|
ZHUJ DEBATES 93 RTVJ Images vs. Issues in Presidential Debates
Does the Audience Learn More about Images
than Issues from Televised Debates?
Effects of the First 1992 Presidential Debate
Jian-Hua Zhu, J. Ronald Milavsky and Rahul Biswas
Dept. of Communication Sciences
University of Connecticut
Storrs, CT 06226
Paper Submitted to the Radio and Television Division of
the Association for Education in Journalism and Mass
Communication,
April 1, 1993
ABSTRACT
This paper reviews all previous studies of debate effects on
issue learning and image formation, and reports the results of a
new study of the effects of the first presidential debate in the
1992 election. The study used a between-subjects design with
repeated measures. The study was replicated under both
experimental and natural viewing conditions. Results show that
the viewers learned a great deal about candidates' issue
positions that were discussed in the debate, but no learning
took place of issues that were not debated. The debate did not
affect the viewers' perception of the two well-known candidates'
personalities, but improved the perception of the less well-
known candidate, Ross Perot, on several debate-related
personality traits.
Does the audience of televised presidential debates learn
more about candidates' personal characteristics than their issue
positions? The answer to this question has important
implications for the political process. According to the
classic theory of democracy, a televised debate contributes
positively to the democratic process when it increases the
electorate's rational decision-making by informing them about
issues and where the candidates stand on them. On the other
hand, the debate makes a negative contribution to the extent
that it leads to greater learning about candidates'
personalities than it does about their stands on issues.
This latter function of debates has been criticized since
1960 when presidential debates were first televised. Based on
his observation of the 1960 debates between Kennedy and Nixon,
Lubell (1962) believed that many voters could not process the
complicated issues discussed in the debates, so they "settled
back and judged the debates as a personality contest" (p. 153).
He concluded that televised debates, a novel format at the time,
threatened the established political process because "the
debates tend to lessen somewhat the importance of issues and
party and to elevate the significance of personality,
particularly on its theatrical side" (p. 152). This view has
been echoed by other critics of debates (e.g., Lanoue and
Schrott, 1991).
LITERATURE REVIEW
To determine how much such criticism is supported by
empirical research, we have conducted a comprehensive, and
quantitative review of the debate literature. Among all the
published studies of presidential debates we could locate, 35
appear to be relevant according to our selection criteria: (1)
using exposure to the debate as an independent variable and (2)
treating issue learning and image formation as dependent
variables. A number of studies, although containing measures of
issue learning and/or image formation, were excluded from this
review because they either used issue/image as independent
variables, or did not include debate viewing as an independent
variable. Table 1 summarizes the major features of these
studies (the bibliographical list is reported in Appendix A).
--------------------
Table 1 about here
--------------------
Overall Trends in Debate Research
Column 1 of Table 1 reveals several overall trends in
debate effects research. The data show clearly that most of
what is now known about debate effects came from studies of the
first televised presidential debates in the 1960 election, and
from the debates in the 1976 election. The latter election
inspired the most debate effects studies, with 19 studies. Each
of the three elections in the 1980s inspired only 4 or fewer
studies. The increasing importance of the media, particularly
television, in election campaigns calls for continued study of
media effects. The new study, reported below, was designed to
determine if effects found in earlier studies replicate under
current conditions and to contribute further understanding of
the role of debates in elections.
Most of the studies reviewed were drawn from either
geographically local adult (16 studies) or college student
populations (13). Only four studies were based on national
general public samples. Two studies used selected respondents,
such as members of the League of Women Voters (Nimmo et al.,
1980), or females living in married-student houses (Tannenbaum
et al., 1962). The sample size was generally small or medium,
with the median being about 200. A quarter of the studies had
400 or more subjects or respondents, and a similar proportion
had 100 or fewer.
An overwhelming majority (31) of the studies focused on the
debates during the general election campaign. Only four dealt
with primary debates. This concentration on general debates was
somewhat compensated by the fact that most studies (28) examined
the first debate in a campaign, which has been found to have
greater impact than the subsequent debates (e.g., L. Becker et
al., 1980).
Methodologically, many studies share several common
features. Most studies (26) used surveys to collect data, and
only six were based on experimental designs in which exposure
was controlled or manipulated. Only 3 of these (Hawkins, et al.
1979; Lemert et al., 1983; Pfau, 1987) used a random procedure
to assign subjects to groups. Three studies (Bothwell and
Brigham, 1983; Lang and Lang, 1979, 1980) combined both survey
and experimental design features. From a research standpoint,
televised debates have one advantage over other campaign formats
(e.g., news coverage and political advertising): debates are
always a distinct event with an easily defined beginning and
end, whereas news stories and campaign commercials are
continuously unfolding throughout the entire campaign. Most of
the debate studies (30) apparently took advantage of this by
using a before-after design. Still there were five studies
using an after-only design, in which cases the impact of a
debate would be confounded by the audience's prior knowledge or
attitudes.
Confounding problems may also exist among the 30 studies
with a before-after design. About one third of them (9) did not
measure whether the audience actually watched a debate. By
comparing issue/image scores measured before and after a debate,
some studies basically assumed that the debate was the single
source of the change, or lack of it.1 Such an inference is
especially problematic when the two points of measurement were
one week or more apart (e.g., L. Becker et al. 1979; Hagner and
Rieselbach, 1980).
An equally critical problem of confounding effects is that
three-quarters (23) of the before-after studies did not include
a control group (i.e., nonviewers of a debate). Most of these
studies administered a pre-test a few days in advance (to offset
any sensitization effect), and a post-test right after the
debate (to eliminate post-debate influences). However, the
audience might still be influenced by other sources of campaign
information between the pre- and post-tests. This "contaminated
learning" effect can be controlled by comparing change scores
between viewers and nonviewers. Only seven studies did so,
however.2
Another general weakness of the studies under review is the
failure to link the audience's responses to actual debate
content. Like other communication events, some debates can be
linked to specific messages presented in the debate while others
may not have any specific debate message references. In
general, content-specific effects should be given more weight,
because they are more interpretable and less subject to
confounding influences. In addition, references to specific
messages in a debate can be found readily, either by reading a
transcript or reviewing visual tape. Unfortunately, only 7 out
of the 35 studies took advantage of this. Without considering
the original debate content, the remaining studies used a
variety of other measures, which are more subjective (to be
discussed below), to assess debate effects.
Variants of Issue/Image Measures
Even though we used a fairly narrow definition of debate
effects to select the 35 studies, there were still diverse
measures of issue learning and image formation in these studies.
We have identified at least 7 variants of issue learning, 4
variants of image formation, and 4 variants of other attitudinal
effects. The variants of issue learning include issue knowledge
(i.e., the audience's judgment of a candidate's issue position,
the accuracy of which can be verified with actual debate
content); issue opinion (a subjective measure of the audience's
own views on an issue); issue congruency (the similarity between
the audience's own position on an issue and a candidate's
position on the issue as perceived by the audience); issue
discrimination (the difference between the perceived positions
take by two or more candidates); audience's issue salience (the
audience's own perception of the importance of an issue, which
is a standard agenda-setting question), candidate's issue
salience (the audience's perceived issue importance to a
candidate), and issue-salience congruency (the distance between
the audience's own issue salience and his/her perceived
candidate issue-salience).
The variants of image formation are image perception (the
audience's rating of a candidate on certain personality traits),
image congruency (the difference between the perceived image of
a candidate and the perceived ideal president), image
discrimination (the difference in the perceived traits of two or
image salience (the audience's perceived
importance of certain traits of a candidate to his/her vote
decision). In addition, many of the studies also measured other
attitudinal effects of watching a debate, such as affective
response toward a candidate, candidate choice, intention to
vote, and interest in the campaign.
It turned out that 21 studies had at least one issue
measure and 27 had at least one image measure. Two studies
(Lang and Lang, 1979; McLeod et al., 1979) measured three or
more variants of issue learning, and another two studies (Miller
and MacKuen, 1979; Tannenbaum et al. 1962) had three image
measures. Overall, four measures stood out as the most
frequently used: image perception (by 23 studies), issue
discrimination (9), issue knowledge (8), and image
discrimination (6). Two of these measures, issue discrimination
and image discrimination, deserve further consideration, because
we believe that they were inappropriately used in studies of
presidential debates.
The theory behind the two discrimination measures is that
an informed citizenry rests on "cognitive discrimination
ability" to distinguish two candidates' positions or
personalities (Choi and Becker, 1987; Clarke and Kline, 1974),
and that a televised debate is effective if it helps polarize
two candidates in the audience's mind (Miller and MacKuen,
1979). When applying the theory to presidential debates, these
studies implicitly made two strong assumptions: First, that the
two competing candidates indeed have clear-cut differences in
issue positions. Second, that the audience's perception of two
candidates' personalities is a "zero-sum" process in which the
rise of one's image is always at the price of the competing
candidate. Both assumptions are questionable in light of
empirical evidence.
In media-dominated modern politics, viable presidential
candidates have to be middle-of-the-road to appeal to a large
base of constituents. Consequently, the major candidates tend
to share some similar centrist positions. As will be described
in detail later, we found in the 1992 election that the three
major candidates had similar stands on a number of issues. For
example, they all favored retraining defense industry workers
and opposed sending troops to Bosnia. It is possible that an
audience guesses where a candidate stands on some issues based
on general assumptions about such matters as the candidate's
party affiliation or past record. Such guesses might lead to
the belief that the candidates are farther apart than they
really are. This would show up in a pre-debate study as large
(but false) issue discrimination. After the debate in which
candidates discuss the issue, the similarity of their position
becomes better known, and the post-debate measures accurately
show a decrease in issue-discrimination. Positing issue
discrimination as a theoretical debate objective would lead to
the conclusion that the debate failed to educate, when it in
fact did.
In addition, there also has been ample evidence from the
debate literature that all candidates usually improve their
images among the audience in the same debate (e.g., Simons and
Leibowitz, 1979). Based on such findings, Lanoue and Schrott
(1991, p. 120) concluded that the audience formed images of each
candidate in a non-zero sum manner. For this reason, too, a
televised debate could lead to convergent, rather than
polarized, images of candidates, which again would be considered
a non- or failed-effect within the discrimination ability theory
framework.
Debate Impact on Issue Knowledge and Image Perception
Because of these problems, we decided to focus on issue
knowledge and image perception. Both are direct measures of
debate effects. The accuracy of issue knowledge also can be
objectively verified by reference to debate content. Therefore,
they have higher validity than the discrimination measures. Our
analysis of previous studies shows that issue knowledge has been
used in only 8 of the 35 studies, compared to 23 which measured
image perception. Columns 2 and 3 in Table 1 display a
comparison of these studies.
A few common features emerge from both issue knowledge
studies and image perception studies: Both were more likely to
use local adult and student samples, and were also more likely
to focus on the first debate in the general election. These
similarities were outnumbered by the differences between the
two, however. One difference lies in an overall trend. While
image perception studies were done in most elections, recent
studies (1980 and thereafter) were more likely to include issue
knowledge. This is in line with the general trend in mass
communication research toward paying more attention to cognitive
effects (Severin and Tankard, 1992).
The two types of studies appear to be quite different on
several methodological dimensions. For example, the studies of
issue knowledge were more balanced between survey and experiment
(4 vs. 4) than the studies of image perception (20 vs. 3 in
favor of survey). Exposure was measured in all studies of issue
knowledge, but only in 14 out of 23 studies of image perception.
Half of the issue knowledge studies included a control group
(i.e., nonviewers), while the control group was used by only a
quarter of image perception studies. Three studies of issue
knowledge compared the audience's responses to actual debate
content, but none of image perception studies attempted to do so
(of course, image-related content is more difficult to trace).
The only strength that image perception studies have over issue
knowledge studies is that the former are more likely to use
before-after design than the latter. Overall, then, the studies
of issue knowledge appear, though fewer, to be based on firmer
methodological ground than the studies of image perception.
With these methodological differences in mind, let us take a
look at the relative effects televised debates have had on issue
knowledge and image perception.
Because of the diverse measurements and analyses used in
these studies, we developed a classification scheme to
categorize the debate effects found. The effects are considered
"substantial/clear" if exposure to a debate (or its surrogate
such as time) produces statistically significant, and consistent
(in the case of multiple analyses), effects on issue knowledge
or image perception. The effects are coded "none" if exposure
is consistently nonsignificant. Many studies turned out to lie
in between, and are labelled "moderate" in effect size, "mixed"
across analyses, or "maybe" if the original author(s) only
provided speculative descriptions. According to this
classification, 3 of the 8 issue knowledge studies found
televised debates to have substantial/clear effects, and 4 other
knowledge studies found moderate/mixed effects. On the other
hand, substantial effects were found in 5 out of 23 image
perception studies, and moderate/mixed effects were found in
another 10 image studies.
Note that not all these studies examined debate effects on
issue knowledge and image perception simultaneously. In fact,
only 5 studies included measures of both issue knowledge and
image perception. Of them, three (L. Becker et al., 1980; Drew
and Weaver, 1991; McLeod et al. 1979) reported that the impact
of exposure to a debate was greater on issue knowledge than on
image perception. One study (Lang and Lang, 1979) found the
debate impact between issue knowledge and image perception to be
similar. In only one study (Carter, 1962), a debate was found
to produce a greater impact on image perception than on issue
knowledge.
Summary
Several generalizations have emerged from this quantitative
review of the debate effects research literature: First, the
direct evidence is quite limited about the relative impact of
televised debates on issue learning versus image formation,
because most of the studies either have used inappropriate
measures of the two effects, or have not compared the two
simultaneously. The five studies that did conduct direct
comparisons suggest that issue knowledge learning is greater
than image perception formation. The findings are inconclusive,
however, given the small number of studies and the
methodological weaknesses mentioned above.
Based on almost the same literature reviewed here, Lanoue
and Schrott (1991, p. 96) concluded that "In reality, though,
viewers are far more likely to use debates to gain insight into
each candidate's personality and character. ... a superior
'personal' presentation appears to be more important to voters
than accumulation of issue-oriented debating 'points'." Our
analysis suggests that this generalization is at best only very
weakly supported by the existing data, if at all.
This literature review also suggests several ways to
improve studies of debate effects. Specifically, there appear
to be four criteria for a minimally necessary and sufficient
test of debate effects on issue knowledge and image perception:
(1) the basic design should be a between-subjects design
contrasting viewers and nonviewers, (2) the dependent variables
should be measured both pre- and post-debate, (3) the results
should be verified with content analysis of the debate to
ascertain content-specific effects, and (4) the data should be
collected from both the forced viewing condition (experiment, to
assure that exposure does take place) and the natural viewing
condition (survey, to offset laboratory artifacts). Some of
these issues have been noted by Joslyn (1984) and Lanoue and
Schrott (1991) as well. The study reported below was designed
based on these requirements.
METHOD
Subjects
The context of this study was the first 1992 presidential
debate between George Bush, Bill Clinton, and Ross Perot, which
took place on October 11, 1992. The subjects are 185
undergraduate students enrolled in introductory courses in
communication and public speaking at a state university in
Connecticut.3 The students received class credits for
participation in the experiment. Slightly more than half of the
subjects (58%) were female. Roughly one third were freshmen,
one third sophomores and the remaining juniors or seniors.
About half of them had either a part-time or full-time job.
As in many other experimental studies using student
subjects, our student sample does not represent the general
population. However, the resulting data suggest that our
student sample is comparable to the general electorate in
several important aspects. For example, 32% of the students
described themselves as "very interested," 54% "moderately
interested" and 13% "not interested" in national politics.
These proportion are comparable to a national survey done during
the same period.4 Likewise, 77% of the students said that they
had registered to vote in the election, and 83% of these
intended to vote, as compared to about 80% of the general
population being "likely voters," defined somewhat differently
by various national pollsters in early October, 1992.5 The
proportion of Democrats, Republicans, or Independents among the
students is similar to the pattern among the general public in
Connecticut.6
In addition, the students' knowledge about candidates'
issue positions, one of the key variables under study, was
similar to the national level. Before the debate, the students
answered correctly, on the average, 29% of 21 knowledge
questions about the three candidates' issue positions. This
knowledge score was very close to the 32% correct rate of all
knowledge questions, or 27% of "non-trivial" questions, reported
by Justin Lewis and Michael Morgan (1992) based on their
nationwide sample interviewed in the same period.
Procedure
During the recruitment, the study was described as an
"election" study. The debate was not specifically referred to
in any way. The students were asked to participate in two
sessions, three to five days apart. The first session (pre-
test) was held between October 7 and 9, two to four days prior
to the debate . The subjects came at a time of their choice to
answer the pre-test questionnaire (to be described in the
measurement section), which lasted about 20 minutes.
The post-test took place either on Sunday evening, October
11 when the debate aired, or throughout the following day.
According to their appointment made before the pre-test, 53
students came to the post-test on October 11. The first time
students became aware that the presidential debate was involved
in the study was upon arrival for the second session when they
were asked to watch the 90-minute long telecast. Right after
the debate, they answered the post-test questionnaire, which was
identical to the pre-test. This group of 53 will be called
"experimental viewers."
When the remaining 132 subjects arrived next day, we first
asked them a series of filter questions to determine whether
they had watched the debate on TV, listened to it on radio at
their own initiation, or had watched, heard or read any news
reports about the debate. Roughly half of them (65) had watched
or listened to the debate,7 while the other half (67) had
neither watched nor been exposed to news about the debate. We
randomly assigned, in a ratio of 2 to 1, the uncontaminated
subjects into a newspaper group (46) and a control group (21).
The newspaper group was asked to read the full transcript of the
debate printed in The New York Times of the day, while the
control group was not. Both groups then completed the post-test
questionnaire. The newspaper group was included for another
purpose unrelated to the current study, and will be excluded
from the analysis reported below. The basic analysis will
compare the experimental group with the "uncontaminated" control
group. For convenience, we will call the members of the control
group "nonviewers."
The "contaminated" subjects who had watched the debate on
the previous evening at their homes or dormitories also answered
the post-test questionnaire. They will be labeled "natural
viewers." In the filter-questions, we asked them to indicate
which of six 15-minute segments of the debate they watched, and
which other activities they engaged in while watching, from
which we derived a measure of attention to the debate. These
questions were later combined to form an exposure measure. The
experimental subjects and the natural subjects were analyzed
separately. Thus the analysis of natural viewers provides an
independent check on the effects of experimental manipulation.
It should be noted that the assignment to the initial
experimental exposure condition was not strictly random. At the
beginning of the study, each subject was asked to choose a time
for both pre- and post-tests. While the selection of the pre-
test sessions did not affect the outcome at all, the choice of
October 11 for the post-test session determined placement in the
experimental viewing condition, while the choice of October 12
placed them in either the newspaper, control, or natural viewing
conditions. Of course, no subject was aware of this because the
sign-up occurred long before they even came to the pre-test, and
in addition, the subjects had no way of knowing that the debate
was involved at all. Therefore, such assignment can be called a
"blind self-selection" procedure, as opposed to the "conscious
self-selection" procedure used in almost all previous
experimental studies in which the subjects were told that a
debate was being studied and allowed to choose a session to
attend.
However, there was still a possibility of systematic
selection biases. To detect these, we compared various pre-test
items between experimental viewers, nonviewers and natural
viewers. Since the pre-test was taken prior to exposure to the
stimulus (i.e., the debate), significant difference in the pre-
test scores among these groups would indicate a biased
assignment. Otherwise, we could reasonably assume no systematic
bias involved in the assignment, and any significant difference
detected in the post-test could be attributed to the stimulus.
--------------------
Table 2 about here
--------------------
As shown in Table 2, except for political interest, we did
not find any significant difference among the three groups, or
between any pair of them, in every important characteristics in
the pre-test. The subject's sex, class standing, party
affiliation, election participation, and media use habits were
similar across groups. The groups differed on political
interest, however, with the natural viewers being most
interested in national politics, which explains at least in part
why they voluntarily watched the debate. Therefore, political
interest was controlled in the final analysis. In addition,
since there was the possibility of contingent effects, other
variables, such gender, employment, and party affiliation, were
also introduced into the final analysis as control variables.
Equally similar were their scores on the key dependent
variables (to be described later): knowledge of candidates'
issue positions, perception of candidates' characteristics, and
vote preference. The only exception among the 21 comparison is
the perception of Perot's communication style, in which the
natural viewers scored significantly higher than the other two
groups. This did not affect the final results because we only
compared experimental viewers with nonviewers, or heavy natural
viewers with light natural viewers.
Measures
As mentioned above, pre- and post-test questionnaires
contained a set of identical questions, which could be divided
into four sections: (1) perception of candidates'
characteristics; (2) knowledge of candidates' issue positions,
(3) rating of the salience of election issues and the
candidates' performance on these issues, and (4) the subject's
personal background information. Because section 3 will not be
used for the current study, and section 4 is self explanatory,
we will describe below only the first two sections.
Issue knowledge. To test the effects of all specific
messages presented in the debate on the subject's knowledge, we
would have had to know in advance exactly what issues would come
up in the debate, which was of course impossible. After
consulting various election news reports, we complied a list of
24 "most-likely-to-be-debated" items for both pre- and post-
tests. These items asked the subject to indicate whether a
candidate supported, opposed, or had no position on 24 specific
policy issues. The same 24 items were asked for Bush, Clinton,
and Perot, respectively. Therefore, the total number of issue
knowledge items is 72 (24 items by 3 candidates).
For obvious reasons, we did not anticipate that all three
candidates would discuss all these issues. On the contrary, it
would increase the power of the study if a portion of the issues
would not be addressed. It would enable examination of
differences between the debated and undebated issues.
Theoretically, learning should occur only of the debated issues,
not of the undebated issues. If there was knowledge gain in the
latter, we should suspect the presence of some confounding
influences.
Our content analysis of the debate transcript shows that 21
out of the 72 items were explicitly debated: 8 were addressed
by Bush, 9 by Clinton, and 4 by Perot. Although these items
accounted for less than one third of the original list, there
are sufficient numbers to form a composite knowledge score for
each candidate. Specifically, based on actual statements made
by the candidates in the debate, we gave each correct answer a
score of "1," and summed them for each candidate respectively
(the wording and the key for these knowledge items are in
Appendix B). Because the total number of issue items varies
from one candidate to another, we converted the scores into
percentages, to make the knowledge score about each candidate's
positions comparable.
Image perception. The research literature reports a wide
variety of measures of presidential candidate images. We chose
to follow a five-dimensional scheme developed by Miller and his
associates (Miller, Wattenberg, and Malanchuk, 1984; Rusk and
Weisberg, 1972). Our choice of this scheme was based on several
considerations. First, its external validity has been tested by
replications within many national probability samples. Second,
its measures are derived from responses to open-ended questions.
In other words, the assessment items are actually used by
ordinary voters, rather than imposed by researchers (see Clarke
and Kline, 1974, for a criticism of researcher-oriented
measures). Third, the resulting cognitive structure of
candidate assessments has been verified by eight National
Election Studies over 32 years (1952-80). To quote Miller et
al. (p. 208), "The empirical evidence ... provides consistent
support for the theory that Americans evaluate presidential
candidates on the basis of a limited set of general and enduring
criteria. In short, people have a prototype of what a president
should be like and they judge real candidates by these
standards."
Miller et al. labeled the five dimensions as "competence,"
"integrity," "reliability," "charisma," and "personal."
Competence refers to a candidate's past political experience,
comprehension of political issues, and intelligence. Integrity
includes remarks such as trustworthy, honest, and sincere.
Reliability is a label that does not well represent the traits
involved, such as strong, decisive, and aggressive. The authors
distinguished reliability from integrity in this way (p. 197):
"Reliability thus serves as a bridge between integrity and
competence. Perhaps the best definition of it [i.e.,
reliability] would be as trust in [a] capability sense rather
than in the honesty sense." We have adopted the term "potency"
to replace "reliability" in this study. Charisma (called
"leadership" in an earlier version, Miller and MacKuen, 1979)
deals with a candidate's ability to get along with and
communicate with people. Personal is a catch-all category that
includes various observable features of a candidate's
appearance, such as age, health, smile, and speech pattern. In
our study, we decided to call the last dimension "communication
style." Miller et al. (1984) report that although all of these
dimensions involved in the way voters judge presidential
candidates, competence accounts for the most variance, followed
by integrity. Charisma seems to be the least important among
the five dimensions.
In both pre- and post-test, we asked the subjects to rate
each of the three candidates along 34 bipolar items,
representing various specific personality traits. The
evaluations were scored on 7-point scales with 7 being most
positive and 1 most negative. The resulting scores were used to
fit a five-dimension model. Since the model will repeat 6 times
(=3 candidates x 2 points), not all of the 34 items fit all 6
models. Some items were salient only in the pre- or post-test.
Others items were salient only for a particular candidate.
Still other items loaded on almost all five dimensions across
all candidates and all time points. Therefore, we selected 3
"best" items for each of the dimensions, or a total 15 items for
each candidate at each point (see Appendix C for the wording of
these items). This resulted in a quite satisfactory structure,
in which each of the five image dimensions for each candidate at
each point is made up by the same items, with a high internal
consistency score ( Cronbach's alpha) around .80 (Appendix C).
Analysis
Our research question is fairly straightforward: whether
the debate has a greater impact on image perception than issue
knowledge, or vice versa. This question can be answered by
simply comparing the coefficients of the independent variable
(i.e., exposure to the debate) regressed on the two dependent
variables (knowledge scores and image scores). The coefficients
can be obtained from either an ANOVA approach which uses the
change score of knowledge and image from pre-test to post-test
as the dependent variable, or a regression approach which uses
the post-test score as the dependent variable and the pre-test
score as an covariate. We chose to follow the regression
strategy because change score has limiting statistical
properties (Heilzer, 1959; Rogosa, Brandt and Zimowski, 1982).
On the other hand, regression analysis with pre-test score as a
covariate can precisely measure the impact of pre-test on post-
test (i.e., autocorrelation), which is often the best
explanation for the lack of debate effects.
For each candidate, 7 regression analyses were performed,
each with one of the following post-test scores as the dependent
variables: (1) knowledge of a candidate's position on the
issues discussed by the candidate in the debate; (2) knowledge
of the candidate's position on the issues he did not discuss in
the debate; (3) perception of the candidate's competence, (4)
potency, (5) integrity, (6) charisma, and (7) communication
style.
Exposure to the debate was the independent variable in
every regression analysis. Only one pre-test score (e.g.,
Perception of Bush's Competence at Time 1) was entered in the
corresponding regression (using Perception of Bush's Competence
at Time 2 as the dependent variable). Gender, Employment, Party
Affiliation, and Political Interest were included as control
variables in every regression equation.
The same seven regressions were repeated for three
candidates, and for both experimental sample and natural viewer
sample. In total, 42 (=7 x 3 x 2) regressions were conducted.
For simplicity, we will report in Table 3 only the two central
coefficients for each regression: exposure and the
autoregressor (i.e., pre-test score). The coefficients of all
other control variables and the overall model statistics are
reported in Appendix D.
FINDINGS
--------------------
Table 3 about here
-------------------
Learning Debated Issues
The evidence is unambiguously strong for the effect of
watching the debate on learning the issue positions discussed by
the candidates in the debate. As the top panel ("Debated
Issues") in Table 3 shows, after controlling for gender,
employment, political interest, and party affiliation (not shown
here), those watching the debate under the experimental
condition know, on the average, 34% more about Bush's issue
position, 24% more about Clinton's, and 39% more about Perot's
position, as compared with the nonviewers. The learning is
quite impressive, if one considers their knowledge level prior
to the debate. Recall from Table 2 that, before the debate, the
experimental viewers correctly answered 22% of Bush's positions,
39% of Clinton's, and 11% of Perot's, all of which were not
different from the nonviewers' prior knowledge level.
The same learning effects are detected among the natural
viewers. An explanation is in order to help interpret the
coefficients of exposure for both the experimental and natural
samples. In the experimental sample, exposure is simply a dummy
variable, coded 1 for the viewers and 0 for the nonviewers. The
coefficient, therefore, indicates the average difference in
learning between the viewers and the nonviewers, other things
being equal. In the natural sample, as described before,
exposure is an interval scale comprised from the number of 15-
minute segments of the debate watched and the level of attention
paid to the debate. To make the coefficients of exposure
compatible between the two samples, we scaled the natural
viewers' exposure to the range from 0 to 1, with 0 being minimal
watching plus least attention, and 1 being maximum watching plus
most attention. Therefore, the coefficient of 38.572 in column
4 of first line indicates that the most attentive viewers
learned 39% more about Bush's issue positions than the least
attentive viewers under the natural viewing condition, other
things being equal. Qualitatively, the difference between the
most attentive viewers and the least attentive viewers in the
natural sample is not exactly the same as the difference between
the experimental viewers and the nonviewers. However, as the
top panel ("Debated Issues") of Table 3 shows, the differences
appear to be comparable at least quantitatively. The
corroboration between the experimental data and the survey data
seems to rule out an artifactal effect from the experimental
stimulus.
It is interesting to note that while prior knowledge
strongly affects the learning of Bush's or Clinton's issue
position, the autocorrelation is nonsignificant for Perot who
was little known before going into the debate. The absence of
autocorrelation explains why exposure to the debate has the
strongest effect on learning Perot's issue stance, among both
experimental and natural viewers. Conversely, exposure has the
relatively weakest impact on both experimental and natural
viewers' learning of Clinton's position.
The statement that the audience learned most about Perot's
issue position but least about Clinton's position from the
debate may obscure an important fact, however. The reason the
viewers acquired the least issue knowledge about Clinton is
simply that they had been most familiar with his positions prior
to the debate (see Table 2). When considering both the baseline
knowledge and the acquired knowledge, we can see an interesting
pattern of convergence (Figure 1). For the viewers, despite
quite different starting values, the debate "equalizes" the
level of knowledge about all three candidates' issue positions.
This pattern does not hold for the nonviewers who, as should be
the case, knew more about Clinton's issue position than about
the other two candidates both before and after the debate.
--------------------
Figure 1 about here
--------------------
Learning Undebated Issues
While the viewers, either under an experimental condition
or a natural setting, learned a great deal of the candidates'
positions on the issues debated, they ought to have learned
little about the undebated issues. As the second panel
("Undebated Issues") of Table 3 shows, neither the experimental
viewers nor the natural viewers significantly improved their
knowledge of any of the candidates' positions on any issue not
discussed in the debate. This finding offers an additional
validity check for the effect of the debate on learning the
debated issues described above. In other words, the impact of
debate on issue learning is only content-specific. Television
does not have the magic power to generalize the knowledge gain
beyond what is said in the debate.
Image Perception
The effect of watching the debate on the students'
perception of candidates' character is much more limited. As
the last five panels of Table 3 show, the impact of watching the
debate is largely confined to the perception of Perot's
character among the experimental viewers. An exception is that
the experimental viewers also have a more positive view of
Bush's communication style than the nonviewers, other things
being equal. No difference can be found in Bush's other four
image dimensions. Nor is there any difference in any of
Clinton's five image dimensions. For the natural viewer sample,
exposure to the debate virtually does not have any impact on any
image dimension for any candidate.
Nevertheless, the effect of watching the debate on Perot's
image is substantial among the experimental sample. Other
things being equal, the experimental viewers scored him about
one to one and half points (on a 7-point scale) higher than the
nonviewers for his integrity, charisma, and communication.
However, there is no difference in perceptions of Perot's
competence and potency, the two dimensions which may be more
affected by a candidate's performance over a total career than
his performance in a 90-minute debate (Miller et al., 1984).
Why did the debate fail to bring about any significant
change to Bush's or Clinton's images? The answer can be found
in the stability of image perception, as indicated by the high
autocorrelation coefficients between pre- and post-test image
scores which range from about .50 to .80 for Bush and Clinton,
respectively. The strong high autocorrelation shows that the
subjects had already formed fairly firm perceptions of the two
well-known politicians, after having been exposed to the nearly
year-long campaign. Consequently, there should be minimal room
for change.
CONCLUSIONS AND DISCUSSION
Summary of Previous Research
A central concern of previous researchers of effects of
presidential debates on the electorate has been whether the
debates contribute to learning about issues in the campaign and
whether they increase the salience of candidates' personality in
voter decision making. The literature review reported here has
found that a variety of research designs, concepts and measures
have been used to study the effects of presidential debates.
The review also has found that image formation has received
attention in more studies than has issue learning, but that the
trend is toward more study of issue learning. Few studies have
focused on both issues and images. Generally, the methods used
in study of issues appear to be somewhat stronger than those
used to study images.
Reviewing the variety of methods used, we have found that
surveys have outnumbered experimental designs and that when
experiments were used, they seldom have employed random
assignment. Surveys have rarely employed national probability
samples. Sample sizes tend to be small in both surveys and
experiments. And most studies have used samples of students in
both surveys and experiments. Most studies used a before-after
design without a control group (i.e., "within-subjects design).
A few studies even did not measure actual exposure to the debate
studied. Results of these studies reflect the great variety of
methods and approaches used. Nevertheless, most studies have
found at least some effects of debate on issue learning, and
more mixed effects on image formation.
Results In This Study
This study has found that the audience learned a great deal
about the candidates' issue positions from the first
presidential debate in 1992 election. The issue learning took
place for all three candidates, with differential rates. The
audience learned the most about the least known candidate,
Perot, but learned the least about the most known candidate,
Clinton. In the end, the debate equalized the level of issue
knowledge across the three candidates. In addition, the issue
learning occurred only to those issues discussed by the
candidates in the debate. In other words, only message-specific
effects have been found in this study. These patterns hold up
in both the experimental and natural viewing conditions, and are
consistent with expectations based on results of previous
research.
The results pertaining to the effects of the first
televised presidential debate on image formation were mixed,
which also is consistent with previous studies. The audience's
perceptions of the two well-known candidates, the Incumbent
president Bush, and the Democrat Party candidate Clinton who had
run an effective campaign for almost a year, were not affected
by the debate. Exposure to the debate, however, did
substantially change the viewers' minds about the least known
and least popular candidate, Ross Perot, in the positive
direction. Perot had entered the race late (July 1992), had
campaigned, withdrawn from the race, and recentered just before
the first debate. As late as in early October, 1992, public
opinion polls showed that a sizable portion of the electorate
still were unsure, or negative about him.8 With these
conditions providing the context, his performance in the debate
could have been critical and our study results showed that the
debate helped him improve his image considerably. However, of
the five image dimensions we measured, the debate was more
likely to affect the perception of three debate-related
performance dimensions (i.e., integrity, charisma, and
communication style), and two other dimensions which perhaps are
better indicators of performance in office -- competence and
potency.
At least two important generalizations might be drawn from
these results: First, debates are least effective in changing
the public's mind about the character of a presidential
candidate if he is well known; they are most effective when
candidates are not well known. This generalization is
consistent with findings reported in Table 1 that show greater
impact in primary debates (in which several candidates, at least
some of whom are not well known, compete for public attention)
than in general election debates (in which a smaller number of
well-known candidates participate).
Second, when a candidate's image is changed by performance
in a debate, the change occurs in dimensions of personality that
are more evident in a debate format, and not in dimensions that
are more relevant to performance in office. This
generalization, if it holds up in future studies, could ease the
concern of those who believe that a debate has potential to
introduce irrational or non-rational considerations into a
presidential campaign by shifting focus from issue
considerations to personality considerations. It suggests that
the public does not leap to conclusions about capabilities as
president from proficiency as debater.
Note that these findings are based on a fairly rigorous
research design. Contaminated learning from external sources
has been controlled by contrasting the change scores between the
viewers and the nonviewers. Experimental artifacts have been
offset by replicating the same measures in an experimental
sample and a natural viewer sample. Knowledge learning has been
verified with content analysis of actual debate transcript.
Measures of image perceptions were developed based on a well-
established literature, and have proven to have high reliability
in this study. Given these characteristics, we believe the
findings reported above are substantial and reliable.
Based on our findings, we would like to draw several
general comments on the implications of televised debates for
the democratic process. First, debates appear to make positive
contributions to rational vote decision-making by helping the
audience learn substantial amounts about the candidates' issue
stands. Note that the 1992 first presidential debate took place
less than four weeks before the election. However, as shown by
Lewis and Morgan (1992) and this study, the electorate had, on
the average, only a 20% rate of accuracy about candidates' issue
positions at that point in the campaign. Our study suggests
that the first debate improved the accuracy rate over 50%.
On the other hand, there seems to be less convincing
evidence suggesting the fear that televised debates may have a
negative impact on the democratic process by focusing the
audience's attention on the candidates' personalities and
character. As found in this study, debates do not have much
impact on the audience's perception of well-known candidates.
This should ease much of the concern about the negative impact
of debates, because the presidential candidates are usually
well-known by the time a debate takes place. As this study
shows, when a less well-known candidate, usually an independent
candidate, surges in the campaign, debates can enhance the
candidate's images.
Where this debate did influence the electorate's perception
of a less well-known candidate, it was on traits that are
relatively more related to debate performance, such as charisma
and communication style, than on relatively job-related
performance traits, such as competence and potency. However, as
Miller et al. (1984) found, when it comes to the vote decision,
the job-related performance dimension (i.e., competence) carries
more weight than the debate-related dimensions.
A final caution about the limits of our study is in order
here. We used a student sample, and measured only short-term
effects (i.e., right after exposure to the debate). While our
design improves various shortcomings in the previous studies, it
is still not clear whether our findings hold up in the general
electorate and in the long-term (e.g., a week or more after the
debate).
ENDNOTES
__________________________________________
1 A study by Hagner and Rieselbach (1980) measured exposure, but
did not use it to test the impact of debate for some unknown
reason. Instead, they inferred the debate effects by
comparing change scores between September and November.
2 Bothwell and Brigham (1983) had a group of nonviewers in their
sample, but excluded it from the analysis of debate effects.
We have considered this as a study without a control group.
3 Another 23 participating in the pre-test but did not come back
for the post-test. They are excluded from this study.
4 Conducted by the Los Angeles Times from October 2 to 5, 1992,
achieved at the Roper Center for Public Opinion Research of
the University of Connecticut). In the survey, 49% reported
"very interested," 43% "somewhat interested," and 7%
"very/somewhat uninterested" in following the election
campaigns.
5 For example, Gallup found 79% of those surveyed between
October 1 and 3, 1992 were registered voters. The findings
are achieved at the Roper Center.
6 See a survey conducted by Hartford Courant and Institute for
Social Inquiry of University of Connecticut between August 27
and September 2, 1992, which is achieved at the Roper Center.
7 In fact, only one subject listened to the debate on radio,
while the other 64 watched it on TV. For convenience, we will
not distinguish the listener from the rest.
8 For example, CBS News/New York Times poll between October 2
and 4, 1992 found that while 7% of the registered voters were
favorable about Ross Porot, 59% were not favorable. Another
36% were either not sure or did not hear enough about Ross
Perot. The poll results are achieved at the Roper Center of
the University of Connecticut.
References
Choi, Hayeon C., and Samuel L. Becker (1987). Media Use,
Issue/Image Discriminations, and Voting. Communication
Research, 14, 267-291.
Clarke, Peter, and F. Gerald Kline (1974). Media Effects
Reconsidered: Some New Strategies for Communication
Research. Communication Research, 1, 224-240.
Heilizer, Fred (1959). Some Cautions Concerning the Use of
Change Scores. Journal of Clinical Psychology, 15, 447-
449.
Joslyn, Richard A. (1984). Mass Media and Elections. Reading,
MA: Addison-Wesley.
Lewis, Justin, and Michael Morgan (1992). Images/Issues/Impact:
The media and campaign '92. Unpublished report. Elmhurst,
MA: The Center for the Study of Communication, University
of Massachusetts.
Lanoue, David J., and Peter R. Schrott (1991). The Joint Press
Conference: The History, Impact, and Prospects of American
Presidential Debates. New York: Greenwood Press.
Lubell, Samuel (1962). Personalities vs. Issues. In Sidney
Kraus (Ed.), The Great Debates: Background - Perspective
-- Effects. Indiana University Press, 151-162.
Miller, Arthur H., Martin P. Wattenberg, and Oksana Malanchuk
(1984). Cognitive Representations of Candidate
Assessments. In Keith R. Sanders, Landa L. Kaid, and Dan
Nimmo (Eds.), Political Communication Yearbook, 1984.
Carbondale, IL: Southern Illinois University Press. 183-
210.
Miller, Arthur H., and Michael MacKuen (1979). Learning about
the Candidates: The 1976 Presidential Debates. Public
Opinion Quarterly, 43, 326-346.
Rogosa, David, David Brandt and Michelle Zimowski (1982). A
Growth Curve Approach to Measurement of Change.
Psychological Bulletin, 92, 726-748.
Rusk, J.G., and H.F. Weisberg (1972). Perceptions of
Presidential Candidates: Implications for Electoral
Change. Midwest Journal of Political Science, 16, 388-410.
Several, Werner J., and James W. Tankard, Jr. (1992).
Communication Theories: Origins, Methods, And Uses in The
Mass Media, 3rd Ed. New York: Longman.
TABLE 1
STUDIES OF DEBATE EFFECTS ON ISSUE LEARNING AND IMAGE FORMATION
Total Issue Image
Studies Knowledge Perception
(N=35) (N=8) (N=23)
Year of the Study
1960 5 1 5
1976 19 4 12
1980-88 11 3 6
Data Collection
Survey26 4 20
Experiment 6 2 1
Exprmnt. & Survey 3 2 2
Study Design
Before-After 30 5 20
After-Only 5 3 3
Exposure to Debate
Measured 25 8 14
Unmeasured 10 0 9
Including Nonviewers
Yes 10 4 5
No 25 4 18
Sample Characteristics
National Adults4 0 3
Local Adults 16 6 11
Selected Groups2 0 2
Students 13 2 7
Sample Size
Under 100 7 0 5
100-300 15 4 8
300+ 13 4 10
Campaign Context
Primary Season 4 1 1
General Season31 7 22
Debate Sequence
First Debate 28 7 20
Later Debate 7 1 3
Linking Debate Content
Yes 5 2 0
No 30 6 0
Effects of Debate
Substantial/Clear na 3 5
Moderate/Mixedna 4 10
None na 1 8
Note: Four studies which examined the impact of debate on affective
response or
vote decision are included in the overall 35 studies, but excluded from
the issue-
knowledge group and the image-formation group.
TABLE 2
CHARACTERISTICS OF THREE GROUPS BEFORE THE DEBATE
Exprmnt Non- Natural
Viewers Viewers Viewers
(N=53) (N=21) (N=65)
Female 54.7% 71.4% 50.8%
Employed 50.9 42.9 33.8
Class Standing
Freshman 39.6 42.9 32.3
Jr. & Sr. 41.5 28.6 30.8
Media Reliance
TV > Newspapers 43.4 28.6 35.4
TV < Newspapers 15.1 28.6 16.9
Election Particptn
Registered to Vote 73.6 81.0 80.0
Intend to Vote 81.1 81.0 86.2
Political Interest*
Very Much 30.2 23.8 40.6
Not At All 24.5 28.6 4.7
Party Affiliation
Democrat 26.4 19.0 23.4
Republican 17.0 23.8 29.7
Independenta 56.6 57.1 46.9
TV = Newspapers 39.6 33.3 44.6
Pre-test Scores on Bush
Debated Issues 22.10 21.76 25.27
Undebated Issues 25.10 21.61 30.41
Competence 4.88 4.96 5.37
Potency 5.11 4.95 5.38
Integrity 3.76 4.01 4.18
Charisma 4.53 4.90 4.94
Communication 4.12 4.03 4.35
Vote Choice .16 .14 .32
Pre-test Scores on Clinton
Debated Issues 39.38 41.07 42.69
Undebated Issues 29.08 26.98 31.92
Competence 5.22 4.96 5.14
Potency 5.23 5.31 5.27
Integrity 4.21 4.52 4.35
Charisma 5.66 5.92 5.93
Communication 5.03 5.06 4.95
Vote Choice .50 .61 .41
Pre-test Scores on Perot
Debated Issues 10.69 12.69 14.87
Undebated Issues 21.69 9.52 26.92
Competence 3.74 3.71 4.17
Potency 4.14 3.71 4.54
Integrity 3.31 2.84 3.66
Charisma 3.93 3.71 3.90
Communication* 3.15 2.98 3.81
Vote Choice .01 .00 .00
* Difference between the natural viewers and the experimental
viewers/nonviewers
is significant at .05.
a "Independent" includes A Connecticut Party and People for Perot.
TABLE 3
UNSTANDARDIZED OLS REGRESSION COEFFICIENTS
(Standard Errors in Parentheses)
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Debated Issues
Exposure 33.996*** 24.703*** 38.900*** 38.572*** 16.522** 38.894**
to Debate (4.697) (4.090) (6.288) (7.188) (5.947) (11.904)
Pretest .389** .562*** .248 .391*** .526*** .026
Score (.144) (.096) (.151) (.106) (.082) (.168)
Undebated issues
Exposure 1.937 2.672 8.677 -4.337 .558 16.147
to Debate (2.782) (2.722) (6.661) (5.812) (5.819) (12.019)
Pretest .604*** .520*** .352** .719*** .642*** .474***
Score (.087) (.109) (.113) (.098) (.103) (.130)
Competence
Exposure.169 -.045 .400 .399 .381 .301
to Debate (.229) (.180) (.332) (.392) (.473) (.537)
Pretest .796*** .544*** .679*** .493*** .584*** .466***
Score (.088) (.079) (.116) (.127) (.135) (.131)
Potency
Exposure.333 .076 .396 -.287 -.085 .328
to Debate (.265) (.230) (.333) (.371) (.382) (.598)
Pretest .623*** .532*** .373*** .524*** .559*** .441***
Score (.091) (.092) (.090) (.098) (.092) (.108)
Integrity
Exposure.328 -.072 1.175*** .233 .766 .329
to Debate (.254) (.289) (.337) (.385) (.462) (.459)
Pretest .828*** .772*** .542*** .681*** .573*** .768***
Score (.076) (.089) (.101) (.082) (.094) (.115)
Charisma
Exposure.218 .246 1.498*** .379 .240 .655
to Debate (.263) (.257) (.276) (.441) (.405) (.519)
Pretest .838*** .573*** .553*** .696*** .493*** .295*
Score (.092) (.110) (.082) (.096) (.112) (.123)
Communication
Exposure.561* .387 .843** -.092 .378 .430
to Debate (.265) (.263) (.304) (.439) (.461) (.584)
Pretest .406*** .356*** .427*** .452*** .582*** .490***
Score (.087) (.103) (.110) (.106) (.113) (.116)
Note: Exposure is a dummy variable for the experimental subjects with 1
being the
viewer group and 0 being control group; for the natural subjects, it is
an
interval scale varying from 0 to 1 with 1 being maximum watching and
attention and
0 being minimal watching and attention.
APPENDIX A
A LIST OF DEBATE EFFECTS STUDIES
Barnett, G. A. (1981). A multidimensional analysis of the 1976
presidential
campaign. Communication Quarterly, 29 (3), 156-165.
Becker, L. B., Sobowale, I. A., Cobbey, R. E., and Eyal, C. H. (1980).
Debates'
effects on voters' understanding of candidates and issues. In G.
F. Bishop,
R. G. Meadow, and M. Jackson-Beeck (Eds.), The Presidential
Debates: Media,
Electoral, and Policy Perspectives. New York, NY: Praeger
Publishers, pp.
126-139.
Becker, L. B., Weaver, D. H., Graber, D. A., and McCombs, M. E. (1979).
Influence
of public agendas. In S. Kraus (Ed.), The Great Debates: Carter
vs. Ford,
1976. Bloomington, IN: Indiana University Press, pp. 418-428.
Becker, S. L., Pepper, R., Wenner, L. A., and Kime, J. K. (1979).
Information
flow and the shaping of meanings. In S. Kraus (Ed.), The Great
Debates:
Carter vs. Ford, 1976. Bloomington, IN: Indiana University Press,
pp. 384-
397.
Ben-Zeev, S. and White, I. (1962). In S. Kraus (ED.), The Great
Debates:
Background, Perspective, Effects. Bloomington, IN: Indiana
University Press,
pp. 331-337.
Bishop, G.F., R. W. Oldendick, & A. J. Tuchfarber (1980). The
presidential
debates as a device for increasing the "rationality" of electoral
behavior.
In G. F. Bishop, R. G. Meadow, and M. Jackson-Beeck (Eds.), The
Presidential
Debates: Media, Electoral, and Policy Perspectives. New York, NY:
Praeger
Publishers, pp. 179-196.
Bothwell, R. K. and Brigham, J. C. (1983). Selective evaluation and
recall during
the 1980 Reagan-Carter debate. Journal of Applied Social
Psychology, 13 (5),
427-442.
Bowes, J. E. and Strentz, H. (1979). Candidate images: stereotyping and
the 1976
debates. In B. D. Ruben (Ed.), Communication Yearbook II, New
Brunswick, NJ:
Transaction Books, pp. 391-406.
Carter, R. F. (1962). Some effects of the debates. In S. Kraus (ED.),
The Great
Debates: Background, Perspective, Effects. Bloomington, IN:
Indiana
University Press, pp. 253-270.
Davis, D. K. (1979). Influence on vote decisions. In S. Kraus (Ed.),
The Great
Debates: Carter vs. Ford, 1976. Bloomington, IN: Indiana
University Press,
pp. 331-347.
Davis, M. (1982). Voting intentions and the 1980 Carter-Reagan debate.
Journal of
Applied Social Psychology, 12 (6), 481-492.
Dennis, J., Chaffee, S. H., and Choe, S. Y. (1979). Impact on partisan,
image,
and issue voting. In S. Kraus (Ed.), The Great Debates: Carter vs.
Ford,
1976. Bloomington, IN: Indiana University Press, pp. 314-330.
Drew, D., and Weaver, D.H. (1991). Voter learning in the 1988
presidential
election: Did the debates and the media matter? Journalism
Quarterly, 68,
27-37.
Hagner, P. R. and Rieselbach, L. N. (1980). The impact of the 1976
presidential
debates: conversion or reinforcement. In G. F. Bishop, R. G.
Meadow, and M.
Jackson-Beeck (Eds.), The Presidential Debates: Media, Electoral,
and Policy
Perspectives. New York, NY: Praeger Publishers, pp. 157-178.
Hawkins, R. P., Pingree, S., Smith, K. A., and Bechtolt, W. E., Jr.
(1979).
Adolescents' responses to issues and images. In S. Kraus (Ed.),
The Great
Debates: Carter vs. Ford, 1976. Bloomington, IN: Indiana
University Press,
pp. 368-383.
Jones, B. V. (1982). Attitude change toward the candidates and of the
1980
presidential election. The Journal of Social Psychology, 116,
297-298.
Kraus, S. and Smith, R. G. (1962). Issues and Images. In S. Kraus
(ED.), The
Great Debates: Background, Perspective, Effects. Bloomington, IN:
Indiana
University Press, pp. 289-312.
Lang, K. and Lang, G. E. (1962). Reactions of viewers. In S. Kraus
(ED.), The
Great Debates: Background, Perspective, Effects. Bloomington, IN:
Indiana
University Press, pp. 313-330.
Lang, G. E. and Lang, K. (1979). Immediate and mediated responses. In
S. Kraus
(Ed.), The Great Debates: Carter vs. Ford, 1976. Bloomington, IN:
Indiana
University Press, pp. 298-313.
Lang, G. E. and Lang, K. (1980). The formation of public opinion:
Direct and
mediated effects of the first debate. In G. F. Bishop, R. G.
Meadow, and M.
Jackson-Beeck (Eds.), The Presidential Debates: Media, Electoral,
and Policy
Perspectives. New York, NY: Praeger Publishers, pp. 61-80.
Lanoue, D. J. and Schrott, P. R. (1989a). Voters' reactions to
televised
presidential debates: measurement of the source and magnitude of
opinion
change. Political Psychology, 10 (2), 275-285.
Lanoue, D. J. and Schrott, P. R. (1989b). The effects of primary season
debates
on public opinion. Political Behavior, 11 (3), 289-306.
Lemert, J. B., Elliot, W. R., Nestvold, K. J., and Rarick, G. R. (1983).
Effects
of viewing a presidential primary debate: an experiment.
Communication
Research, 10 (2), 155-173.
McIntosh, E. G. (1989). Clarification of issues during the presidential
debates.
Psychological Reports, 64, 1124.
McLeod, J. M., Durall, J. A., Ziemke, D. A., and Bybee, C. R. (1979).
Reactions
of young and older voters: Expanding the context of effects. In S.
Kraus
(Ed.), The Great Debates: Carter vs. Ford, 1976. Bloomington, IN:
Indiana
University Press, pp. 348-367.
Miller, A. H. and MacKuen. (1979). Informing the electorate: A national
study.
In S. Kraus (Ed.), The Great Debates: Carter vs. Ford, 1976.
Bloomington,
IN: Indiana University Press, pp. 269-297.
Nimmo, D., Mansfield, M., and Curry, J. (1980). Persistence and change
in
candidate images. In G. F. Bishop, R. G. Meadow, and M.
Jackson-Beeck
(Eds.), The Presidential Debates: Media, Electoral, and Policy
Perspectives.
New York, NY: Praeger Publishers, pp. 140-156.
Oshagan, H. (1988). Looking at voting as a decisional processes: What
factors
determine initial preference? Paper presented at the Annual
Meeting of the
Association for Education in Journalism and Mass Communication,
71st,
Portland, OR, July 2-5.
Payne, J. G., Golden, J. L., Marlier, J., and Ratzan, S. C. (1989).
Perceptions
of the 1988 presidential and vice-presidential debates. American
Behavioral
Scientist, 32 (4), 425-435.
Pfau, M. (1987). The influence of intraparty political debates on
candidate
preference. Communication Research, 14 (6), 687-697.
Simons, H. W. and Leibowitz, K. (1979). Shifts in candidate images. In
S. Kraus
(Ed.), The Great Debates: Carter vs. Ford, 1976. Bloomington, IN:
Indiana
University Press, pp. 398-404.
Stamm, K. R. (1985). Effects of the Bush/Ferraro debate on candidate
characterization. Paper presented at the Annual Meeting of the
Association
for Education in Journalism and Mass Communication, 68th, Memphis,
TN, August
3-6.
Steeper, F.T. (1980). Public response to Gerald Ford's statements on
Eastern
Europe in the second debate. In G. F. Bishop, R. G. Meadow, and M.
Jackson-
Beeck (Eds.), The Presidential Debates: Media, Electoral, and
Policy
Perspectives. New York, NY: Praeger Publishers, pp. 81-104.
Tannenbaum, P. H., Greenberg, B. S., and Silverman, F. R. (1962).
Candidate
Images. In S. Kraus (ED.), The Great Debates: Background,
Perspective,
Effects. Bloomington, IN: Indiana University Press, pp. 271-288.
APPENDIX B
WORDING OF ISSUE KNOWLEDGE ITEMS
(Numbered as original, correct answer in parentheses)
#3. Wants to lower taxes on the profits obtained from investment in new
business
ventures only. (Clinton: Yes.)
#7. Would require employers of 50 or more people to provide employees
unpaid leave
from their jobs when there are family emergencies or to care for new
born
children. (Bush: No; Clinton: Yes.)
#9. Backs a major government program to hire people to work on building
and
rebuilding roads, bridges and railbeds. (Clinton: Yes; Perot: Yes.)
#10. Would avoid sending troops to stop the violence between Serbs,
Muslims and
Bosnians and wants the violence stopped through diplomatic means.
(Bush: Yes;
Clinton: Yes; Perot: Yes.)
#11. Wants legislation to reduce the number of civil (not criminal)
lawsuits.
(Bush: Yes).
#15. Wants to increase taxes on "wealthier" Americans. (Bush: No;
Clinton: Yes.)
#16. Supports full and active participation of Afro-Americans in
American social
and economic life. (Bush: Yes; Clinton: Yes; Perot: Yes.)
#18. Has presented a plan that would eliminate the national debt
completely in 5
years. (Clinton: No.)
#19. Advocates a government program to retrain workers who have become
unemployed.
(Bush: Yes; Clinton: Yes.)
#20. Would increase government spending to treat AIDS patients and to
find a cure
for AIDS. (Bush: No; Clinton: Yes.)
#24. Would increase the tax on gasoline 10 cents a gallon each year for
the next
five years. (Bush: No; Perot: Yes.)
APPENDIX C
WORDING OF IMAGE PERCEPTION ITEMS
(Numbered as original, Cronbach's Alpha in parentheses)
Competence:
#1. Intelligent ... Not intelligent
#2. Competent ... Incompetent
#4. He knows a lot ... He doesn't know much
(The Cronbach's Alpha of the three items for Bush is .76 at time 1 and
.83 at time
2; for Clinton is .75 at time 1 and .75 at time 2; for Perot is .76 at
time 1 and
.90 at time 2.)
Potency:
#21. Aggressive .. Passive
#22. Strong ... Weak
#24. Dominant ... Subservient
(The Cronbach's Alpha of the three items for Bush is .80 at time 1 and
.85 at time
2; for Clinton is .75 at time 1 and .77 at time 2; for Perot is .81 at
time 1 and
.85 at time 2.)
Integrity:
#14. Trustworthy ... Untrustworthy
#15. Sincere ... Insincere
#33. Honest ... Dishonest
(The Cronbach's Alpha of the three items for Bush is .87 at time 1 and
.87 at time
2; for Clinton is .79 at time 1 and .84 at time 2; for Perot is .81 at
time 1 and
.80 at time 2.)
Charisma
#16. Pleasant ... Unpleasant
#17. Warm personality ... Cold personality
#31. Friendly ... Unfriendly
(The Cronbach's Alpha of the three items for Bush is .82 at time 1 and
.89 at time
2; for Clinton is .80 at time 1 and .87 at time 2; for Perot is .80 at
time 1 and
.89 at time 2.)
Communication Skills
#9. Expressive ... Unexpressive
#10. Clear ... Vague
#12. Speaks in specifics ... Speaks in generalities
(The Cronbach's Alpha of the three items for Bush is .70 at time 1 and
.67 at time
2; for Clinton is .65 at time 1 and .72 at time 2; for Perot is .65 at
time 1 and
.71 at time 2.)
APPENDIX D
TABLE 1
DETERMINANTS OF KNOWLEDGE SCORE ON DEBATED ISSUES
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure 33.996*** 24.703*** 38.900*** 38.572*** 16.522** 38.894**
to Debate (4.697) (4.090) (6.288) (7.188) (5.947) (11.904)
Pretest .389** .562*** .248 .391*** .526*** .026
Score (.144) (.096) (.151) (.106) (.082) (.168)
Gender 1.073 10.831** -8.699 -10.991* -4.102 -6.201
(male=1) (4.513) (4.067) (6.084) (4.480) (3.746) (7.332)
Employed 5.903 8.502* -6.027 8.684 8.133* -9.114
(=1) (4.574) (3.889) (5.921) (4.798) (3.967) (7.878)
Political 3.900 1.608 6.067 11.405** 7.150* 6.064
Interest (3.062) (2.676) (3.936) (4.480) (3.441) (6.172)
Republican -3.435 -6.554 -7.801 -12.531* 2.633 12.986
(=1) (6.130) (5.237) (7.951) (5.408) (4.480) (8.818)
Democrat -1.045 -.015 10.321 -8.245 .951 2.493
(=1) (5.175) (4.461) (6.815) (5.308) (4.437) (8.725)
Constant 1.349 -.334 15.079 -7.558 3.450 6.357
(8.646) (7.966) ( 11.171) (9.516) (7.855) (15.436)
Adj. R2 .484 .544 .413 .594 .607 .243
Note: Exposure is a dummy variable for the experimental subjects with 1
being the
viewer group and 0 being control group; for the natural subjects, it is
an
interval scale varying from 0 to 1 with 1 being maximum watching and
attention and
0 being minimal watching and attention.
TABLE 2
DETERMINANTS OF KNOWLEDGE SCORE ON UNDEBATED ISSUES
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure 1.937 2.672 8.677 -4.337 .558 16.147
to Debate (2.782) (2.722) (6.661) (5.812) (5.819) (12.019)
Pretest .604*** .520*** .352** .719*** .642*** .474***
Score (.087) (.109) (.113) (.098) (.103) (.130)
Gender-2.220 1.729 -10.411 -6.637 -.685 7.088
(male=1) (2.737) (2.582) (6.359) (3.651) (3.644) (7.776)
Employed 1.650 -1.212 11.956 -.170 -3.208 -4.678
(=1) (2.628) (2.574) (6.325) (3.863) (3.859) (8.004)
Political 5.138** 5.114** 5.171 6.014 7.121* 5.154
Interest (1.845) (1.777) (4.080) (3.445) (3.241) (6.907)
Republican 1.980 -2.078 -1.531 -1.246 -2.277 1.660
(=1) (3.494) (3.441) (8.425) (4.469) (4.360) (9.092)
Democrat -.957 -1.579 .529 -5.126 .331 1.407
(=1) (3.036) (3.011) (7.412) (4.319) (4.292) (9.119)
Constant -1.997 .263 5.794 .817 -4.251 8.393
(4.979) (5.255) (11.792) (7.829) (7.614) (15.909)
Adj. R2 .564 .378 .247 .612 .508 .212
For an explanation of Exposure, see note a in Table 1.
TABLE 3
DETERMINANTS OF IMAGE SCORE ON COMPETENCE
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure.169 -.045 .400 .399 .381 .301
to Debate (.229) (.180) (.332) (.392) (.473) (.537)
Pretest .796*** .544*** .679*** .493*** .584*** .466***
Score (.088) (.079) (.116) (.127) (.135) (.131)
Gender .447* .313 -.093 .165 .056 -.018
(male=1) (.217) (.169) (.316) (.251) (.289) (.347)
Employed.193 -.059 -.179 -.249 -.057 -.124
(=1) (.217) (.170) (.316) (.271) (.316) (.359)
Political -.000 -.022 .555** -.061 -.154 -.134
Interest (.142) (.115) (.206) (.206) (.243) (.287)
Republican .163 -.094 -.596 .768* -.184 .178
(=1) (.296) (.230) (.417) (.295) (.362) (.403)
Democrat -.352 .312 -.497 -.354 .178 -.029
(=1) (.250) (.198) (.373) (.299) (.343) (.406)
Constant 1.082 2.815 .853 2.633 2.612 3.108
(.632) (.479) (.724) (.805) (.823) (.852)
Adj. R2 .564 .444 .416 .361 .256 .139
For an explanation of Exposure, see note a in Table 1.
TABLE 4
DETERMINANTS OF IMAGE SCORE ON POTENCY
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure.333 .076 .396 -.287 -.085 .328
to Debate (.265) (.230) (.333) (.371) (.382) (.598)
Pretest .623*** .532*** .373*** .524*** .559*** .441***
Score (.091) (.092) (.090) (.098) (.092) (.108)
Gender .480 .073 -.117 -.082 -.086 -.400
(male=1) (.249) (.214) (.310) (.235) (.235) (.376)
Employed -.144 -.034 .447 .079 .654* -.195
(=1) (.249) (.216) (.313) (.244) (.249) (.394)
Political -.034 -.081 .597** -.279 -.263 .099
Interest (.163) (.142) (.206) (.186) (.193) (.315)
Republican .166 .101 .054 .746** -.165 .032
(=1) (.337) (.295) (.416) (.269) (.290) (.446)
Democrat -.539 .406 -.002 -.096 .006 .215
(=1) (.288) (.253) (.361) (.271) (.276) (.444)
Constant 1.711 2.583 1.482 3.098 2.839 2.817
(.666) (.616) (.651) (.681) (.681) (.852)
Adj. R2 .426 .333 .291 .471 .443 .236
For an explanation of Exposure, see note a in Table 1.
TABLE 5
DETERMINANTS OF IMAGE SCORE ON INTEGRITY
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure.328 -.072 1.175*** .233 .766 .329
to Debate (.254) (.289) (.337) (.385) (.462) (.459)
Pretest .828*** .772*** .542*** .681*** .573*** .768***
Score (.076) (.089) (.101) (.082) (.094) (.115)
Gender -.252 .002 .138 -.329 .035 .208
(male=1) (.243) (.274) (.312) (.241) (.289) (.299)
Employed -.211 .261 .108 .088 .245 -.248
(=1) (.241) (.272) (.313) (.259) (.310) (.316)
Political .072 -.100 .322 -.678** -.444 .174
Interest (.157) (.183) (.205) (.205) (.241) (.239)
Republican .341 .075 -.108 .274 -.190 -.360
(=1) (.334) (.361) (.417) (.297) (.363) (.345)
Democrat -.163 .344 .033 -.450 .322 .522
(=1) (.279) (.323) (.370) (.300) (.351) (.364)
Constant 1.228 1.376 1.115 3.237 2.589 1.023
(.503) (.562) (.631) (.557) (.689) (.799)
Adj. R2 .668 .560 .391 .648 .480 .388
For an explanation of Exposure, see note a in Table 1.
TABLE 6
DETERMINANTS OF IMAGE SCORE ON CHARISMA
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure.218 .246 1.498*** .379 .240 .655
to Debate (.263) (.257) (.276) (.441) (.405) (.519)
Pretest .838*** .573*** .553*** .696*** .493*** .295*
Score (.092) (.110) (.082) (.096) (.112) (.123)
Gender -.409 .070 .232 -.113 .048 .259
(male=1) (.245) (.238) (.257) (.282) (.254) (.326)
Employed -.206 -.244 .016 -.048 .336 .053
(=1) (.249) (.242) (.261) (.290) (.273) (.350)
Political -.098 -.051 .344 -.482* -.139 -.183
Interest (.162) (.161) (.173) (.226) (.219) (.274)
Republican -.187 .125 .082 .461 -.701* .533
(=1) (.348) (.325) (.348) (.337) (.311) (.389)
Democrat -.522 .360 .232 -.443 .214 .664
(=1) (.286) (.296) (.301) (.336) (.302) (.410)
Constant 1.304 2.409 1.096 2.404 3.038 3.696
(.626) (.698) (.595) (.794) (.729) (.953)
Adj. R2 .570 .313 .514 .622 .374 .053
For an explanation of Exposure, see note a in Table 1.
TABLE 7
DETERMINANTS OF IMAGE SCORE ON COMMUNICATION SKILLS
Experimental Sample (N=74) Natural Viewer Sample (N=65)
Bush Clinton Perot Bush Clinton Perot
Exposure.561* .387 .843** -.092 .378 .430
to Debate (.265) (.263) (.304) (.439) (.461) (.584)
Pretest .406*** .356*** .427*** .452*** .582*** .490***
Score (.087) (.103) (.110) (.106) (.113) (.116)
Gender .316 .441 -.659* -.004 .149 .234
(male=1) (.246) (.245) (.287) (.278) (.277) (.356)
Employed.185 .026 .285 -.072 .410 .082
(=1) (.258) (.248) (.293) (.304) (.303) (.381)
Political -.218 -.069 .245 -.029 .137 .211
Interest (.163) (.166) (.188) (.228) (.228) (.295)
Republican .445 .423 -.003 .689* -.693* .006
(=1) (.332) (.330) (.382) (.336) (.333) (.427)
Democrat.130 .869** -.416 -.660* -.674* -.281
(=1) (.292) (.289) (.331) (.329) (.336) (.428)
Constant 2.730 2.880 2.118 2.779 1.935 2.039
(.616) (.640) (.617) (.770) (.769) ( (.826)
Adj. R2 .281 .251 .342 .411 .362 .217
For an explanation of Exposure, see note a in Table 1.
more candidates), and
======================================================================= 58
Return-Path: <@[log in to unmask]:[log in to unmask]>
Received: from CMUVM (NJE origin SMTP@CMUVM) by CMUVM.CSV.CMICH.EDU (LMail
V1.1d/1.7f) with BSMTP id 5336; Tue, 3 Aug 1993 22:35:45 -0400
Received: from lynx.dac.neu.edu by CMUVM.CSV.CMICH.EDU (IBM VM SMTP V2R1)
with TCP; Tue, 03 Aug 93 22:35:43 EDT
Received: by lynx.dac.neu.edu (5.65/DEC-Ultrix/4.3)
id AA00462; Tue, 3 Aug 1993 22:34:36 -0400
From: [log in to unmask] (William Smith)
Message-Id: <[log in to unmask]>
Subject: ZHUJ DEBATES 93 RTVJ Abstract
To: [log in to unmask] (Elliott Parker)
Date: Tue, 3 Aug 1993 22:34:34 -0400 (EDT)
X-Mailer: ELM [version 2.4 PL21]
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Content-Length: 1263
HUJ DEBATES 93 RTVJ Images vs. Issues in Presidential Debates
Does the Audience Learn More about Images
than Issues from Televised Debates?
Effects of the First 1992 Presidential Debate
Jian-Hua Zhu, J. Ronald Milavsky and Rahul Biswas
Dept. of Communication Sciences
University of Connecticut
Storrs, CT 06226
Paper Submitted to the Radio and Television Division of
the Association for Education in Journalism and Mass
Communication,
April 1, 1993
ABSTRACT
This paper reviews all previous studies of debate effects on
issue learning and image formation, and reports the results of a
new study of the effects of the first presidential debate in the
1992 election. The study used a between-subjects design with
repeated measures. The study was replicated under both
experimental and natural viewing conditions. Results show that
the viewers learned a great deal about candidates' issue
positions that were discussed in the debate, but no learning
took place of issues that were not debated. The debate did not
affect the viewers' perception of the two well-known candidates'
personalities, but improved the perception of the less well-
known candidate, Ross Perot, on several debate-related
personality traits.
|
