The Media and Voter Turnout
An Investigation of People's Willingness to Vote
in the 1992 Presidential Election
Nojin Kwak
5169 Vilas Communication Hall
School of Journalism and Mass Communication
University of Wisconsin-Madison
Madison, Wisconsin 53706
(608) 263-7852 / 262-3690
e-mail: [log in to unmask]
The author is a doctoral student in Mass Communication at the
University of Wisconsin-Madison, with primary interests in the role of the media
in electoral process
Paper submitted to the Communication and Theory Division,
Association for Education in Journalism and Mass Communication, 79th Convention,
Anaheim, CA, August, 1996.
Abstract
This study investigated how media-related variables were
associated with people's willingness to vote, synthetically considering
variables investigated in voter turnout literature based on major electoral
theories: sociological, social-psychological, and rational choice. In the
cross-sectional analysis, indirect effects of the media and negative influence
of talk shows were observed; in the panel, the media's direct impact was
identified for some respondents. The importance of the process-oriented approach
on electoral behaviors and campaign media use was corroborated.
The Media and Voter Turnout: An Investigation of People's
Willingness to Vote in the 1992 Presidential Election
The Media and Voter Turnout :
An Investigation of People's Willingness to Vote in the 1992
Presidential Election
Introduction
Various changes in the political environment have made
people's political behaviors less predictable, compelling researchers to
re-conceptualize the role of the media in the political process (McLeod,
Kosicki, & McLeod, 1994; O'Keefe & Atwood, 1981). Voter turnout, which has
continuously declined since 1960, and was barely over 50% in recent presidential
elections, is even regarded as a "puzzle" to researchers (Brody, 1978).
Consequently, voter turnout, as an important phenomenon of political volatility,
becomes an interesting area of research, particularly in terms of the role of
the media (McLeod et al., 1994)
In the political science literature, however, the media's
impact on voter turnout has rarely been addressed (e.g., see Wolfinger and
Rosenstone, 1980). Three major paradigms of voting behavior, sociological
(Lazarsfeld, Berelson, and Gaudet, 1964), social-psychological (Campbell,
Converse, Miller and Stokes, 1960), and rational choice (Downs, 1957)
perspectives, or studies based on these paradigms have paid little attention to
communication variables in their theoretical and empirical analyses (for a
summary and comparison of these three paradigms, see Dennis, 1991). Some
researchers have shown an interest in media factors, but their attempts have
often been limited to including a single indicator such as newspaper reading as
an element comprising people's campaign interest or involvement, rather than
treating it as a distinct theoretical concept (Nichols and Beck, 1995; Teixeira,
1987).
The neglect of media factors in voter turnout models is
ironic, given that we can find a lot of attention to campaign media in the
related literature, from demonstrating the active role of the media in electoral
process (Ranney, 1983) to criticizing the media as trivializing/eroding the
electoral process by focusing on non-substantive issues (Patterson, 1993).
Although these studies do not specifically address the media effects on voter
turnout, one of the basic themes underlying these studies is that the media do
matter in the election process.
In mass communication research, voter turnout has been
relatively neglected, too; studies have been sporadic or are outdated (Bybee,
McLeod, Luetscher, & Garramone, 1981; Glaser, 1965; Simon & Stern, 1955).
Recently, analyzing the 1992 presidential election, Weaver and Drew (1995) found
the effect of people's media use on their willingness to vote was indirect:
people's media use boosted their campaign interest which, in their model, most
significantly affected respondents' stated likelihood of voting. However, their
study is inconclusive, because their model of voter turnout and their choice of
variables were not guided by theoretical concerns developed in voter turnout
literature. Thus, their findings concerning people's intention to vote and media
use need to be re-tested in a theoretical model composed of various
turnout-related, both long- and short-term, variables.
Ansolabehere and Iyengar (1994) also attempted to assess the
influence of media on people's turnout behavior in an experimental setting.
Based on the rational choice model, they tested whether the closeness of the
election, reflected in poll-related news reports, affected respondents'
likelihood to vote. They failed to find any significant effects on individuals'
inclination to vote.
What is interesting in their study, though, is that they
treated people's perception of opinion trends or climate of opinion (which was
reflected in the media polls) as a main mechanism helping them to calculate the
extent of election closeness. On what basis people's calculation of election
closeness is processed has not been carefully conceptualized in traditional
rational choice research (e.g., Ricker and Ordeshook, 1968). Mostly these
studies simply ask respondents to estimate whether a given election will be
close or not. However, Ansolabehere et al. (1994) designed an experiment in such
a way that people's perception of the closeness of a race resulted from
assessing the climate of opinion as to whether the campaign was/would be
competitive or not.
To some extent, voter turnout can be understood as a product
of the campaign process. Some individuals' likelihood of voting may be
determined by such long-term political orientations as strength of partisanship
and sense of civic duty to vote, well before a campaign begins. However, for
others, various factors related to campaign may increase or decrease their
inclination to vote. A similar view was proposed by Blumler and McLeod (1974),
who emphasized that voter turnout research should be designed to reflect the
dynamics of the phenomenon. In a British election study, they differentiated
respondents based on the (in)consistency between their expressed pre-campaign
intention to vote and actual turnout. However, they failed to address the
dynamic aspects of various short-term forces such as people's campaign interest
and their media use. People's media use pattern, for example, may not be
consistent but vary during the course of a campaign; and such variation may be
accompanied by increase or decrease in people's willingness to turn out at the
polls. This approach may enable researchers to ascertain the effects of
short-term forces more appropriately.
To summarize, this study will investigate how media-related
variables are associated with people's stated willingness to vote, synthetically
considering the effects of various other variables which have been investigated
in voter turnout literature based on three major electoral theories:
sociological, social-psychological, and rational choice. It will present two
separate analyses using cross-sectional and panel data. The former focuses on a
general understanding of how people's intention to vote is predicted by various
variables; and the latter investigates how people's willingness to vote is
influenced by variations in short-term forces during the campaign.
Variables
Five categories of variables will be investigated in the
study: demographics (age, sex, education and income), social-psychological,
campaign involvement, rational choice, and those relevant to the domain of the
media. In the panel data analysis, all of these categories except for the
rational choice variables will be studied. The demographic category partly
represents the sociological tradition in voting literature (Lazarsfeld et al.,
1954; Leighley & Nagler, 1992), which emphasizes the role of social locations
people occupy.
The social-psychological approach of the Michigan School
(Campbell et al., 1960) focuses upon people's long-term political orientations.
In addition to such usual measures as political efficacy and the intensity of
partisanship, people's general political interest is included in this category,
representing people's long-term psychological involvement in politics.
Campaign involvement is a category specifically related to a
particular campaign; which also evolved from the Michigan School. It includes
people's campaign interest and candidate affect. People's campaign interest has
frequently been analyzed in voter turnout research; but people's affective
feeling toward candidates has rarely been addressed in this context, despite its
assumed importance. According to a social-cognitive model (Rahn, Aldrich,
Borgida and Sullivan, 1990), people's feeling about candidates, which are based
on their assessment of candidates' leadership ability and personal qualities,
forms a basis for voters' decisions. Their assertions on the importance of what
they called "an affective summary of candidates" also correspond to the idea of
candidate-centered politics in the context of electoral dealignment (Wattenberg,
1990). If people's summary feelings toward candidates are one of important
factors influencing their voting decisions, it should be more than plausible to
expect that people's affective evaluation about candidates, or candidate affect,
will exert an influence on voter turnout.
The rational choice model assumes that people's individual
economic calculations in an election (costs versus benefits from voting) help
them decide whether to vote or not. Two aspects of the rational choice model are
considered: high cost and the perceived closeness of election (for a thorough
and critical discussion of empirical applications of this model, see Dennis,
1991). High cost as an obstacle to turnout at the polls is represented by the
difficulty expressed in voter decision-making. This refers to the extent to
which people experience difficulty in deciding on their favorite candidate. The
closeness of the election is conceptualized as one's vote value, which is
defined as a perceived importance of one's vote in influencing the election
outcome. Following Ansolabehere et al.(1994), this concept is operationalized as
reflecting one's perception of the opinion climate, in terms of the distribution
of people's support for candidates and the trend of support, as will be
discussed later.
Finally, people's media use will be considered in a broad
voter turnout model discussed above. In addition to people's use of
political/campaign stories in the two major media (TV and newspapers), their use
of important campaign events such as party conventions and candidate debates as
presented in the media, and people's use of TV talk shows with candidate
appearances will be separately analyzed.
Method
Samples
Two kinds of samples, cross-sectional and panel, were
analyzed in this study. Both of these samples were gathered from telephone
interviews conducted by XXXXXXXXXX at the XXXXXXXXX. The interviews for the
cross-sectional data were conducted in late-October, 1992; and the sample is
composed of a random sample of 421 adult residents in XXXXX, XXXXXX. The panel
data were gathered in August and October in 1992 from the same population; and
the size of the sample is 131.
Operationalization
In the cross-sectional analysis, the dependent variable,
people's willingness to vote, was measured by asking respondents their
likelihood of voting on election day using a 100 percentile scale. Including
demographic variables (age, sex, education and income), five categories of
independent variables were considered: social-psychological, campaign
involvement, rational choice, general campaign media use, and major campaign
event media use.
The social-psychological block includes such variables as
partisan support (whether a respondent is a party-supporter or not), political
efficacy, and general political interest. Political efficacy is composed of two
factors: personal efficacy and general efficacy. Five variables were
factor-analyzed using principle component extraction method and oblique
rotation. The first factor, personal efficacy, refers to one's perception about
her/his own political effectiveness, based on people's responses to the
following questions: "Public officials don't care about what people like me
think," and, "Sometimes politics and government are so complicated that people
like me can't really understand what's going on." The second factor, general
efficacy, concerns the extent to which a respondent evaluated the political
importance of general public including her/himself. People's responses to three
questions indexed the factor: : "Every vote counts in an election including
yours and mine," "Most of the time we can trust our government to do what's
right," and "In America, everyone who wants to has a voice what the government
does." Respondents were asked about their degree of agreement on a ten point
scale with the statements. People's general political interest was measured by
asking how much respondents were interested in politics in general, apart from
the campaign, on a 10-point scale.
Campaign involvement includes two variables: candidate affect
and campaign interest. Candidate affect refers to how favorably one feels toward
his/her supported candidate on a 100 degree thermometer measure. Campaign
interest indicates to what extent respondents are interested in following the
current presidential campaign; it was measured on a 10-point scale.
Two variables are considered which represent the rational
choice model: difficulty in decision-making as an indicator of cost aspect and
one's vote value in terms of affecting the election outcome. Respondents whose
stated that their likelihood of voting was greater than 0% were asked about
their degree of difficulty in deciding for whom to vote for president. The index
for vote value was constructed by combining two questions: 1) asking the
estimated current distribution of votes for candidates and 2) asking about the
trend of support, i.e., whether each candidate was gaining or losing public
support, or support was staying the same. One's vote value was coded 1, where
one's vote was least likely to affect the outcome of the election, if one
considered his/her supporting candidate as the front-runner and as gaining
support, or if one believed his/her candidate was not the front-runner and was
losing support. One's vote value was coded 2, if one's candidate was perceived
as being behind another candidate and as receiving invariable public support. If
one/s supporting candidate was considered as the front-runner but as maintaining
the same level of public support, his/her vote value for the campaign was coded
3. Finally, one's vote value was coded 4, where one's vote value was most likely
to affect the election outcome, if one's candidate was perceived as leading the
race but as losing public support, or one's candidate was thought of as a
underdog but as gaining public support.
People's general campaign media use consisted of 5 variables:
newspaper political news exposure, newspaper political news attention, TV
political news exposure, TV political news attention and exposure to TV talk
shows with candidate appearances. Newspaper political news exposure was
constructed by combining respondents' exposure to international, national and
local news (alpha=.82), and newspaper political news attention is a composite
index of respondents' attention to these three types of news contents
(alpha=.82). People's exposure to morning, local and network news constituted a
composite index of TV political news exposure (alpha=.63), and TV political news
attention was composed of respondents' attention to stories about international
affairs, national government and local government (alpha=.76). All of original
indicators were measured on a 10-point scale from "rarely read (watch)" to
"read (watch) all the time" for exposure questions, and from "little attention"
to "very close attention" for attention questions. If respondents answered they
did not read or watch, their responses were coded as zero for each variable.
Finally, major campaign event media use measured people's
media use of televised candidate debates. Respondents' exposure to these
debates was indexed by constructing a composite measure of their exposure to
three presidential and one vice presidential candidate debates, which were asked
separately (alpha=.76). Respondents' attention to televised debates was measured
by a question asking their level of attention to televised debates on a 10-point
scale.
The use of variables in the panel data analysis was rather
limited. Because panel analysis is interested in how changes in one's likelihood
of voting are influenced by dynamic factors over the course of the campaign,
variables in campaign involvement and media categories were constructed by using
the data of both waves 1 and 2. Stable variables such as demographic and
social-psychological variables were analyzed using either wave 1 or 2 data,
depending on when the relevant questions were asked. Political efficacy and the
rational choice model were not investigated due to unavailability of relevant
data.
The dependent variable, changes in one's likelihood of
voting, was constructed by subtracting one's vote intention at wave 1 from that
at wave 2. Campaign involvement includes two variables: change in one's campaign
interest and change in one's candidate affect. Both variables represent the
difference between people's responses at waves 1 and 2, with wave 1 information
used as a base line. Four variables belong to general campaign media use. These
are the degree to which one was exposed to presidential campaign stories, and
the extent to which one paid attention to campaign stories, all of which were
asked separately for newspapers and TV. These variables were computed according
to how different people's attention or exposure at wave 1 is from those at wave
2. Finally, people's media use of major campaign events reflects changes in
people's media use of major campaign events. Two major events were selected
which occurred at the early and late stages of the general campaign period:
party conventions and candidate debates. Thus, people's changed use of those
events was constructed by subtracting their attention to televised party
conventions from that to debates.
Results
The order of entering each category in regression analyses
was decided by the relative distance from the point of decision-making as to
whether to turn out at the polls. Thus, demographic variables were entered
first, followed by social-psychological block, which represents long-term
political orientations. Campaign-related variables (i.e., campaign involvement)
were entered as the third block and the rational choice category was entered
next. Media variables were entered last, with general campaign media use first
and major campaign event media use next.
All the regression coefficients reported were standardized so
that each variable's relative contribution can be compared with each other.
Simple bivariate correlations and regression analyses of each category after
controlling for demographic variables are first, followed by regression
analyses with all the blocks of variables.
Cross-sectional data analysis
All the indicators of the social-psychological approach and
campaign involvement had significant associations with the people's intention to
vote (Table 1). People's interest in the campaign (r=.33) showed the strongest
relationship, followed by respondents' general interest in politics (r=.22).
Partisan support, political efficacy and candidate affect also demonstrated
significant correlations. The incremental R2 produced by the
social-psychological block controlling for the demographic block was .054, with
partisan support and political interest having significant betas. Campaign
involvement explained 9.3% of total variance in people's willingness to vote,
which was the greatest magnitude, with only campaign interest being
independently significant.
Among rational choice indicators, difficulty in
decision-making showed a significant initial correlation (.16) and beta (.18)
coefficients. The incremental R2 of the model was .035, which was primarily due
to the contribution of the difficulty indicator, given the insignificant
association of one's vote value with the dependent variable. However, the
direction of the difficulty indicator's coefficient was the opposite to what we
had expected. The greater one's difficulty in deciding who to vote for, the more
likely one was to state s/he would vote. This result directed the investigator
to re-evaluate the conceptual meaning of the indicator; and it was concluded
that what the indicator had measured was not the intended high cost aspect , but
the degree of investment one perceived s/he had made in the decision-making
process during the campaign. That is, because the question was asked of
respondents having more than a zero percent intention to vote, people expressing
difficulty might be those who already had paid a high price, not those who
calculated the expected cost for their future behavior. Because of such a
conceptual difference of the current difficulty measure from the indicators of
the original rational choice model, each indicator (i.e., vote value and
difficulty variables) was entered separately in subsequent regression analyses.
Some indicators in both media use categories showed
significant associations at the zero-order level; but none of betas and
incremental R2 turned out to be significant. Among demographic variables,
education (r=.18) remained significant when considered along with other
demographic variables (beta=.16). The demographic category's R2 (.046) is
significant.
Table 2 shows a hierarchical multiple regression result with
7 blocks entered. Overall, the whole model explained 22.0% of the total variance
in the people's intention to vote. The incremental R2 of social-psychological
approach (5.4%), campaign involvement (4.9%), and the difficulty block (4.0%)
were significant even after the preceding blocks' contribution was considered.
However, the R2 changes produced by the other blocks were not significant.
Looking at the final betas, we can find people's campaign
interest had the strongest association with their willingness to vote (.30).
One's perceived difficulty in decision-making, the campaign investment one made
in deciding who was his/her favorite candidate, appeared to be the next best
indicator(.23). Education (.13) and one's identification with a party (.11)
showed positive, significant relationships with the dependent variable.
People's talk show exposure, whose initial coefficient (-.01)
was not significant, appeared as a significant indicator after all the variables
in the model were entered (final beta=-.11). Being a rather weak indicator
(t=-2.150), it suggested that exposure to candidate appearances on TV talk shows
had a negative influence on people's willingness to vote.
Overall, those who had high interest in following the
campaign process and who had experienced difficulty in choosing their favorite
candidate tended to have a greater intention to vote. Also, those with a higher
education level, stronger identification with either party and less exposure to
campaign-related TV talk shows were more likely to say they would turn out at
the polls.
In Tables 3 and 4, I extended the investigation to see how
people's campaign interest, the strongest predictor of people's intention to
vote, was associated with other variables. In Table 3, people's media use of a
major campaign event, exposure and attention to televised presidential
candidates' debates (r=.45 and r=.61, respectively), showed the strongest
associations with campaign interest. This category alone, after controlling for
the demographic variables, explained 35.5% of the dependent variable's variance,
with debate attention having a particularly strong beta (.57). Variables
included under the social-psychological approach also showed a strong
relationship. All the variables, partisan support (r=.21, beta=.15), general
political interest (r=.51, beta=.50), personal efficacy (r=.18, beta=.11), and
general efficacy (r=.17, beta=.09), were significantly associated with people's
willingness to vote; and the incremental R2 of this category was .306. People's
ordinal campaign media use explained 18.1% of the total variance of the
dependent variable. All five variables had significant relationships (r ranges
from .13 to .35), with newspaper exposure (beta=.27), TV attention (beta=.26),
and talk show exposure (beta=.18) remaining significant after demographic
controls. Candidate affect, which was introduced to represent the phenomenon of
candidate-centered election, was also significantly related to people's campaign
interest (r=.32, beta=.28, the incremental R2= 7.8%). However, the difficulty of
decision making, the vote value and the demographic variables' relationships
with people's campaign interest were not significant.
Thus, we see that 52.2% of the total variance in campaign
interest was explained by the 7 blocks (Table 4). The incremental R2 of
social-psychological dimension (.306), candidate affect (.023), general campaign
media use (.045), and debate media use (.130) were significant. Although the
social-psychological block's incremental R2 looks impressive, what is more
significant is the debate media use block's independent contribution obtained
after entering all the other 6 blocks. When the social-psychological block was
entered as the last block, the incremental R2 produced by the block was only
.076 (this finding is not shown in tables).The two media blocks together
independently explained almost 18% of the total variance in people's campaign
interest, after all the other blocks' contributions were considered.
The final betas indicate similar patterns of those discussed
above. It was debate attention (.43), which was most strongly related to
people's campaign interest. People's political interest (.29), both of political
efficacy (.10 and .07), and candidate affect (.14) also showed significant
coefficients. People's newspaper exposure's final beta (.15) was significant as
well.
Overall, in terms of people's long-term political
orientation, those who had high political efficacy and political interest tended
to have a keen interest in the election campaign. In the campaign-related
domain, people's favorable feeling toward their favorite candidates tended to
boost their campaign interest. In terms of media use, those who read newspapers
more often and who paid closer attention to major campaign events such as
candidates' debates on TV were more likely to be interested in the campaign
process.
Panel data analysis
The bivariate correlations between the independent variables
and people's voting intention change during the general campaign were
significant for education, political interest, and campaign interest change
(Table 5). The campaign interest change had the strongest relationship with the
dependent variable (r=.33); and the relationship remained significant after
demographic control (beta=.30). The incremental R2 of campaign involvement was
.086.
Interpretation of the negative association of education
(r=-.18, beta=-.17) and political interest (r=-.25, beta=-.24) with change in
one's willingness to vote should be approached with caution. Since only 4.7% of
the whole respondents' voting intention decreased during the campaign, these
associations should be interpreted as indicating that people with higher
education and greater political interest were more likely to have a consistent
degree of voting intention throughout the campaign. In other words, those with
lower education and less interest in politics tended to develop their
willingness to vote during the campaign. In terms of incremental R2, the
social-psychological category explained 5.3% of the dependent variable's
variance. However, the demographic block's R2 change was not significant.
Changed patterns in people's media use of both general and
major campaign events did not significantly affect the change in their stated
likelihood of voting. Nor did the two categories' incremental R2s appear to be
significant.
The portion of the total variance in people's voting
intention change explained by the five blocks analyzed was 20.4% (Table 6). The
social-psychological (.047) and campaign involvement (.080) blocks' R2 changes
were significant. The other blocks did not show any independently significant
contribution after relevant controls. Only two variables' final betas were
significant, political interest (-.24) and campaign interest change (.27), of
which people's campaign interest change was a stronger indicator.
Thus, those who came to be more interested in the campaign
became more willing to participate in voting during the campaign period; and
people whose political interest was low were more likely to experience an
increased inclination to vote during the campaign.
The relationship between people's political interest and
voting intention change presented a simple, but important question: why was it
that people with low political interest were increasingly willing to vote as the
campaign unfolded? The frequency distribution of vote intention change for the
high political interest group (i.e., respondents located at the upper half of
the scale) revealed that almost 86% of the sample had demonstrated no change in
their stated intention to vote, while 42% of the low political interest group
(i.e., respondents of the lower half of the scale) indicated that their
intention to vote had changed. These frequency differences again mandated
further inquiry. Thus, a separate regression analysis was employed to see which
factors affected those with low political interest in terms of their voting
intention change (Table 7).
Among the low political interest group, as expected, people's
campaign interest change showed a strong positive association with the dependent
variable in terms of both r (.47) and final beta (.39). What is more interesting
is, however, the significant correlation between people's TV use pattern and
change in their willingness to vote (r=.40 and final beta=.52). These findings
demonstrate that, among people with a low level of political interest, those
whose television viewing related to the campaign increased tended to be more
inclined to vote as the campaign proceeded.
As an individual indicator, this TV use variable was a
stronger predictor than the campaign interest variable, based on the comparison
between final betas. As a block, the media block's incremental R2 (.149) was
marginally significant (sig. of F change=.065). However, it needs to be
emphasized that the incremental R2 of the media block was produced after
controlling for 8 variables in the three preceding blocks including campaign
involvement. Also, in a separate analysis, the incremental R2 of the media
block, after controlling for demographic and social-psychological blocks, was as
large as .262 (sig. F change=.012, this finding is not shown in tables). Thus,
it is fair to conclude the media block was as significant as the campaign
involvement in explaining the variance in the dependent variable.
As Table 8 shows, the media variables' strong relationship
with people's campaign interest change was apparent. People's general campaign
media use category explained 19.4% of the total variance of the dependent
variable after demographic controls. People's increasing use of campaign stories
in the media in such areas as newspaper attention (.32), TV exposure (.35) and
TV attention (.38) led to people's increasing campaign interest. Newspaper
attention (beta=.25) and TV exposure (beta=.19) retained significant
associations even after the demographic variables were controlled for. People's
increasing attention to major campaign events on TV also showed a significant
relationship (r=.33, beta=.34). Its incremental R2 (.111) is significant as
well. However, except for age (r=-.18, beta=-.15), all the other variables were
not significantly related to people's campaign interest change; the incremental
R2 was not significant, either.
The hierarchical regression in Table 9 reassures us that
there is a strong relationship between the media variables and campaign interest
change. The two media blocks, if combined, independently explained more than 23%
of the total variance in the dependent variable. The total R2 of the model was
.292.
The final betas of TV exposure (.23), newspaper attention (.21)
and major campaign event media attention (.22) changes showed significant
associations, thereby indicating that, as the campaign proceeded, those who came
to watch more campaign stories on TV, read more closely campaign-related
articles in newspapers, and paid more attention to major campaign events in the
media became more interested in the campaign.
Conclusion and Discussion
In the cross-sectional data analysis, findings supported the
results arrived at by Weaver et al. (1995). Campaign interest was the most
significant variable predicting people's stated willingness to participate in
voting, and the media variables were strong indicators explaining people's
campaign interest. Such media impact remained significant even though various
long-term and campaign-related political variables were simultaneously
considered in a regression equation.
The significant influences of education and the strength of
partisanship were replicated, as found in voter turnout literature (Aldrich &
Simon, 1986). Other long-term political orientations such as political efficacy
and general political interest, and people's affirmative feeling toward their
favorite candidates were found to bolster their interest in the campaign, but
not their willingness to vote. Among all the variables significantly related to
respondents' campaign interest, in particular, people's viewing of major
election events via the media (i.e., televised candidate debates) was most
impressive in terms of predicting how strongly people were interested in the
ongoing campaign process.
In addition, people's tendency not to "waste" what they have
invested during the campaign (i.e., resources so invested as to perceive some
difficulty in deciding on their favorite candidate) was observed, whereby people
who perceived more difficulty tended to be more willing to vote. Difficulty
indicator's final beta was the second most significant coefficient; and it
explained 4.0% of the variance of the dependent variable, even after controlling
for demographic and social-psychological variables. In future research, such a
campaign investment dimension and its relationship with people's media use in
explaining their electoral behaviors may need more attention.
One interesting finding is the negative relationship between
people's exposure to TV talk shows and their stated willingness to vote. This
relationship remained significant even after 18 variables in 7 blocks were
considered. One simple interpretation may be that the more strongly one intended
to vote, the less likely s/he watched TV talks shows, because one might have
thought that these shows dealt with less substantive political issues. However,
a recent investigation on this relatively new component of political
communication evinces a different explanation.
Hollander (1995) found that for those with low education
level the more they watched talk shows, the more likely they perceived
themselves as being knowledgeable about election issues, without actually
gaining knowledge. That is, watching talk shows might provide people with a
sense of being informed, thereby making people less likely to seek other
informational sources. In the same line of reasoning, the interactive nature of
talk shows might give heavy viewers a feeling of electoral participation, which
might replace their actual participation in voting.
Some findings in this study suggest the impact of talk shows
may be rather complex. Before people's major event media use was considered,
people's TV talk show exposure appeared significantly related to their campaign
interest (regression equation 6 in Table 4), which predicted greater
participation at the polls. Based on these findings, we can generalize that
among talk show viewers, if their watching contributed to their campaign
interest, their voter turnout was likely to be boosted by such viewing; but if
they could not make the connection between talk shows and an on-going campaign,
their increased viewing further isolated them from politics in the real world.
Whether the connection was realized or not should be dependent to a great degree
upon how talk show contents were composed.
In the panel, it was again found that those who became more
interested in the campaign tended to be increasingly willing to vote as the
campaign proceeded. Also, as people became more frequent and attentive users of
the media, their interest in the campaign interest grew during the campaign
period.
In sum, the media variables' relationship with people's
stated willingness to vote was found to be indirect, rather than direct. Media
factors appeared to influence the pre-condition (i.e., campaign interest) which
was strongly associated with people's stated likelihood to vote.
In the panel analysis, the concept of voter turnout, or
willingness to vote, was alternatively understood as a dynamic construct. Rather
than simply understanding voter turnout or intention to vote as being realized
at a single point in a campaign or as being immutable as a consequence of
people's long-term political predispositions, this study proposed that the
decision for whether or not to vote might in part mature during the campaign,
that is, would be a product of the campaign process. The panel data revealed
that, while 76% of the sample showed a consistent level of voting intention, the
voting intention of 24% of the respondents changed during the campaign, either
increasing or decreasing. Consequently, questions were raised as to why some
people's willingness to vote changed, while others did not, and why some
people's intention to vote changed to a greater degree. These findings indicated
that the conception of the intention to vote as a dynamic concept was plausible
empirically as well as theoretically. Some variables investigated, including
media variables, were found to be meaningfully associated with the dependent
variable, as discussed above.
One part of the findings which is suggestive for future
research in this process-oriented approach is the media variables' role in
increasing the low political interest group's willingness to vote during the
campaign. Particularly, their increasing TV watching helped to boost their
inclination to vote over the course of campaign, even more strongly than did
their campaign interest.
The process-oriented approach may enable researchers to look
beyond short-term political variables, such as campaign interest, in terms of
media effects on political variables in the electoral process. At least one
study reported findings which suggest possible variation in long-term political
orientation during the campaign period (Brody, 1977, who was interested in the
changes in the intensity of partisan identification). If such a phenomenon,
i.e., the volatility of long-term political variables which are considered
rather stable, can be generalized to other variables, the media's role in voter
turnout process may prove to be even more significant than past studies have
indicated. People's sense of civic duty to vote, which is one of most important
predictors of people's voter turnout (Nichols et al., 1995), may be relevant in
this context. Given many media organizations' explicit efforts to emphasize
civic duties and to increase voter turnout (e.g., MTV's "Rock the Vote"
campaign), the media's influence on people's perceived civic duty during a
campaign may be expected.
Thus, the influence of the media on people's political
behaviors during a campaign period cannot be satisfactorily ascertained without
investigating the dynamic nature of related variables: people's political
orientation, political behaviors, and media use. These processes cannot be
captured by a cross-sectional survey conducted in late October before the
election, when, for example, people's sense of duty should be fully operating.
Longitudinal research such as a panel design is necessary so that we can trace
the changes in people's sense of civic duty over a longer period of time, the
effect of such changes on people's inclination to vote, and the media's role in
these processes.
Reference
Aldrich, J. H. & Simon, D. M. (1986). Turnout in American national
elections. In S. Long (ed.).
Research in Micropolitics. Volume 1. Greenwich: JAI Press Inc.
Ansolabehere, S. & Iyengar, S. (1994). Of horseshoes and horse
race: experimental studies of the
impact of poll results on electoral behavior. Political
Communication 11: 413-430.
Blumler, J. G. & McLeod, J. M. (1974). Communication and voter
turnout in Britain. In T. Leggatt (ed.) Sociological Theory and Survey
Research: Institutional Change and Social
Policy in Great Britain. Beverly Hills: SAGE.
Brody, R. A. (1978). The puzzle of political participation in
America. In A. King (ed.) The New
Political System. Washington, D. C.: American Enterprise
Institute.
Brody, R. A. (1977). Stability and change in party identification:
presidential to off-years.
Paper presented at the annual meeting of the American Political
Science Association.
Washington, D.C.
Bybee, C. R., McLeod, J. M., Luetscher, W. D. & Garramone, G.
(1981). Mass communication
and voter volatility. Public Opinion Quarterly. 45: 69-90.
Campbell, A., Converse, P. E., Miller, W. E., & Stokes, D. E.
(1960). The American Voter.
New York: Wiley.
Dennis, J. (1991). Theories of turnout: an empirical comparison of
alienationist and rationalist
perspectives. In W. Crotty (ed.). Political Participation and
American Democracy.
New York: Greenwood Press.
Dennis, J. (1991). The study of electoral behavior. In W. Crotty
(ed.). Political Science: Looking
to the Future. Vol. 3. Political Behavior. Evanston: Northwestern
Univ. Press
Downs, A. (1957). An Economic Theory of Democracy. New York:
Harper and Row
Glaser, W. A. (1965). Television and voting turnout. Public
Opinion Quarterly. 29.: 71-86.
Hollander, B. A. (1995). The new news and the 1992 presidential
campaign: perceived vs. actual
political knowledge. Journalism and Mass Communication Quarterly.
72: 786-798.
Lazarsfeld, P. F., Berelson, B., & Gaudet, H. (1944). The People's
Choice. New York: Duell,
Sloan & Pearce.
Leighley, J. E. & Nagler, J. (1992). Individual and systematic
influences on turnout: who votes?
1984. The Journal of Politics. 54: 718-739.
McLeod, J. M., Kosicki, G., & McLeod, D. (1994). The expanding
boundaries of political
communication effects. In J. Bryant & D. Zillmann (eds.).
Perspectives on Media
Effects. Hillsdale: L. Erlbaum Associates.
Nichols, S. M. & Beck, P. A. (1995). Reversing the decline: voter
turnout in the 1992 election. In
H. F. Weisberg (ed.). Democracy's Feast: Elections in America.
Chatham: Chatham
House Publishers, Inc.
O'Keefe, G. & Atwood, E.L. (1981). Communication and election
campaigns. In D. D. Nimmo
& K. R. Sanders (eds.). Handbook of Political Communication.
Beverly Hills:
SAGE.
Patterson, T. E. (1994). Out of Order. New York: Vintage Books.
Rahn, W. M., Aldrich, J. H., Borgida, E., & Sullivan, J. L.
(1990). A social-cognitive model of
candidate appraisal. In J. Ferejohn & J. Kuklinski (eds.).
Information and Democratic
Processes. Urbana: Univ. of Illinois Press.
Ranney, A. (1983). Channels of Power: the Impact of Television on
American Politics. New
York: Basic Books.
Ricker, W. H. & Ordeshook, P. C. (1968). A theory of the calculus
of voting. American Political
Science Review. 62: 25-42.
Simon, H. A. & Stern, F. (1955). The effect of television upon
voting behavior in the 1952
Presidential election. American Political Science Review. 49 :
470-477.
Teixeira, R. A. (1987). Why Americans Don't Vote: Turnout Decline
in the United States
1960-1984. New York: Greenwood Press.
Wattenberg, M. P. (1990). From a partisan to a candidate-centered
electorate. In A. King (ed.).
The New American Political System. Washington, D.C.: The AEI
Press.
Wattenberg, M. P. (1990). The Decline of American Political
Parties, 1952-1988. Cambridge:
Harvard Univ. Press.
Weaver, D. & Drew, D. (1995). Voter learning in the 1992
presidential election: did the "nontraditional" media and debates matter?
Journalism and Mass Communication
Quarterly. 72(1): 7-17.
Wolfinger, R. E. & Rosenstone, S. J. (1980). Who Votes? New
Heaven: Yale Univ. Press
Table 1
Preliminary Analysis of People's Willingness to Vote
Bivariate Correlations and Regression Analysis Controlling for
Demographic Variables
r beta Incremental
R2(%)
Demographic
age .06 .06
sex (female) .01 .02
education .18** .16**
income .15** .10 4.6%***
Social-Psychological
partisan support (supporter) .16** .14**
general political interest .22*** .18***
personal efficacy .10* .03
general efficacy .05 .01 5.4%***
Campaign Involvement
candidate affect .17** .02
campaign interest .33*** .30*** 9.3%***
Rational Choice Model
difficulty in decision .16** .18***
value of vote .05 - .07 3.5%***
General Media Use
NP political news exposure .18*** .14
NP political news attention .16** .02
TV political news exposure .07 - .03
TV political news attention - .02 .01
TV talk show exposure - .01 - .01 2.1%
Major Event Media Use
debate exposure .16** .09
debate attention .17** .08 2.4%
n=421
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
the demographic variables were analyzed without
controlling for other blocks of variables.
Table 2
Hierarchical Regression Analysis of People's Willingness to Vote
R1 R2 R3 R4 R5 R6 R7
Demographic
age .06 .05 .05 .07 .06 .06 .05
sex(female) .02 .04 .02 .03 .02 .03 .03
education .16** .11* .13** .15** .15** .13** .13**
income .10 .09 .08 .08 .07 .06 .07
Social-Psychological
partisan support(supporter) .14** .10* .10* .10* .11* .11*
general political interest .18*** .04 .06 .05 .05 .06
personal efficacy .03 .00 .02 .03 .02 .01
general efficacy .01 - .01 - .02 - .02
- .02 - .02
Campaign Involvement
candidate affect .00 .06 .06 .06 .05
campaign interest .27*** .25*** .25*** .28*** .30***
Campaign Investment
difficulty in decision-making .21*** .22*** .23*** .23***
Rational Choice Model
value of vote - .07 - .08 - .08
General Campaign Media Use
NP political news exposure .07 .06
NP political news attention .05 .05
TV political news exposure .01 .00
TV political news attention - .10 - .09
TV talk show exposure - .11* - .11*
Major Event Media Use
debate exposure .09
debate attention - .11
R-square (%) 4.6 10.0 14.8 18.9 19.3 21.4 22.0
Incremental R-square (%) 4.6 5.4 4.9 4.0 0.4 2.2 0.6
Significance of F Change .001 .000 .000 .000 157 .050 .241
n=421
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
Table 3
Preliminary Analysis of People's Interest in Campaign
Bivariate Correlations and Regression Analysis Controlling for
Demographic variables
r beta Incremental
R2
Demographic
age .07 .07
sex (female) .03 .02
education .08 .07
income .08 .06 1.4%
Social-Psychological
partisan support (supporter) .21*** .16***
general political interest .51*** .50***
personal efficacy .18*** .11*
general efficacy .17** .09* 30.6%***
Campaign Involvement
candidate affect .32*** .28*** 7.8%***
Campaign Investment
difficulty in decision -.05 -.03 .10%
Rational Choice Model
value of vote -.02 - .01 .0%
General Media Use
NP political news exposure .33*** .27***
NP political news attention .28*** - .10
TV political news exposure .13** - .11
TV political news attention .35*** .26***
TV talk show exposure .26***
.18*** 18.1%***
Major Event Media Use
debate exposure .45*** .04
debate attention .61*** .57*** 35.5%***
n=421
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
the demographic variables were analyzed without
controlling for other blocks of variables.
Table 4
Hierarchical Regression Analysis of People's Campaign Interest
R1 R2 R3 R4 R5 R6 R7
Demographic
age .07 .01 .01 .02 .02 .00 - .02
sex(female) .02 .07 .05 .05 .06 .04 .04
education .07 - .07 - .05 - .05 - .05
- .04 - .05
income .06 .03 .03 .03 .03 .02 - .01
Social-Psychological
partisan support(supporter) .15*** .12** .12** .12** .10* .06
general political
interest .50*** .48** .48** .48** .37*** .29***
personal efficacy .11* .09* .10* .10* .11* .10*
general efficacy .09* .08 .08 .08* .05 .07*
Campaign Involvement
candidate affect
.16*** .18*** .18*** .20*** .14***
Campaign Investment
difficulty in decision-making .06 .06 .05 .02
Rational Choice Model
value of vote .00 .01 -.01
General Campaign Media Use
NP political news exposure .18* .15*
NP political news attention - .14 - .12
TV political news exposure - .06 - .06
TV political news attention .18** .08
TV talk show exposure .10* .03
Major Event Media Use
debate exposure - .01
debate attention .43***
R-square (%) 1.4 32.0 34.3 34.7 34.7 39.2 52.2
Incremental R-square (%) 1.4 30.6 2.3 0.4 0.0 4.5 13.0
Significance of F Change .204 .000 .000 .130 .907 .000 .000
n=421
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
Table 5
Preliminary Analysis of People's Voting Willingness Change
Bivariate Correlations and Regression Analysis Controlling for
Demographic Variables
r beta Incremental
R2
Demographic
age - .13 - .15
sex (female) .01 - .06
education - .18** - .22*
income - .03 .02 5.8%
Social-Psychological
partisan support (supporter) - .02 .00
political interest - .25** - .23** 4.7%*
Campaign Involvement
candidate affect change .08 .00
campaign interest change .33*** .30*** 8.6%**
General Media Use
NP campaign exposure change .13 .08
NP campaign attention change .12 .04
TV campaign exposure change .16 .17
TV campaign attention change .10
- .05 3.4%
Major Event Media Use
major event attention change .11 .13 1.5%
n=131
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
the demographic variables were analyzed without
controlling for other blocks of variables.
Table 6
Hierarchical Regression Analysis of People's Voting Willingness
Change
R1 R2 R3 R4 R5
Demographic
age -.15 -.15 -.10 -.09 -.09
sex(female) -.06 -.10 -.10 -.09 -.09
education -.22* -.18 -.18 -.18 -.17
income .02 .01 .02 -.02 -.02
Social-Psychological
partisan support(supporter) .00 .03 .03 .03
political interest -.23* -.21* -.23** -.23**
Campaign Involvement
candidate affect .01 .02 .02
campaign interest .29*** .27** .27**
General Campaign Media Use
NP campaign exposure change .08 .08
NP campaign attention change -.02 -.02
TV campaign exposure change .16 .16
TV campaign attention change -.12 -.12
Major Event Media Use
major event attention change -.01
R-square (%) 5.8 10.5 18.5 20.4 20.4
Incremental R-square (%) 5.8 4.7 8.0 1.9 0.0
Significance of F Change .108 .041 .003 .587 .982
n=131
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients.
Table 7
Hierarchical Regression of Low Political Interest Group's Voting
Willingness Change
r R1 R2 R3 R4 R5
Demographic
age -.25 -.22 -.25 -.03 -.04 -.04
sex(female) .01 -.11 -.13 -.12 -.13 -.15
education -.25 -.27 -.24 -.42* -.38* -.38*
income .00 .05 .03 .07 .00 .00
Social-Psychological
partisan support(supporter) .13 .08 .14 .15 .14
political interest -.02 -.09 -.02 -.20 -.21
Campaign Involvement
candidate affect .11 .10 .12 .12
campaign interest .47** .51** .37* .39*
General Campaign Media Use
NP campaign exposure change .24 .11 .13
NP campaign attention change .18 .14 .14
TV campaign exposure change .40** .52* .52*
TV campaign attention change .18 -.31 -.31
Major Event Media Use
major event attention change .16 -.05
R-square (%) 12.0 13.3 35.1 50.0 50.1
Incremental R-square (%) 12.0 1.2 21.9 14.9 0.1
Significance of F Change .250 .760 .005 .065 .784
n=46
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
Table 8
Preliminary Analysis of People's Campaign Interest Change
Bivariate Correlations and Regression Analysis Controlling for
Demographic Variables
r beta Incremental
R2
Demographic
age - .18* - .18*
sex (female) .00 .00
education .02 .00
income - .05 - .05 3.7%
Social-Psychological
partisan support (supporter) - .09 - .09
political interest - .05 - .05 1.1%
Campaign Involvement
candidate affect change .13 .10 1.0%
General Media Use
NP campaign exposure change .13 - .05
NP campaign attention change .32*** .26*
TV campaign exposure change .35*** .20
TV campaign attention change .38***
.15 19.4%***
Major Event Media Use
major event attention change .33*** .34*** 11.1%***
n=131
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
the demographic variables were analyzed without
controlling for other blocks of variables.
Table 9
Hierarchical Regression Analysis of People's Campaign Interest
Change
R1 R2 R3 R4 R5
Demographic
age -.18* -.18* -.17 -.10 -.11
sex(female) .00 -.01 .00 .03 .05
education .01 .02 .04 .06 .03
income -.05 -.06 -.06 -.07 -.06
Social-Psychological
partisan support(supporter) -.09 -.08 -.05 .00
political interest -.05 -.06 -.09 -.07
Campaign Involvement
candidate affect .10 .12 .10
General Campaign Media Use
NP campaign exposure change -.05 -.10
NP campaign attention change .24* .21*
TV campaign exposure change .23* .22*
TV campaign attention change .15 .13
Major Event Media Use
major event attention change .22*
R-square (%) 3.7 4.8 5.8 25.5 29.2
Incremental R-square (%) 3.7 1.1 1.1 19.7 3.7
Significance of F Change .315 .492 .251 .000 .015
n=131
* p<.05 **p<.01 ***p<.001
Note: betas refer to standardized regression coefficients
|
