Mass Media and Public Opinion: An Experimental Study of M edia Effe cts [601287]

Mass Media and Public Opinion: An Experimental Study of M edia Effe cts
Glen Smith
Department of P olitical Science
Washington State University
smith631@w su.edu
1Abstract
The media’s ability to change publ ic opinion ha s long be en a concern in pol itical science
research. Re cent research has found t hat news coverage can influence publ ic opinion on
political issues and vot ing de cisions. One shortfall of traditional media effects research is
that it tends to lump together editorials and opi nion columns with straight news articles; a
practice that prevents the researchers from analyzing the influence of each type
separately. O ne exploration of m edia effects did separate the influence of ne ws from the
influence of opi nion, a nd found t he opinion columns and editorials had more influence on
voting de cisions than news articles. I expand on t his research by pe rform ing an
experiment testing the influence of opi nion columns and editorials. The results indicate
that 1) opi nion columns have a persuasive influence over readers, 2) this influence is
generalizable to readers’ opinions on pol itical issues, 3) the salience of the issue makes
little difference in the persuasive pow er of e ditorials and opi nion columns, and 4) opi nion
columns make readers more ideologically consistent in their issue opinions.

Paper prepared for t he annua l meeting of t he International Society of P olitical
Psychology on J uly 5, 2007.
Introduction
1

The impact of the mass media on publ ic opinion ha s long be en an area of gre at
concern in pol itical science and popul ar culture. Hitler’s propa ganda machine and its
assumed influence on the German people during W orld War II c aused gre at concern
among gove rnment officials and academics. This concern was largely put to rest after
research into media effects found t hat the news media tend to reinforce existing opi nions
rather than persuade people to change opinions (Hovland, 1953; Klapper, 1960).
Although pa st research has mostly examined the influences of straight news articles,
recent research has pointed to the pow er of opi nion columns to influence voting de cisions
(Dalton, Be ck and H uckfeldt, 1998; Beck et al., 2002). U nfort unately, the pervasiveness
of this influence is unclear because the research has been limited to an examination of t he
1992 pre sidential election. M y research aims at testing the persuasive pow er of opi nion
columns in a broa der context. I attempt to expand our unde rstanding of t he persuasive
power of t he opinion section of ne wspapers in the following w ays. First, I test the pow er
of opi nion columns in an experimental setting, a dding a measure of internal validity to
past research which relied on s urveys and content analyses. Second, I t est whether
opinion columns influence publ ic opinions on pol itical issues and pol icy opt ions. Past
research found opi nion columns and editorials influence voting de cisions, but no re search
has investigated whether that persuasive pow er translates to publ ic opinion concerning
political issues. Finally, I investigate whether the salience of the issues (the readers’
levels of interest in the issues) influence the persuasive pow er of opi nion columns or
news articles.
In the present research, I test the influence of opi nion columns by e xpos ing three
groups of pa rticipants to three different types of articles: balanced articles, biased articles,
and opi nion columns. One group i s expos ed to two articles each one containing a two-
2

sided communication flow (pos itive and ne gative aspects) concerning re form plans. One
article discussed Bus h’s Social Security reform plan and the other discussed the 2005
bankrupt cy reform legislation that was eventually passed by Congre ss. A second group
was expos ed to the same articles minus almost all negative aspects of the reform plans,
isolating the absence of ne gative aspects as the only difference between the two groups .
Participants in the third group w ere expos ed to the same (balanced) articles as the first
group, but were also expos ed to opi nion columns oppos ing each of t he issues.
The experimental conditions I have set up pi t manipulation against persuasion. If
a journa list gives only pos itive aspects of a plan to the publ ic without mentioning the
negative aspects and this influences publ ic opinion, it is certainly manipulation. O n the
other hand, if people’s opinions change because they read an argum ent in an opi nion
column that presented a reasoned argum ent, it is difficult to call this manipulation as it
more closely resembles persuasion. A lthough i t is certainly pos sible for opi nion columns
to manipulate readers, I show in the results section that this is not likely the case because
those reading opi nion columns held opi nions that were more ideologically consistent than
either of t he other two groups .
In the end, m y research adds important insights into the media’s pow er to
persuade the publ ic. My findings sugge st that while the news media have little pow er to
manipulate publ ic opinion, they are able to persuade people through t he opinion section
of the newspaper. Opinion columns and editorials also allow people to better unde rstand
how the issue fits with their values and pre dispositions. The results from my experiment
coupled with the findings of pa st survey research sugge st that future research may find
more media effects by s eparating opi nion columns and editorials from straight news
articles.
3

Media Effe cts Research
Are the news media able to persuade readers? If so, that would give them the
power to influence publ ic opinion; a topic at the center of m edia effects research. M ost
past research has found t hat the media are ineffective at changing publ ic opinion
(Klapper, 1960). Ra ther than change opinion, the news media tend to reinforc e people’s
current opinions (Klapper, 1960). Re inforc ement occurs because of such psychological
factors as selective expos ure and selective retention. P eople are more likely to remember
inform ation that they agree with, and tend to avoid expos ure to inform ation that is
contrary to their current beliefs. People generally resist and avoid discordant inform ation
because of the discomfort that usually follows. Because of these proc esses, the news
media are often ineffective at persuading pe ople to change their opinions on pol itical
issues. As we will see, how ever, this is not always the case as certain circumstances have
the potential to increase the news media’s pow er to persuade the publ ic.
One circumstance likely influencing m edia effects is the salience of the issue at
hand. T he salience of a political issue represents the amount of interest one has in an
issue, as well as how often and how recently they have thought about the issue. Less
experience with a political issue allows for m ore media effects because people lack the
previous considerations necessary to reject inform ation contrary to their opinions (Zaller,
1992). It is impossible to reject inform ation contrary to one ’s previous opinion if there
are no pre vious opinions. A lack of pre vious considerations would reduce selective
retention and increase the influence of the inform ation provi ded by t he news media
concerning the issues.
Another factor influencing the media’s pow er to influence opinions are
4

characteristics of the messages. The media are likely to influence publ ic opinion w hen
they consistently pre sent a one-sided communication flow to the publ ic. Zaller (1992,
1996) us ed survey data to show how mass opinion duri ng the Vietnam War was linked to
the amount of di scord a mong pol itical elites. Throughout most of the War, the publ ic
was overwhelmingly favorable to staying the course. Zaller (1992, 1996) a rgues that the
consistent support among t he publ ic was due to a consensus among pol itical elites on the
issue. The intermediate actor in this relationship is the news media. Elite consensus on
the issue resulted in the media presenting a one-sided communication flow to the publ ic;
the publ ic only heard the pro-w ar side of the debate. This occurred because the news
media relied on pol itical elites as inform ation sources, which caused the level of di scord
in the media to be indexed to the level of di scord a mong pol itical elites.1 The
presentation of a one-sided communication flow in the news media caused those paying
attention to the media to have a consensus concerning the War. Thus, when the media
presents a one-sided communication flow concerning pol itical issues, large media effects
on publ ic opinion are possible.
One probl em with Zaller’s (1992, 1996) s tudy i s that the media are taken as a
whole, and not separate parts. Elite opinion can be transferred through s traight news
articles or opi nion columns and editorials. Although i t is very likely that news articles
presented a one-sided communication flow concerning the War, it is unclear whether that
was the cause of publ ic consensus. If opi nion columns and editorials follow elite
opinion, they are also likely to pre sent a consensus view concerning the War. The use of
survey data makes it impossible to determine whether the media’s influence on publ ic
opinion is driven by opi nion columns and editorials or straight news articles.2
1 For more on the concept of indexing in the media see Bennett (1996) . 2 Zaller was not looking at this question, but rather he examined whether elite consensus influenced publ ic
opinion through t he media. Whether the influence was caused by the opinion section or the straight news
5

The news media were also found t o influence retrospective evaluations of the
econom y. H etherington (1996) found t hat increased expos ure to the news media (prior to
the 1992 e lection) c aused people to perceive the econom y more negatively. In t urn, t hose
econom ic perceptions influenced people’s voting de cisions. Because most of the
econom ic news was negative, the incumbent president (George H. W. Bus h) suffered at
the polls. This study, t oo, fa ils to determine whether econom ic evaluations were
influenced by l earning about the econom y from news articles, or from the elites in
opinion columns and editorials. Perhaps people took c ues from the elites in the opinion
section on how the econom y was doing. Brody (1991) s ugge sts that evaluations of
presidential perform ance are influenced by e lite opinion, one source of w hich are the
opinion sections of ne wspapers. In s hort, both Zaller (1992, 1996) a nd H etherington
(1996) found i mportant media effects on publ ic opinion, but it is unclear whether these
effects were driven by t he inform ation contained in straight news articles or the opinions
and pe rsuasive argum ents contained in the opinion sections of the newspaper.
Recent research has also found m edia effects on e valuations of pol itical
candidates. Research on S enate election coverage has indicated that the news media can
influence people’s voting de cisions, although t he influence is minor. K ahn and K enney
(2002) pe rform ed a content analysis of the largest newspaper in each of t he fifty states
and combined it with data from surveys asking pe ople in each state about their media
consumption ha bits and vot ing de cisions. The results indicate that the media’s tone
toward Senate candidates influenced reader’s vote choice. More favorable coverage in
the local newspaper caused candidates’ vote totals to increase. Kahn and K enney (2002)
argued that this resulted from the endors ement decisions of the ownership and editorial
articles was of little concern in his research. Therefore, this takes nothing away from his research, but
merely points to an avenue for future research.
6

staffs because endors ements influenced straight news coverage of S enate candidates.
Editors and ow ners caused more positive straight news coverage, and the slanted news
influenced readers’ voting de cisions. Similar findings were reported in a case study of
the Minnesota Senate race during the 2000 e lection (D ruckman and Parkin, 2005).
The content analysis in bot h Kahn and K enney (2002) a nd D ruckman and Parkin
(2005) fa iled to separate the influence of editorials and opi nion columns from the
influence of straight news articles. Perhaps editorials and opi nion columns had a larger
influence on vot ing de cisions than straight news articles. The true cause of opi nion
change in these studies, then, might be the content of the opinion section ra ther than
straight news articles. Opinion columns and editorials are likely slanted toward the
ideological direction of t he publ isher or e ditor (P age, 1996). If opi nion content is
influenced by t he partisan leanings of the newspaper owners—and this is likely the case
—it would explain the influence of endors ement decisions. In ot her words, the partisan
leanings of ne wspaper editors and ow ners may cause biased news coverage, but they are
also likely to cause biased opi nion content. Without separating the content of the opinion
section from the news articles, it is impossible to determine whether the influence found
in Senate election re search was driven by t he slant of the newspapers’ opinion sections or
the papers’ slanted news coverage of the candidates.
A similar study t o Kahn and K enney (2002) w as done on the 1992 pre sidential
election, but unlike the Senate election re search, the researchers in this study s eparated
the content of opi nion columns and editorials from the content of straight news articles.
Dalton, Be ck and H uckfeldt (1998) pe rform ed a content analysis of ne wspaper and
television coverage of the candidates in the 1992 pre sidential election. T he results
indicated that the content of the election coverage was mostly obj ective, but editorials
7

and opi nion columns were more partisan. They combined the content analysis with a
representative survey of s elect counties throughout the United States. Separating the
opinion content from the content of straight news articles allowed the researchers to
determine which factor was a stronge r influence on pe ople’s voting de cisions. Survey
results indicated that vote choice was influenced by t he content of the newspapers that
people read. Furthermore, the influence of the editorials and opi nion columns was
stronge r than the influence of straight news articles (which was minimal). The
researchers concluded that editorials and opi nion columns influenced vot ing de cisions by
provi ding pa rtisan cues to the readers that simplified the choices in the election.
Although s traight news articles were mostly obj ective, when bias did existed it had little
influence on re aders’ voting de cisions. Rather, it was the more partisan editorials and
opinion columns that influenced vot ing de cisions. In s hort, Dalton, Be ck and H uckfeldt
(1998) provi de strong e vidence that opinion columns influence publ ic opinion m ore than
straight news articles.3
The findings of D alton, Be ck and H uckfeldt (1998) a re limited by t heir reliance
on da ta from the 1992 pre sidential election. H etherington (1996) s ugge sts that the 1992
presidential election ha d unc haracteristically high levels of m edia effects because of the
poor e conom y at the time. The dom inance of ne gative econom ic news coverage during
the campaign caused large media effects on vot ing de cisions, but media effects were
much weaker duri ng the 1984 a nd 1988 pre sidential elections. Thus, sole reliance on the
1992 e lection m ight have exaggerated the influence of the news media on publ ic opinion.
Reliance on s urvey data also pos es many of t he same probl ems with external validity that
plague most survey research. For one , it is impossible to know whether the survey
3 Subsequent analysis showed that opinion columns and editorials influenced vot ing de cisions even when
readers’ interpersonal communications were taken into account (Beck et al., 2002) .
8

participants actually read the opinion columns or editorials. An experimental design is
required to isolate the opinion columns as the only factor influencing opi nion.
In sum, past research sugge sts that the media can influence publ ic opinion on
political issues and vot ing de cisions. It is unclear, how ever, whether these media effects
are driven by t he content of straight news articles or by t he content of the opinion section
of the newspaper. Dalton, Be ck and H uckfeldt (1998) di d find opi nion columns to have a
stronge r influence on publ ic opinion than news articles, but the context of these findings
are limited to one presidential election.
My research expands on pa st research in the following w ays. First, I investigate
whether opinion columns influence publ ic opinion on pol itical issues. Although pa st
research has found t hat opinion columns influence vote choices during elections, it is
unclear whether the same influence exists in less partisan contexts, and on i ssue opinions
in addition to vot ing choices. It is undoubt edly important to unde rstand m edia effects
during elections, but it is also important to unde rstand w hether those effects are similar
on opi nions outside of the election pe riods.4 Second, I provi de an experimental test of the
persuasive pow er of opi nion columns compared to the manipulative pow er of s traight
news articles. My experimental design allows me to isolate opinion columns as the only
stimulus influencing pa rticipants’ opinions. Although pa st research has shown that
opinion content has a strong i nfluence on vot ing de cisions, an experimental design is
required to isolate the influence of opi nion columns as separate from straight news
articles.
4 The rise of going publ ic as a political strategy, as well as the permanent campaign, ha ve both increased
the importance of publ ic opinion in American democracy (Kernell, 1997) . Presidents have heightened their
efforts to influence publ ic opinion, a nd this has made it nearly impossible to pass major policy initiatives
without at least majority suppor t from the American people.
9

Third, I e xpand on pa st research by e xamining w hether media effects are
influenced by t he salience of the issues discussed. If opi nion columns do influence
support for publ ic policies, is the influence magnified on i ssues of low salience? I
examine opinions on two issues: Social Security and ba nkrupt cy reform . Social Security
is a moderately salient issue as two years before the experiment, reform ing the system
was at the top of P resident Bush’s agenda. Bankrupt cy reform took a back seat to Social
Security reform in spring 2005, w hich likely resulted in people having less know ledge or
interest in the bankrupt cy issue. As I discuss in the results section, the lower salience of
the bankrupt cy issue is evidenced by t he lower levels of interest in bankrupt cy reform
(compared to Social Security reform ) among pa rticipants in my experiment. Looking at
these issues allows me to investigate whether media effects are more pow erful on low
salience issues. It is likely that manipulation and pe rsuasion are greater on i ssues of low
interest or w hen people have few previous considerations concerning the issues.
Theory
Given the results of pa st research, I e xpect opinion columns will have a larger
influence on publ ic opinion than straight news articles. Past research has shown that the
news media act to reinforc e, rather than change publ ic opinion (K lapper, 1960). I e xpect
that reading a biased pre sentation of a reform plan will not influence people’s level of
support for t hat plan. Those reading bi ased news articles are likely to have the same
levels of support as those reading ba lanced articles.
Although I do not expect straight news articles will influence publ ic opinion, I do
expect opinion columns will be persuasive in bringing pe ople to the side advoc ated by t he
columnist. As mentioned, pa st research indicates that bias in opi nion content influences
readers’ voting de cisions. This leads me to expect opinion content to have a significant
10

influence on publ ic opinion. T hose reading opi nion columns written in oppos ition to the
plan will have significantly less support for t he plan than those only reading ba lanced
straight news articles.
Opinion columns are likely to have more influence because they are meant to be
persuasive. The influence of the opinion sections results from the fact that opinion
columns and editorials generally argue a position. O pinion columnists are typically
political elites who are skilled at developing the best argum ent for or a gainst a policy.
They also pos sess the ability to frame the argum ent in the best possible light, making
some values and considerations appear more relevant than others (Nelson, Cl awson and
Oxley, 1997). In s hort, the persuasiveness of opi nion columns and editorials results from
the quality of a rgum ent found i n the articles.
Another expectation re lates to the difference in salience between the two issues
involved in my research. I e xpect that media effects will have an inverse relationship
with the salience of the issues discussed. The less someone know s about or cares about
an issue, the more likely they are to be influenced by t he media’s presentation of t hat
issue. Media effects are more likely to occur on l ow salience issues compared to higher
salience issues.
Experimental Procedures
For the experiment, participants were random ly assigned to one of three groups .
All participants were expos ed to articles concerning Social Security and ba nkrupt cy
reform legislation. O ne article was written in spring 2005 a nd concerned adding pri vate
accounts to the Social Security system. Also in the spring of 2005, Congre ss passed
bankrupt cy reform legislation that altered the laws regulating how people could file for
11

bankrupt cy and the rules regarding re payment. In t he control group, pa rticipants were
expos ed to articles presenting bot h pos itive and ne gative aspects of the reform plans; I
refer to this group a s the balanced group. A separate group w as expos ed to the same
articles as the control group, but those articles were modified to exclude almost all
negative considerations concerning the plans.5 This group pre sents a biased account of
the reform plans, and thus I refer to it as the biased group. A rticles for bot h groups
mentioned pos itives concerning the bankrupt cy and pri vate accounts policies, but only
the balanced group re ad the negative aspects of the plans. The experiment effectively
isolated the lack of ne gative considerations as the treatment condition. A comparison of
opinion be tween the balanced group a nd the biased group w ill indicate whether expos ure
to a one-sided communication flow influences support for t he plan.
A third group of pa rticipants were expos ed to bot h the balanced straight news
articles and opi nion columns concerning each issue. In ot her words, the opinion column
group re ad the two-sided articles (just like the balanced group) but also read opi nion
columns concerning each issue.6 The opinion columns were written in the New York
Times and are oppos ed to bot h reform plans. A comparison be tween the balanced group
and the opinion columns group a llows me to determine whether the addition of opi nion
columns influences the participants’ support for t he plans.
The participants in the experiment were recruited from unde rgraduate political
science classes at Washington State University. In a ll, there were 153 pa rticipants in the
5 Articles in bot h the control and experimental groups provided party cues on the issue. This allows them
to use the party as a heuristic to determine whether they suppor t or oppos e the policy. The Social Security
article mentions Bush’s suppor t, but the Bankruptcy article only mentions Republican Senators as
suppor ting the plan; Bush’s name is absent. Providing pa rty cues tests whether people rely more on
information in evaluating pol icy or rely more on he uristics using effective evaluations of parties (Brady and
Sniderman, 1985) .6 Those in the opinion columns group r ead four pieces altogether. They read two articles: one article
concerning ba nkruptcy, one concerning Social Security. In addition, they read an opi nion column
concerning Social Security and an opi nion column concerning ba nkruptcy reform.
12

experiment and all received course credit in exchange for t heir participation. A lthough
the drawbacks to using unde rgraduates are well-documented (Sears, 1986), I do not
expect my results to stand alone. My experiment is designed to supplement past research
indicating that editorials and opi nion columns have more influence on publ ic opinion
than straight news articles (Dalton, Be ck and H uckfeldt, 1998; Beck et al., 2002). T he
experiment allows me to ove rcome the internal validity probl ems of pa st research by
isolating opi nion columns as the causal stimulus influencing any m easured opi nion
change. In s hort, the findings of D alton, Be ck and H uckfeldt (1998) provi de external
validity for t he basic findings of m y experiment, and m y results provi de the necessary
internal validity lacking in their survey-based research.
Perhaps the most important criticism of us ing unde rgraduate students in
experiments is that, because they are younge r, they are typically more susceptible to
opinion change or m anipulation than a representative sample (Sears, 1986). I a ttempt to
overcome this probl em by c omparing the influences of two different stimuli. By
comparing the influence of opi nion columns to the influence of ne ws articles, I control
for the increased susceptibility to manipulation found a mong unde rgraduates. If
unde rgraduates are more susceptible to influence, it will cause opinion change in bot h
groups . If opi nion columns are found t o be persuasive, but news articles are not, it is not
likely due to the increased susceptibility of c ollege students, but rather it is due to the
persuasiveness of the opinion columns. In ot her words, by c omparing the influence of
opinion columns to biased news articles, I am able to control for unde rgraduates’ greater
susceptibility to influence. Essentially, the balanced and bi ased articles groups serve as
tests of the influence of norm al or bi ased articles on unde rgraduate’s perceptions of bi as.
Survey Questions
13

After reading the articles, all participants completed surveys asking va rious
questions concerning their opinions on the issues. The dependent variables in this study
were the participants’ opinions on the Social Security and ba nkrupt cy reform policies.
Support for S ocial Security reform is measured by t he following que stion: “Do you
support or oppos e changing the current Social Security system to include personal
retirement accounts?” Opinion on ba nkrupt cy reform was measured by a sking: “In 2005,
new laws were passed that made it more difficult to file for ba nkrupt cy. Ba sed on w hat
you know about the changes to the past laws, do you s upport or oppos e those changes?”
For each of t hese questions, participants were asked to rate their level of support on a 7-
point scale from strongl y oppos e to strongl y support .
14

Interest in the issues was measured by a sking pa rticipants to: “Please rate how
closely you fol low news concerning the following issues on the scale below.” The
participants rated their interest in the Social Security and ba nkrupt cy issues on 7-poi nt
scales ranging from “not at all interested” to “very interested”. Ide ology w as measured
on a 7-poi nt scale ranging from liberal (one ) to conservative (seven). In a ddition,
participants were asked whether they consider themselves Republicans or D emocrats.
15Table 1 C haracteristics of P articipants Across Groups
Balanced GroupBiased GroupOpinion Columns
Age 20.51 20.71 20.33
Female57% 40% 37%
Ideology4.09 4.25 4.29
S.S. Interest3.78 3.75 3.64
Bank Int erest3.36 3.01 2.96
Republican38% 47% 43%
Democrats38% 32% 39%
Television1.05 1.03 0.77
Newspapers0.83 0.78 0.72
Internet 1.24 0.84 1.03
Radio0.17 0.28 0.27
N 47 55 51
The columns represent the characteristics of the participants in the study according to the groups to which
they were assigned. A ge is in years and female shows the percentage of women in each group. I deology is
the mean for the group a nd is based on a 7-point scale with 1 representing the most liberal position and 7
representing the most conservative position. I nterest in bot h bankruptcy and Social Security reform are
also measured on 7- point scales from ‘not at all interested’ to ‘very interested’. Both the Republican and
Democrat rows represent the percentage of participants in each group i dentifying as Republicans or
Democrats, respectively. The rest of the participants are used as the baseline for comparison. A ll of the
expos ure media sources variables were the mean num ber of hour s in an average day that participants
received news from that source. The bottom row represents the num ber of participants in each group.

Participants were also asked how many hours per day they usually get news from
television, ne wspapers, radio, or t he internet.
Accompanying these measures were other standard de mographic variables such
as age and ge nder. Table 1 reports the basic demographics, ideology, pa rtisanship and
media attention of t he participants by e xperimental group. T he groups were similar
across most independent variables. One exception to this was the distribution of fe male
participants across the groups , which ranged from 57% i n the balanced group t o 37% i n
group opi nion columns group. Be cause of these differences, all of the models below
include a dum my variable indicating ge nder. Age was not included in the final models
because it was found t o have little influence on the overall results or on t he dependent
variables.
Results
Are opinion columns more persuasive than straight news articles? Before getting
to the influence of opi nion columns, it is important to unde rstand the influence (or l ack of
influence) of s traight news articles. Table 2 reports the results from an orde red probi t
regression m odel with support for S ocial Security and ba nkrupt cy reform as the
dependent variables. The results indicate that the contents of straight news articles have
little influence on pa rticipants’ support for re form . On bot h issues, the coefficients for
membership in the biased group w ere negative, indicating that those reading articles
biased in favor of re form had less support than those reading ba lanced articles.
Therefore, taking away all negative aspects of a policy did not cause more support for
that policy. This is consistent with my expectations given the results of pre vious research
finding m inimal media effects (Klapper, 1960). S urpri singly, on bot h issues the biased
16

group s howed less support for t he plans than the balanced group w hen other predictors
are taken into account. This is an interesting finding as the biased articles did not present
any ne gative aspects of the reform plans. One would expect that expos ure to onl y
positive aspects of the plans would cause at least a modest increase in support for t he
reform plan. The direction of t he influence provi des strong e vidence that the content (or
slant) of s traight news articles does not influence publ ic opinion on pol itical issues. In
short, the contents of straight news articles are not persuasive even when they pre sent a
biased account of publ ic policies.
Table 2 I nfluences on support for Social Security an d bankruptcy reform.
Social Security OpinionBankruptcy Opinion
Biased Group -.240 (.219) -.031 (.216)
Opinion Col umns Group-.525 (.227)* -.477 (.222)*
Ideology .113 (.072) .122 (.071)
Republican .566 (.264)* -.038 (.260)
Democrat .052 (.252) -.544 (.249)*
Interest in the Issue.183 (.065)** -.020 (.068)
Female .199 (.188) -.345 (.185)
Television -.280 (.112)* -.112 (.109)
Paper .179 (.129) -.059 (.125)
Internet -.131 (.091) -.101 (.092)
Radio .082 (.117) .090 (.115)
Cut 1 -0.878 (.495) -2.823 (.528)
Cut 2 0.039 (.483) -1.825 (.475)
17

Cut 3 0.322 (.481) -1.27 (.468)
Cut 4 1.647 (.492) -0.858 (.466)
Cut 5 2.679 (.515) 0.136 (.463)
Cut 6 1.308 (.468)
N 150 150
Log Likelihood -207.93 -232.26
Chi-Square (11) (p-va lue)35.44 (p< .001)35.69 (p< .001)
The results in the first column are from an ordered-probit with opi nion toward Social Security as the
dependent variable. The second c olumn reports the results with opi nion toward bankruptcy as the
dependent variable. The columns report the coefficients for both models with standard errors in
parentheses. Both dependent variables are on a 7-point scale, with 1 representing strong oppos ition and 7
representing strong s uppor t for the Social Security and ba nkruptcy reform measures discussed in the
articles and throughout spring of 2005. T he biased group a nd opi nion column group va riables are dum my
variables indicating the group of the participants. The balanced group i s used as the base-line for
comparison to the other groups . Interest in the issues is based on a 7-point scale (with 7 representing ve ry
interested and 0 i ndicating no i nterest) and w as asked for each issue separately. Ideology represents a 7-
point scale (with 7 a s most conservative). The Republican and D emocrat variables are dum my variables
indicating identification w ith a political party. Those not identifying w ith either party are used as the
baseline for comparison. F inally, the television, pa per, internet and radio variables represent how many
hours in the average day that participants get news from those sources.
*p< .05 **p< .01
If straight news articles have no influence in such an extreme case, how can I
explain the results of pa st research indicating that slanted news influences publ ic opinion
(Druckman and Parkin, 2004; Kahn and K enney, 2002)? In a ddition to my results, the
results in Dalton, Be ck and H uckfeldt (1998) i ndicate that news articles have less
influence on the vote than editorials and opi nion columns. There are at least two pos sible
explanations for t he differences between my results (and D alton, Be ck and H uckfeldt,
1998 re sults) and the findings during Senate elections (Kahn and K enney, 2002;
Druckman and Parkin, 2005). O ne explanation is that the media effects found i n past
research were driven by t he influence of opi nion columns and editorials and not straight
news articles. Perhaps researchers found m edia effects in the past because they did not
separate the influence of the two sections of the papers.
18

Another explanation is that news articles have more influence on publ ic opinion
in Senate elections compared to pre sidential elections. The lower levels of interest and
know ledge about Senate elections might cause the impressions supplied in news coverage
to be more important in those elections than in pre sidential elections. Also, in Senate
elections, people are generally restricted to one newspaper for c overage of their Senator,
but there are many di fferent sources in pre sidential elections that cover the race. Another
factor could be the generally favorable relationship between most Congre ssmen and their
local reporters (Arnold, 2004; Prior, 2006). T his is a matter for fut ure research, as I will
turn to an investigation of t he pow er of opi nion columns to influence publ ic opinion.
Although a rticles have little pow er to influence publ ic opinion, opi nion columns
appear to have some persuasive pow er ove r readers. Participants in the opinion columns
group re ad the same balanced articles read by t he control group, a nd also read one
opinion column on e ach issue. The results indicate that reading opi nion columns
decreased support for t he reform plans. This effect is in the expected direction, a s the
opinion columns were written in oppos ition to bot h plans. Thus, the opinion columns
were persuasive in moving opi nion toward oppos ition (the desired direction) to bot h
Social Security and ba nkrupt cy reform . This provi des strong e vidence that opinion
columns can influence publ ic opinion m ore than straight news articles. In a ddition, these
findings show that the influences of opi nion columns are not confined to vot ing de cisions
in pre sidential elections, as previous research has found (D alton, Be ck and H uckfeldt,
1998). O pinion columns can also have an influence on publ ic opinion concerning such
issues as Social Security and ba nkrupt cy reform . These findings, coupled with the results
in Dalton, Be ck and H uckfeldt (1998) s ugge st that opinion columns have broa d
persuasive pow er to influence publ ic opinion.
19

The influence of opi nion columns on s upport for re form is similar for bot h issues.
Bankrupt cy reform ’s status as a lower salience issue did not influence the persuasiveness
of the opinion columns. The low salience of ba nkrupt cy in comparison to Social Security
is evidenced by t he lower levels of interest in the bankrupt cy issue. When all the groups
are combined, ba nkrupt cy had a mean score of 3.10 on a 7-poi nt scale. This score is
significantly lower than the participants’ interest in the Social Security issue (3.72).7
Although ba nkrupt cy is a lower salience issue than Social Security, reading opi nion
columns had a similar influence on bot h issues. Opinion columns had the same influence
on pe ople’s support for ba nkrupt cy reform as they did on s upport for S ocial Security
reform . The salience of the issues also had little influence the manipulative pow er of
straight news articles. In s hort, the lack of i nterest in the bankrupt cy issue did not appear
to have any impact on pe ople’s susceptibility to influence from either straight news
articles or opi nion columns.
What other factors influence support for re form ? Surpri singly, ideology ha d very
little influence on pe ople’s level of support for t he plans. Republicans were more likely
to support reform ing Social Security than Democrats or Inde pendents, but Democrats
were no less likely to oppos e reform than Inde pendents. On the bankrupt cy issue,
Republicans were no m ore likely to support reform than Inde pendents, but Democrats
were significantly more likely to support reform than either Republicans or Inde pendents.
The stronge st predictor of s upport for S ocial Security reform was interest in the issue, but
this had little influence on opi nions toward ba nkrupt cy reform .
An interesting finding is the influence of w atching television on pe ople’s support
for S ocial Security reform . The more television one watched, the more likely they were
7 A difference in means test showed a significant difference between interest in the two issues (t = 5.82, p<
.05).
20

to oppos e Social Security reform , and this effect exists even when party identification,
ideology a nd interest are controlled. A n explanation for t his is that those watching
television re gularly were more likely to have previous know ledge of Bus h’s Social
Security reform plan. This effect did not exist on the bankrupt cy issue because there was
less news coverage of that issue in the mainstream media during 2005. It is likely that
watching television increased participants’ previous know ledge of the Social Security
issue and m ade them less reliant on the articles read in the experiment for i nform ation
concerning Bus h’s plan. Thus, watching m ore television appeared to increase the
participants’ know ledge of Bus h’s plan and this made them less willing to support the
reform measures that he propos ed.
In addition to the findings given above showing that opinion columns are
persuasive, my results also indicate that opinion columns make people more ideologically
consistent. Those reading opi nion columns showed a stronge r correlation be tween their
opinions on the issues and their self-reported ideological scores. The correlation be tween
support for S ocial Security reform and ideology w ent from .13 i n the balanced articles
group t o .24 i n the biased articles group, but jumped to .38 i n the opinion columns group.
Although l ess pronounc ed, a similar change occurred in the bankrupt cy issue, as the
correlations went from .35 i n the balanced articles group t o .36 i n the biased articles
groups , but then jumped to .43 i n the group e xpos ed to opi nion columns. These results
indicate that opinion columns allow people to conne ct the issues to their ideological
predispositions. Perhaps opinion columns simplify the issues for t he readers. It is likely
harder for re aders to unde rstand the considerations and va lues involved with complex
political issues when reading ne ws articles that typically pre sent fact-based or s trategy
coverage of the news. Opinion columns provi de clear partisan and ideological cues to the
21

readers, but news articles often lack these cues. That opinion columns increase the
readers’ ideological consistency with their issue opinions appears to add furt her evidence
that opinion columns persuade rather than manipulate publ ic opinion. .
Discussion
The results from my experiment add important insights into the media’s pow er to
influence publ ic opinion. F irst and fore most, my results add support to past research
indicating that opinion columns and editorials have a significant influence on publ ic
opinion. P ast research relied on s urvey techniques which lack internal validity (Dalton,
Beck and H uckfeldt, 1998), but my experimental design allowed me to isolate opinion
columns as the only variable influencing publ ic opinion. W hen coupled with the findings
of Dalton, Be ck and H uckfeldt (1998), m y results sugge st that the content of the opinion
section is more likely to influence publ ic opinion than the content of straight news
articles. Future research into media effects in the aggregate would likely find it useful to
differentiate between opi nion columns and straight news articles, as the form er are more
likely to influence opinion than the latter.
In addition to adding internal validity to past survey results, my results also
sugge st that the pow er of opi nion columns exist outside of the election seasons. Not only
do opi nion columns influence voting de cisions (Dalton, Be ck and H uckfeldt, 1998; Beck
et al., 2002), but they also influence publ ic opinion on pol icy opt ions. Little research has
examined whether the media can influence publ ic opinion on pol itical issues or pol icy
options.8 The results presented here sugge st that the persuasive argum ents contained in
opinion columns can increase support for re form plans. It is unclear, how ever, whether
8 Past research sugge sts that how the media frames an issue can influence publ ic opinion (Nelson, C lawson
and O xley, 1997; Iyengar, 1991) .
22

competing opi nion columns in the same newspaper (advoc ating bot h sides of the debate)
cancel out that influence, but that is likely the case. Given the increased partisanship of
opinion sections (Dalton, Be ck and H uckfeldt,1998), c ompeting argum ents are more
likely in the news section ra ther than the opinion section of t he newspapers. This is
troubl esome, as the results presented here sugge st that competing argum ents are needed
more in the opinion sections.
A somewhat surpri sing re sult from my experiment is that salience has little
influence on the persuasive pow er of t he opinion columns or straight news articles.
Straight news articles and opi nion columns were no m ore persuasive on a low salience
issue (bankrupt cy reform ) than they were on a moderately salient issue (Social Security
reform ). I t ested whether the difference in media effects increased by runni ng the models
without interest in the issues as independent variables. If t he interest variable controlled
for the influence of salience across the two issues, I would expect the articles and opi nion
columns to be more persuasive in bankrupt cy than Social Security. This was not the
case, as the results were no di fferent from the results reported in Table 2. T he lack of
difference between the two models indicates that issue salience had little influence on the
persuasive pow er of opi nion columns.
My results also sugge st that media effects on publ ic opinion are better
characterized as persuasion than manipulation. O pinion columns influence publ ic
opinion be cause they expos e readers to persuasive argum ents. This is certainly not an
instance of m anipulation, e specially when the readers know that they are reading
someone’s opinion. O pinion columns and editorials simulate the deliberative proc ess by
expos ing pe ople to well-crafted argum ents concerning pol itical issues. Those reading
opinion columns held opi nions on the issues that were more consistent with their self-
23

identified ideological positions. Thus, reading opi nion columns increases one’s
likelihood of hol ding an opi nion that is consistent with their ideology. A possible
explanation for t his effect is that opinion columns provi de cues to the readers that allow
them to unde rstand how the issues relate to their predispositions. In ot her words, opinion
columns simplify the issues by c onne cting them to values and pre dispositions.
The persuasive pow er of t he opinion section ha s the potential to become
manipulation if only one side of an issue is given. Persuasive argum ents advoc ating one
position ne ed to be collided with persuasive argum ents on the oppos ing side. By
increasing the perceived importance of certain values concerning the issues, opinion
columns and editorials are framing the issues in certain ways. Framing effects have been
shown to influence opinions on pol itical issues when people are expos ed to onl y one
frame (Nelson, Cl awson and O xley, 1997; Druckman and N elson, 2003). O n the other
hand, w hen people are expos ed to competing fra mes there is little influence on publ ic
opinion (D ruckman, 2004). T he pow er of opi nion columns is very similar to the pow er
of fra ming; they bot h pre sent persuasive argum ents. Presenting opi nion columns on bot h
sides of pol itical issues is an important step in reducing the chance of m edia
manipulation.
24

1Bibliography
Arnold, D.A. 2004. Congr ess, the press, and pol itical accountability. Princeton:
Princeton U niversity Press.
Beck, P. A., Dalton, R. J ., Greene, S. and H uckfeldt, R. 2002. “ The Social Calculus of
Voting: Interpersonal, Media, and O rganizational Influences on P residential
Choi ces.” American Political Science Review 96(1): 57-73.
Bennett, W. L. 1996. News: The politics of Illusion. New York: Longm an.
Brody, R. A . 1991. Assessing the President: The media, elite opinion and publ ic
suppor t. Stanford: Stanford U niversity Press.
Dalton, R. J ., Beck P.A., and H uckfeldt R. 1998. “ Partisan Cue s and the Media:
Inform ation Flows in the 1992 P residential Election.” American Political Science
Review, 92(1): 111-126.
Day, K. April 15, 2005. “ Bankrupt cy Bill Passes; Bush Expected to Sign; For M any,
Erasing D ebt Would be Harder.” Washington Post, April 15, 2005, p. E 1.
Druckman, J.N. 2004. “ Political Preference Formation: Com petition, D eliberation, a nd
the (Ir)re levance of F raming Effects.” American Political Science Review, 98(4):
671-686.
Druckman, J.N., and N elson, K . 2003. “ Framing and D eliberation: How citizens’
conve rsations limit elite influence.” American Journal of Political Science 47(4):
729-45.
Druckman, J.N., and Parkin, M . 2005. “ The Impact of M edia Bias: How Editorial Slant
Affects Voters.” Journal of Politics. 67 (4): 1030-49.
Hetherington, M . J. 1996. “ The Media’s Role in Forming V oters’ National Econom ic
Evaluations in 1992.” American Journal of Political Science 40(2): 372-95.
Hovland, C. L ., Janis I.L., and K elley, H. H. 1953. Com munications and pe rsuasion:
Psychological Studies in opi nion change. New Haven, CT : Yale University
Press.
Iyengar, S. 1991. Is Anyone Respons ible. Chi cago: University of Chi cago Press.
25

Kahn, K . F., and K enney, P. J. 2002. “ The Slant of the News: How Editorial
Endors ements Influence Campaign Cove rage and Ci tizen’s Views of Ca ndidates.”
American Political Science Review. 96 (2): 381-94.
Kernell, S. H. 1997. Going Public: New strategies in presidential leadership.
Washington, D .C.: CQ Press.
Klapper, J. T. 1960. The Effects of Mas s Com munication. New York: Free Press.
Krugm an, P., “The Debt-Peonage Society.” New York Times, March 8, 2005, p. A 23.
Krugm an, P., “Three-Card M aestro.” New York Times, February 18, 2005, p. A 27.

Nelson, T . E., Clawson, R. A ., and O xley, Z. 1997. “ Media Framing of a Civil Liberties
Confl ict and Its Effect on T olerance.” American Political Science Review. 91 (3):
567-84
Page, B. I. 1996. Who D eliberates? Chi cago: University of Chi cago Press.
Porter, E., “Working out the num bers.” New York Times, February 6, 2005, S ec. 1, p. 15.
Prior, M . 2006. “ The Incumbent in the Living Room : The rise of television and the
incumbency advantage in U.S. House elections.” Journal of Politics 68(3): 657-
73.
Sears, D. O. 1986. “ College Sophom ores in the Laboratory: Influences of a narrow
database on ps ychology’s view of hum an nature.” Journal of Personality and
Social Psychology. 51 (3): 515-30.
Zaller, J. R. 1992. The Nature and O rigins of Mas s Opinion. New York: Cambridge
University Press.
Zaller, J. R. 1996. “ The Myth of a Massive Media Impact Revisited: New Support for a
Discredited Ide a.” In D . Mutz, P. Sniderman, and R. Brody, Political Persuasion
and A ttitude Change. Ann A rbor, M I: University of M ichigan Press.

26

Appendix A
Survey Questions
Social Security Opinion
Do you s upport or oppos e changing the current Social Security system to include
personal retirement accounts? Strongl y Oppos e = 1, 4 = No Opinion, S trongl y Support =
7
Bankruptcy Opinion
In 2005, ne w laws were passed that made it more difficult to file for ba nkrupt cy. Ba sed
on w hat you know about the changes to the past laws, do you s upport or oppos e those
changes? Strongl y Oppos e = 1, 4 = No Opinion, S trongl y Support = 7

Interest in the Issues
How closely do you fol low the following issues [Social Security or ba nkrupt cy reform ]?
Not interested = 1, Somewhat Interested = 4, V ery Int erested = 7
Ideology
Do you c onsider yours elf more conservative or m ore liberal in ideology? Liberal = 1,
Neutral = 4, Cons ervative = 7
Party Identification
Despite where you m ay rank yours elf in ideology, w hat party do you t ypically support ?
Democratic Party, Re publican Party, No Party
Age
What is your a ge?
Female
What is your G ender? Female = 1, M ale = 0
News Sour ces
How often do you ge t your ne ws from the following sources?
Television, N ewspaper, Int ernet, Radio.
27