models of media and
politics. New York, NY: Cambridge University Press.
Hanitzsch, T., Hanusch, F., Mellado, C., Anikina, M., Berganza, R., Cangoz, I., . . . Wang Yuen,
E. K. (2011). Mapping journalism cultures across nations: A comparative study of 18 coun-
tries. Journalism Studies, 12, 273-293.
Hedges, L. V., & Olkin, I. (1985). Statistical methods for meta-analysis. Orlando, FL: Academic
Press.
Hester, J. B., & Gibson, R. (2007). The agenda-setting function of national versus local media:
A time-series analysis for the issue of same-sex marriage. Mass Communication and
Society, 10, 299-317.
Higgins, J., & Green, S. (2011). Cochrane handbook for systematic reviews of interventions
(Version 5.0). The Cochrane Collaboration. Available from http://handbook.cochrane.org
Hunter, J. E., & Schmidt, F. L. (1990). Methods of meta-analysis: Correcting error and bias in
research findings. Newbury Park, CA: SAGE.
Iyengar, S. (1979). Television news and issue salience: A reexamination of the agenda-setting
hypothesis. American Politics Quarterly, 7, 395-416.
Iyengar, S., Curran, J., Lund, A. B., Salovaara-Moring, I., Hahn, K. S., & Coen, S. (2010).
Cross-National versus individual-level differences in political information: A media sys-
tems perspective. Journal of Elections, Public Opinion and Parties, 20, 291-309.
Iyengar, S., & Kinder, D. R. (1987). News that matters: Television and American opinion.
Chicago, IL: The University of Chicago Press.
Kim, K., & McCombs, M. (2007). News story descriptions and the public’s opinions of political
candidates. Journalism & Mass Communication Quarterly, 84, 299-314.
Kim, Y., Kim, Y., & Zhou, S. (2017). Theoretical and methodological trends of agenda-setting
theory: A thematic analysis of the last four decades. The Agenda Setting Journal, 1, 5-22
King, P. T. (1997). The press, candidate images, and voter perceptions. In M. E. McCombs, D.
L. Shaw, & D. Weaver (Eds.), Communication and democracy: Exploring the intellectual
frontiers in agenda-setting theory (pp. 29-40). Mahwah, NJ: Lawrence Erlbaum.
170 Journalism & Mass Communication Quarterly 96(1)

Kiousis, S., & Wu, X. (2008). International agenda-building and agenda-setting: Exploring the
influence of public relations counsel on US news media and public perceptions of foreign
nations. International Communication Gazette, 70, 58-75. doi:10.1177/1748048507084578
Krippendorff, K. (2004). Content analysis: An introduction to its methodology (2nd ed.).
Thousand Oaks, CA: SAGE.
Lippmann, W. (1946). Public opinion. New Brunswick, NJ: Transaction Publishers.
Lipsey, M., & Wilson, D. (2001). Practical meta-analysis. Thousand Oaks, CA: SAGE.
Lopez-Escobar, E., Llamas, J. P., & McCombs, M. (1998). Agenda setting and community con-
sensus: First and second level effects. International Journal of Public Opinion Research,
10, 335-348.
Luo, Y. (2014). The Internet and agenda setting in China: The influence of online public opinion
on media coverage and government policy. International Journal of Communication, 8,
1289-1312.
McClure, R. D., & Patterson, T. E. (1976). Print vs. network news. Journal of Communication,
26(2), 23-28.
McCombs, M. E. (1981). The agenda-setting approach. In D. D. Nimmo & K. R. Sanders (Eds.),
Handbook of political communication (pp. 121-140). Beverly Hills, CA: SAGE.
McCombs, M. E. (2004). Setting the agenda: The mass media and public opinion. Cambridge,
UK: Blackwell.
McCombs, M. E., Llamas, J. P., Lopez-Escobar, E., & Rey, F. (1997). Candidate images in
Spanish elections: Second level agenda-setting effects. Journalism & Mass Communication
Quarterly, 74, 703-717.
McCombs, M. E., Lopez-Escobar, E., & Llamas, J. P. (2000). Setting the agenda of attributes in
the 1996 Spanish general election. Journal of Communication, 50, 703-717.
McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public
Opinion Quarterly, 36, 176-187.
McCombs, M. E., Shaw, D. L., & Weaver, D. H. (1997). New directions in agenda-setting
theory and research. Mass Communication and Society, 17, 781-802.
Mellado, C., & Lagos, C. (2013). Redefining comparative analyses of media systems from the
perspective of new democracies. Information, Communication & Society, 26(4), 1-24.
Muddiman, A., Stroud, N. J., & McCombs, M. (2014). Media fragmentation, attribute agenda
setting, and political opinions about Iraq. Journal of Broadcasting & Electronic Media, 58,
215-233. doi:10.1080/08838151.2014.906433
Peter, J. (2003). Country characteristics as contingent conditions of agenda setting: The moder-
ating influence of polarized elite opinion. Communication Research, 30, 683-712.
Peterson, R., & Brown, S. (2005). On the use of beta coefficients in meta-analysis. Journal of
Applied Psychology, 90, 175-181.
Pralle, S. B. (2006). Timing and sequence in agenda-setting and policy change: A comparative
study of lawn care pesticide politics in Canada and the US. Journal of European Public
Policy, 13, 987-1005.
Price, V., & Tewksbury, D. (1997). News values and public opinion: A theoretical account of
media priming and framing. In G. A. Barnett & F. J. Boster (Eds.), Progress in the com-
munication sciences (Vol. 13, pp. 173-212). New York, NY: Ablex.
Prior, M. (2013). Media and political polarization. Annual Review of Political Science, 16, 101-
127.
Protess, D., & McCombs, M. E. (2016). Agenda setting: Readings on media, public opinion,
and policymaking. New York, NY: Routledge.
Luo et al. 171

Randolph, J., & Edmondson, R. S. (2005). Using the Binomial Effect Size Display (BESD)
to present the magnitude of effect sizes to the evaluation audience. Practical Assessment,
Research and Evaluation, 10(14), 1-7.
Roberts, M., Wanta, W., & Dzwo, T. H. D. (2002). Agenda setting and issue salience online.
Communication Research, 29, 452-465.
Rosenthal, R. (1984). Meta-analytic procedures for social research. Beverly Hills, CA: SAGE.
Rosenthal, R., & Rosnow, R. L. (2008). Essentials of behavioral research: Methods and data
analysis. Boston, MA: McGraw-Hill.
Russell Neuman, W. R., Guggenheim, L., Mo Jang, S., & Bae, S. Y. (2014). The dynamics of
public attention: Agenda-setting theory meets big data. Journal of Communication, 64,
193-214. doi:10.1111/jcom.12088
Shehata, A., & Strömbäck, J. (2013). Not (yet) a new era of minimal effects: A study of agenda
setting at the aggregate and individual levels. The International Journal of Press/Politics,
18, 234-255. doi:10.1177/1940161212473831
Shoemaker, P. J., Wanta, W., & Leggett, D. (1989). Drug coverage and public opinion, 1972-
1986. In P. J. Shoemaker (Ed.), Communication campaigns about drugs: Government,
media and the public (pp. 67-80). Hillsdale, NJ: Lawrence Erlbaum.
Stone, G. C., & McCombs, M. E. (1981). Tracing the time lag in agenda-setting. Journalism
Quarterly, 58, 51-55.
Strömbäck, J., & Kiousis, S. (2010). A new look at agenda-setting effects—Comparing the pre-
dictive power of overall political news consumption and specific news media consumption
across different media channels and media types. Journal of Communication, 60, 271-292.
doi:10.1111/j.1460-2466.2010.01482.x
Takeshita, T., & Mikami, S. (1995). How did mass media influence the voters’ choice in the
1993 general election in Japan? Keio Communication Review, 17, 27-41.
Tan, Y., & Weaver, D. H. (2007). Agenda-setting effects among the media, the public, and con-
gress, 1946–2004. Journalism & Mass Communication Quarterly, 84, 729-744.
Tan, Y., & Weaver, D. H. (2013). Agenda diversity and agenda setting from 1956 to 2004:
What are the trends over time? Journalism Studies, 14, 773-789. doi:10.1080/14616
70X.2012.748516
Wanta, W., & Ghanem, S. (2007). Effects of agenda setting. In R. W. Preiss, B. M. Gayle, N.
Burrell, M. Allen, & J. Bryant (Eds.), Mass media effects research: Advances through
meta-analysis (pp. 37-52). Mahwah, NJ: Lawrence Erlbaum.
Wanta, W., Golan, G., & Lee, C. (2004). Agenda setting and international news: Media influence
on public perceptions of foreign nations. Journalism & Mass Communication Quarterly,
81, 364-377.
Wanta, W., & Hu, Y. W. (1994). Time-lag differences in the agenda-setting process: An exami-
nation of five news media. International Journal of Public Opinion Research, 6, 225-240.
Wanta, W., & Wu, Y. C. (1992). Interpersonal communication and the agenda-setting process.
Journalism & Mass Communication Quarterly, 69, 847-855.
Weaver, D. H. (1980). Audience need for orientation and media effects. Communication
Research, 7, 361-376.
Weaver, D. H., Graber, D., McCombs, M., & Eyal, C. (1981). Media agenda-setting in a presi-
dential election: Issues, images, and interest. New York, NY: Praeger.
Wilson, D. B. (2005). Metareg.sps. Retrieved from http://mason.gmu.edu/~dwilsonb/ma.html
Winter, J. P., & Eyal, C. H. (1981). Agenda-setting for the civil rights issue. Public Opinion
Quarterly, 45, 376-383.
172 Journalism & Mass Communication Quarterly 96(1)

Wu, H. D., & Coleman, R. (2009). Advancing agenda-setting theory: The comparative strength
and new contingent conditions of the two levels of agenda-setting effects. Journalism &
Mass Communication Quarterly, 86, 775-789.

Author Biographies
Yunjuan Luo (PhD, Indiana University) is a professor of communication in the School of
Journalism and Communication at South China University of Technology. Her research inter-
ests include new media use and effects, international communication, health communication,
and communication theory and methodology.
Hansel Burley (PhD, Texas A&M University) is a professor of educational psychology and
chair of the Educational Psychology and Leadership Department at Texas Tech University. His
research interests include exploring educational trends and related policies, particularly those
related to diversity, postsecondary access, and remedial education.
Alexander Moe (PhD, Texas Tech University) is an assistant professor of communication at
Coastal Carolina University. His research interests include exploring how various forms of
media play in forming perceptions about issues associated with sports in the current age.
Mingxiao Sui (PhD, Louisiana State University) is an assistant professor of media and com-
munication in the School of Arts and Humanities at Ferrum College, Ferrum, VA. Her research
interests include political communication, selective exposure/partisan media, ethnic media/race
and ethnicity, digital and mobile news technology, and political public relations.

You might also like