Polls and Elections The Conditional Effects of Competing Messages during Presidential Nominating Conventions

Page created by Marion Thornton
 
CONTINUE READING
Polls and Elections The Conditional Effects of Competing Messages during Presidential Nominating Conventions
Polls and Elections

 The Conditional Effects of Competing Messages
  during Presidential Nominating Conventions

                                                    JOSEPH CERA
                                            AARON C. WEINSCHENK
                                           University of Wisconsin—Milwaukee

               Past research demonstrates that presidential nominating conventions can exercise multi-
        ple effects on individual-level opinion; consumption of convention speeches yields opinion more
        favorable to the convening candidate, while exposure to partisan messaging in the surrounding
        information environment can trigger a general partisan bias. In this article, we demonstrate
        that the persuasive power of speeches made by candidates during the second convention in a given
        election cycle can be attenuated by exposure to information from the initial convention. Such
        conditional effects persist even when individual partisan affiliation and preconvention opinion
        are controlled. Notably, positive impressions made by candidates appear to be more impactful
        than negative messaging aimed at candidates by their opposition.

                                                   Introduction

      To what degree can positive impressions made by presidential candidates be suc-
cessfully counteracted by subsequent negative messaging? Can the impact of negative
messaging be nullified by subsequent positive impressions? These are questions that
matter to candidates and parties during the presidential nominating conventions, and
they are questions that scholars have yet to examine fully. Conventions are unique events
where the convening party has a few days of breathing room to campaign with the full
attention of the electorate while the opposition takes a traditional step back. In effect, the
parties take turns ceding a massive information advantage to one another. As a result,
there is a strong incentive for the party that convenes first to anticipate and preempt the

          Joseph Cera is a PhD candidate at the University of Wisconsin—Milwaukee. His research focuses on campaign
effects, political behavior, and survey methodology, and has appeared in American Politics Research.
          Aaron C. Weinschenk is a PhD candidate at the University of Wisconsin—Milwaukee. His research focuses on
political behavior and has appeared in Political Behavior, Political Research Quarterly, and American Politics
Research.

Presidential Studies Quarterly 42, no. 1 (March)         161
© 2012 Center for the Study of the Presidency
Polls and Elections The Conditional Effects of Competing Messages during Presidential Nominating Conventions
162   |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

opposing party’s messaging as much as possible, and it is in the best interests of the party
that convenes second to try and counteract any messaging presented by the first party.
While scholars have devoted attention to modeling the impact of single conventions, how
the conventions jointly impact opinion at the individual level has not been studied. We
think that considered together, the conventions within a given election cycle present
an ideal opportunity to study how individuals deal with sequential competing messages
and a compelling way to examine the extent to which information screening and biased
processing are involved in the formation of individual opinion about candidates for the
American presidency.
      With a few exceptions, most empirical investigation of campaign effects during
conventions has been concerned with aggregate-level outcomes. Unfortunately, conclu-
sions about the different ways that individuals are impacted by conventions cannot
reliably be drawn from aggregate-level data. Recently, researchers have recognized that
measuring variation in both individual characteristics and exposure to information during
conventions is necessary before a complete picture of how conventions shape public
opinion can be gained (Cera and Weinschenk 2012; Hillygus and Jackman 2003). Panel
data collected around single conventions has made it possible to isolate a persuasive effect
triggered by convention speech consumption from a general partisan bias-activating effect
associated with exposure to the rest of the intensely political convention information
atmosphere. In this article, we attempt to further advance individual-level investigation
of convention effects by isolating and considering individual exposure to information from
both conventions within a given election cycle.

                   The Informational Role of Conventions

      The study of campaign effects came to occupy an important place in the political
behavior literature as scholars began to focus on their informational role. However,
because of their complexity and length, a comprehensive model of campaign effects has
eluded us. Focusing on the specific events and activities that occur during election cycles,
such as debates and conventions, has made the study of campaigns more tractable
(Hillygus and Jackman 2003; Holbrook 1996; Shaw 1999). Shafer succinctly described
the usefulness of conventions to the study of campaign effects: “The convention is unique
in the degree to which it is spatially and temporally bounded, and hence intellectually
manageable. The context for its actions, then, can be examined in an unusually compre-
hensive fashion” (1988, 1). During a convention, the convening party captures nearly all
conventional media coverage and monopolizes it for several days. The result is a one-sided
onslaught of campaign information; “the only time during the campaign when parties
are able to exercise clear control over the flow of information” (Holbrook 1996, 80). This
aspect of conventions is particularly important because the audience during conventions
is generally vast and diverse, consisting of partisans of both stripes and independents
alike. In this way, conventions have the potential to both energize partisan supporters
and influence independents’ and opposition partisans’ perceptions of the candidates.
Trent and Friedenberg elaborate:
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                           |   163

     Since 1952 the presence of television has restructured convention programming so that the
     party’s ‘important’ events occur during ‘prime time.’ To make certain that this happens,
     the convention chair often ignores the activities of the delegates on the convention floor
     and rushes through any official party business to make certain that those events planned to
     give the party the most favorable image (for example, ecumenical prayers, civic greetings,
     performances by show business personalities, keynote and acceptance speeches, and con-
     trolled and planned ‘spontaneous’ demonstrations for candidates) will be seen during the
     hours in which most people watch television (2000, 45 emphasis added).

Panagopoulos (2007b) frames this information bonanza as beneficial to the electorate,
noting that “conventions continue to play an important role in the campaign process
partly because they arm citizens with more political information that can be used to
evaluate candidates and make wise vote choices” (92). The timing of conventions (early
in the general election cycle, when information about candidates is relatively scarce)
allows this flow of political information to exercise the largest impact (Holbrook 1996).

              How Conventions Impact Individual Opinion

      To date, most studies aimed at assessing the effects of presidential conventions on
opinion have focused on aggregate shifts in candidate support (commonly referred to
as convention “bumps” or “bounces”) (Campbell, Cherry, and Wink 1992; Gelman and
King 1993; Holbrook 1996; Panagopoulos 2007a; Shaw 1999; Stimson 2004). Much of
this work is outcome-oriented, focused on establishing that convention effects matter to
election outcomes. However, aggregate outcomes are always driven by individual-level
change, and there has been a dearth of research aimed at understanding how convention
information impacts individual members of the electorate. A notable exception is
the work of Hillygus and Jackman (2003), whose key innovation was the use of panel
data collected around the 2000 conventions. They focused on individual characteristics;
on who was likely to change after the conventions. However, they did not address the
fact that information exposure can vary from person to person. Building on their work,
Cera and Weinschenk (2012) used panel data from the 2004 election to study how
individuals interact with the information environment during conventions. They inte-
grated a measure of self-exposure to the major convention speeches into models of
postconvention candidate evaluations, which allowed them to separate out the effects of
information delivered directly by candidates to individual members of the electorate from
the effects of involuntary exposure to the general information environment surrounding
conventions. Cera and Weinschenk found that convention speeches exercise a persuasive
effect on consumers that benefits the convening candidate, even when individual parti-
sanship and preconvention opinion are controlled. Meanwhile, conventions exercise an
atmospheric effect that activates a general partisan bias, making partisanship more
relevant to political evaluations. Thus, postconvention opinion is driven by these separate
but simultaneous effects, which can reinforce or counteract each other, depending on
whether individuals share a party label with the convening candidate. By measuring both
individual partisanship and information consumption, they were also able to examine the
164   |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

role of bias in dealing with convention information. Cera and Weinschenk concluded that
bias in information search was more likely to influence postconvention opinion than
biased processing of information.

      The Next Step: Consideration of Information Asymmetries
      While progress has been made on isolating the effects single conventions have
on individual opinion, we have yet to consider variation in patterns of exposure to both
conventions or the impact of that variation. Previous research reveals significant dif-
ferences in postconvention opinion between individuals falling into different categories
of information consumption at a given convention. However, with respect to the
second convention in a given election cycle, among individuals within each category of
information consumption, there is a wide amount of variation in the amount of infor-
mation consumed during the prior convention. The potential for exposure to informa-
tion from the first convention to influence opinion after the second cannot be ignored.
Contemporary models of political learning and the literature on bias in dealing with
information suggest that additional insight about how individual opinion is shaped
by the conventions within a given election can be gained by considering their joint
impact.
      The “on-line” model of information processing introduced by Lodge, McGraw, and
Stroh (1989) suggests that individuals constantly integrate new information into their
candidate evaluations as it is encountered. The addition of each subsequent piece of
information has a lower relative impact as the amount of information being averaged into
the evaluation grows. The implication is that information encountered during the second
convention in a given election cycle will not make as much of an impact on individuals
who have already taken in relatively larger amounts of information from the initial
convention. Modeling a direct effect of consumption of information from the first
convention on candidate evaluations after the second convention, then, is likely to yield
additional explanatory power.
      The possible conditional effect of information from the first convention on the
relationship between information encountered during the second convention and
postconvention evaluations is another important consideration. People do not deal with
incoming information in an unbiased manner. Individuals make a cognitive investment
in information that culminates in a particular point of view. Subsequent conflicting
information is not always equitably incorporated; it can be discounted or ignored
(Ditto and Lopez 1992; Edwards and Smith 1996; Taber and Lodge 2006). Nyhan and
Reifler (2010) describe the mechanism underlying this behavior: “we expect that citi-
zens are more likely to generate counter-arguments against new information that con-
tradicts their beliefs than information that is consistent with their preexisting views.
As such, they are less likely to accept contradictory information than information that
reinforces their existing beliefs” (307). Researchers have identified some instances
where exposure to messaging that contradicts previously held views regarding political
candidates or policies can trigger an intensified attachment to those previously held
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                       |   165

views—a “backfire effect” (Nyhan and Reifler 2010; Redlawsk 2002). While this phe-
nomenon is not yet fully understood, there is a growing consensus that the process
of constructing counterarguments to newly encountered incongruent information leads
individuals to develop a greater degree of mental organization and structure in support
of their original beliefs than they normally would if left unchallenged (Lodge and
Taber 2000; Nyhan and Reifler 2010). Given the institutionalized sequential delivery
of competing messages that goes on during the presidential conventions, it would be
difficult to design a better petri dish for the culture (and study) of this type of political
behavior.

                                      Hypotheses

       Following the lead of recent research on the individual-level impact of conventions
on opinion, our first expectation is that respondent partisanship will be a significant
predictor of postconvention candidate evaluations, even when preconvention evaluations
are controlled, and after exposure to convention speeches are held constant. As the
convention atmosphere activates partisanship as a more salient decision basis, and as the
convening party’s dominance in the convention atmosphere sways political independents,
we expect the opinions of opposition partisans to significantly separate from the opinion
of independents and convening partisans.
       Our second expectation is that voluntary consumption of convention speeches
will exercise a persuasive pull on opinion in favor of the convening candidate, even
when individual partisanship and preconvention opinion are controlled. Speeches by the
candidates provide easily accessible avenues for members of the electorate to engage in
political learning through direct observation and unmatched opportunities for candidates
to make positive impressions on viewers.
       Our third expectation is that consumption of information from the first convention
(convention A) will exercise a direct effect on opinion after the second convention
(convention B), even after holding other factors constant. As information accumulates,
the impact of each new piece of information exercises less relative impact as it is averaged
in. Therefore, even after holding opinion prior to convention B constant, we expect those
who took in more information during convention A to hold opinions after convention B
less favorable to convening candidate B.
       Our fourth expectation is that the persistent influence of information from conven-
tion A will also be visible as a conditional effect on the impact of consumption of
candidate B’s convention speech on postconvention B opinion. Candidate B’s speeches
will have the largest persuasive impact on those who took in the least information during
convention A, but as investment in the impressions and messages from convention A
increases, the persuasive impact of exposure to candidate B’s speech will decrease.
Increased consumption of candidate B’s speech among individuals with the highest levels
of investment in information from convention A may even yield opinion less favorable
to candidate B. We expect this conditional effect to be present even after respondent
partisanship is controlled.
166    |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

                                        Data and Method

       With paired data points collected just before and immediately after each conven-
tion, and a variety of questions regarding individual characteristics and information
exposure, the 2000 National Annenberg Election Study (NAES) conventions panels
allowed us to investigate how both conventions impact individuals. What sets the 2000
NAES convention datasets apart from other years is that respondents for the second
(Democratic) convention panel were asked questions about information consumption
during the first (Republican) convention. This crucial difference allowed us to examine
both the direct effects of exposure to information from the Republican convention on
post-Democratic convention opinion, as well as the conditional effect of GOP1 conven-
tion information consumption on the relationship between Democratic information
consumption and post-Democratic convention opinion regarding the candidates.
       In 2000, a national cross-section of 1,197 respondents were interviewed immediately
before the Republican presidential nominating convention ( July 21 through July 30),
and were reinterviewed immediately afterward (August 4 through August 13). For the
Democratic convention panel, 1,230 different respondents were interviewed for the national
cross-section study immediately before the Democratic presidential nominating convention
(August 4 through August 13), and were reinterviewed afterward (August 18 through
August 27). The number of observations is lower in the models reported below simply
because the NAES did not always deliver every question to the full sample. Respon-
dents from the Democratic panel were asked about their experiences during the earlier
Republican convention as well as their experiences during the Democratic convention.
       We use ordinary least squares regression (OLS) to examine how the conventions
influenced individual evaluations of Al Gore (the Democratic nominee) and George W.
Bush (The Republican nominee).2 While the candidate evaluations we use to construct our
dependent variables were measured on ordinal scales, we combined them into indexes that
are best treated as continuous; therefore OLS is the appropriate method. We employed two
series of related regression models to test our hypotheses. The first series of models focuses
on evaluations of Gore and Bush after the 2000 Republican convention. The second series
focuses on evaluations of Gore and Bush after the 2000 Democratic convention.

                               Variables and Measurement

      Within each convention panel, once before the convention and once afterward,
respondents were presented with a series of traits and asked how well each one applied to
the respective candidates on a four-point scale ranging from “not well” to “extremely well.”
In order to strike a balance between thorough investigation of opinion regarding candidates

      1. GOP = Grand Old Party (Republican Party)
      2. Our selection of candidate evaluations as a measure of opinion regarding the candidates over
commonly used alternatives such as candidate support is based primarily on increased sensitivity. For a more
complete discussion of the benefits of using candidate evaluations to measure campaign effects, see Cera and
Weinschenk (2012).
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                              |   167

and concise reporting of results, we selected the four available positive traits (“really cares
about people like me,” “honest,” “inspiring,” and “knowledgeable”) and collapsed them into
candidate evaluation indexes.3 For each convention panel, we created two paired sets of pre
and postconvention evaluation indexes—one set for Gore and one for Bush. Higher index
values represent more favorable evaluations. Each index represents the average score across
all answered trait questions for a particular respondent. Scores for a given trait were only
included in a respondent’s index if that respondent provided a data point in both waves of
her or his panel. Cronbach’s alpha for each index never dropped below .8, indicating
intermeasure reliability. We use the postconvention indexes as dependent variables in our
regressions, while the preconvention indexes are used as control variables.
      We used two variables to capture the effects of voluntary exposure to convention
speeches; a measure of consumption of Bush’s convention speech, captured on a four-
point ordinal response scale ranging from “did not watch” to “all,” and an identically
scaled measure of consumption of Gore’s convention speech. These measures were trans-
lated directly from responses to survey questions about how much of each speech, if any,
respondents watched. For our regressions involving the Democratic convention, we
included a measure of exposure to the broadcasted parts of the Republican convention,
captured on a five-point ordinal response scale ranging from “did not watch” to “two
hours or more,” again based directly on responses to a corresponding survey question.
      A measure of party identification was also included. We combined the responses
from two related preconvention survey questions and recoded them into three mutually
exclusive dichotomous variables: Democrat (including Democrats and Democratic-
leaning independents); Republican (including Republicans and Republican-leaning
independents); and Independent (encompassing pure, nonleaning independents only).

                                Results and Discussion

       A brief examination of the aggregate-level effects of the 2000 conventions on
aggregate opinion gives context to our individual-level analysis. In order to examine the
effects of the conventions on candidate evaluations at the aggregate level, we compared
the pre- and postconvention indexes of candidate evaluations and applied significance
tests. Table 1 summarizes the effects of 2000 Republican convention on aggregate-level
candidate evaluations. Asterisks denote significant shifts.
       Prior to the Republican convention, Bush enjoyed slightly more favorable evalua-
tions than Gore. After the convention ended, Bush’s aggregate evaluation index increased
significantly from 2.55 to 2.61, while Gore’s evaluation index dropped from 2.48 to 2.43.
Aggregate change after the Republican convention suggests that Bush made a positive
impression on the electorate, while Gore’s favorability suffered. Table 2 summarizes the
impact of the subsequent 2000 Democratic convention.
       Given that the conventions were held only a week apart, it is interesting to note that
Gore’s pre-Democratic convention index from the Democratic panel was noticeably

       3. A fifth trait—“hypocritical”—was available, but was not included in the evaluation indexes
because of the conceptual distinction between positive and negative traits.
168    |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

TABLE 1
Aggregate Change in Candidate Evaluations, 2000 Republican Convention
                  Mean preconvention       Mean postconvention                              Difference
                   evaluation index         evaluation index              N                (Post–Pre)
Bush                     2.55                        2.61                1,163                .06*
Gore                     2.48                        2.43                1,163               -.05*
* p ⱕ .05 (two-tailed tests).

TABLE 2
Aggregate Change in Candidate Evaluations, 2000 Democratic Convention
               Mean preconvention      Mean postconvention
                evaluation index        evaluation index          N              Difference (Post–Pre)
Bush                  2.61                    2.54               1,163                  -.07*
Gore                  2.50                    2.70               1,163                  .20*
* p ⱕ .05 (two-tailed tests).

higher than his post-Republican convention index, while Bush’s favorability remained
static across both samples. Prior to the Democratic convention, Bush still held a favorabil-
ity advantage over Gore. However, Gore experienced a surge in aggregate favorability after
the Democratic convention, increasing from 2.50 to 2.70, while Bush’s index decreased
from 2.61 down to 2.54. From an aggregate standpoint, it appears that Gore was able to
make an impression strong enough to not only repair the damage done to his aggregate
favorability during the Republican convention, but also to knock Bush back down to
preconvention favorability levels. Taken together, the conventions allowed Gore to secure
a large increase in favorability, overtaking Bush, while perceptions of Bush experienced
little net change.
       It is also worthwhile to precede the individual-level analysis with a look at patterns
of exposure to information from the conventions. Table 3 displays a cross-tabulation of
exposure to the Democratic convention by exposure to the Republican convention. Data
are drawn from two questions from the 2000 NAES Democratic convention panel study.
Each cell contains the raw count of individuals and the percentage of the total sample.
       Of respondents, 59.8% reported watching at least some of the GOP convention,
and 63.8% reported watching at least some of the Democratic convention; it seems that
both conventions reached a wide section of the electorate. While the largest single
category of respondents (26.4%) did not watch any of either convention, the second-
largest single category of respondents (17.9%) were those who reported watching the
maximum amount of both conventions (two hours or more). Exactly 50% of the sample
reported watching at least some of both conventions, indicating a great deal of potential
for information from both conventions to exercise an impact on opinion across individu-
als. Just 9.7% of respondents reported watching at least some of the GOP convention but
none of the Democratic convention, while 13.9% of respondents reported watching at
least some of the Democratic convention but none of the Republican convention.
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                                 |     169

TABLE 3
Cross-tabulation of Exposure to Broadcast Portions of the 2000 Republican and Democratic
Conventions
                                                 Watched Democratic Convention
                                    Did not     Few             Half              Two hours
                                     watch     minutes          hour   One hour      +             Total
Watched GOP      Did not watch     324          48           40         35         46          493
 convention                         26.4%        3.9%         3.3%       2.9%       3.8%        40.2%
                 Few minutes        28          41           15          8         15          107
                                     2.3%        3.3%         1.2%         .7%      1.2%         8.7%
                 Half hour          26          19           45         18         22          130
                                     2.1%        1.6%         3.7%       1.5%       1.8%        10.6%
                 One hour           28           9           26         53         47          163
                                     2.3%          .7%        2.1%       4.3%       3.8%        13.3%
                 Two hours +        37          17           31         28        219          332
                                     3.0%        1.4%         2.5%       2.3%      17.9%        27.1%
                 Total             443         134          157        142        349         1225
                                    36.2%       10.9%        12.8%      11.6%      28.5%       100.0%

TABLE 4
Determinants of Postconvention Candidate Evaluation Indexes, 2000 Republican Convention
                                                         Bush                                 Gore
Preconvention index                               .663 (.026)*                            .813 (.050)*
Bush speech consumption                           .032 (.015)*                           -.049 (.018)*
Republican                                        .274 (.061)*                           -.240 (.071)*
Democrat                                         -.050 (.062)                             .287 (.075)*
Constant                                          .790 (.084)*                            .511 (.141)*
N                                                     694                                     693
Adjusted R2                                       .67                                     .50
OLS Regression.
Dependent variable = Postconvention candidate evaluation index.
* p ⱕ .05 (two-tailed tests).

       To explore how information from both conventions interacts to produce postcon-
vention opinion, we move on to our series of regressions. First, we model the individual-
level determinants of opinion regarding the candidates following the 2000 Republican
convention (see Table 4).
       Table 4 shows the regression model estimates for evaluations of George W. Bush
and Al Gore, respectively, after the 2000 Republican convention. Coefficients are shown
for each independent variable with standard errors in parentheses. Asterisks denote
statistically significant coefficients.
       Our model is similar to the model form used by Cera and Weinschenk (2012) to
analyze the 2004 conventions, and the results are in line with both the expectations
set by that work and our current hypotheses. Increased consumption of Bush’s
170    |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

convention speech yields significantly better postconvention evaluations for Bush and
worse evaluations for Gore, holding preconvention evaluations and individual partisan
affiliation constant. It appears that Bush successfully swayed opinion with his speech and
that the effect was felt across partisan lines.4
       As expected, individual partisanship is also a significant predictor of post-
convention evaluations; after controlling for preconvention evaluations and exposure
to Bush’s speech, Republicans are significantly more likely than other groups to give
Bush better ratings and Gore worse ratings. Democrats, meanwhile, gave Bush sig-
nificantly lower ratings than Republicans and Gore significantly higher ratings than
both Republicans and Independents (the excluded category) after the GOP convention.
These differences are evidence that the partisan convention atmosphere lent individual
partisanship more weight within individuals’ decision-making calculus after the 2000
Republican convention.
       Next, we model candidate evaluations for Gore and Bush after the 2000 Demo-
cratic convention. Because we expect post-Democratic convention opinion to be a func-
tion of exposure to information during the earlier Republican convention in addition to
the other factors, we add two additional models to this series of regressions: one in which
consumption of GOP convention information is allowed to exercise a direct linear effect
and one in which the slope coefficient for the speech exposure variable is allowed to vary
according to different values of consumption of GOP convention information. F-tests are
then used to determine whether or not allowing for information to exercise a conditional
effect on the persuasiveness of Gore’s convention speech adds additional explanatory
power to our models. Finally, we present figures illustrating how different levels of
exposure to GOP convention information impact the relationship between consumption
of Gore’s speech and post-Democratic convention opinion of the candidates (see Table 5).
       Table 5 displays three sets of regression model estimates for evaluations of Al Gore
and George W. Bush, respectively, after the 2000 Democratic convention. Coefficients
are shown for each independent variable with standard errors in parentheses. Asterisks
denote statistically significant coefficients and F-statistics.
       The first model specification mirrors that of the model for the earlier Republican
convention. Consumption of Gore’s speech yields improved evaluations of him, holding
other factors constant, but does not significantly impact individual evaluations of Bush.
While we find that individual partisanship serves as a significant predictor for postcon-
vention evaluations of Bush, it does not appear that there are significant differences
between groups of partisan identifiers when it comes to postconvention evaluations of
Gore. It is possible that this discrepancy is explained by the fact that partisanship was
activated as a strong predictor of views of Gore during the earlier GOP convention, which
was only a week earlier.
       The second model specification adds consumption of the earlier Republican
convention, which is significantly predictive of postconvention evaluations of both

        4. Analysis of alternative models, where the slope of the speech exposure variables are allowed to vary
by respondent partisanship, suggest that the effects of Bush’s speech on opinions of Gore cut across party lines
but that the positive impact of Bush’s speech on his own favorability was mostly limited to Republican
identifiers.
TABLE 5
Determinants of Postconvention Candidate Evaluation Indexes, 2000 Democratic Convention
                                                      Gore—model            Gore—conditional                        Bush—model             Bush—conditional
                                Gore—base           including GOP              information      Bush—base         including GOP               information
                                  model          convention consumption       effects model       model        convention consumption         effects model
Preconv. index                  .632(.030)*          .618 (.030)*             .621 (.030)*      .663 (.026)*       .657 (.026)*              .658 (.026)*
Gore speech                     .110(.016)*          .128 (.017)*             .162 (.029)*     -.015 (.015)       -.025 (.016)              -.074 (.027)*
Democrat                      .0089 (.073)           .084 (.073)              .079 (.073)       .029 (.067)        .033 (.067)               .040 (.067)
Republican                    -.140 (.074)          -.112 (.074)             -.122 (.074)       .282 (.070)*       .267 (.070)*              .281 (.070)*
GOP conv.                          —                -.041 (.012)*            -.016 (.022)           —              .024 (.011)*             -.013 (.020)
GOP conv. * Gore speech             —                     —                  -.015 (.010)           —                   —                    .021 (.009)*
Constant                      1.011 (.101)*         1.105 (.104)*            1.052 (.110)*      .672 (.094)*       .653 (.095)*              .717 (.099)*
N                                  774                   774                      774               773                773                       773
Adjusted R2                       0.58                  0.58                     0.58              0.62               0.62                      0.62
F-statistic                         —                                2.10                            —                             4.95*
OLS Regression.
Dependent variable = Postconvention candidate evaluation index.
* p ⱕ .05 (two-tailed tests).
                                                                                                                                                              Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS
                                                                                                                                                              |
                                                                                                                                                              171
172   |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

candidates, and in directions consistent with the expectations we laid out. Increased
consumption of the GOP convention yields lower post-Democratic convention evalua-
tions of Gore, and higher post-Democratic convention evaluations of Bush, even after
preconvention evaluations, Gore speech consumption, and individual partisan affiliation
are held constant. Consistent with our expectations, individuals who incorporated more
(ostensibly pro-Bush and anti-Gore) information from the earlier GOP convention into
their evaluations of the candidates assign less weight to any information incorporated
during the Democratic convention.
       The third specification is a multiplicative-conditional model that allows the impact
of the variable measuring exposure to Gore’s convention speech to vary based on values
of the variable measuring exposure to the earlier Republican convention. This modeling
decision is based on evidence that investment in a certain viewpoint generates bias
against incorporation of information inconsistent with that viewpoint encountered in
the future. After generating the model parameters for this specification, we ran F-tests
to determine if these alternative multiplicative-conditional models offered enhanced
explanatory power over the earlier linear-additive models. We find that for evaluations of
Bush, modeling a conditional effect of consumption of earlier convention information on
the relationship between exposure to Gore’s speech and postconvention evaluations allow
us to make significantly better predictions. Unfortunately, the regression parameters
alone do not allow for a meaningful discussion of these relationships, since the coefficients
for the interactions (as well as the coefficients for any terms included in the interactions)
do not represent average effects (Brambor, Clark, and Golder 2006). To overcome this
limitation, we graphically present the relevant effects in Figures 1 and 2.
       Figures 1 and 2 show how the impact of increased consumption of Gore’s conven-
tion speech on evaluations of Gore and Bush, respectively, varies based on different levels
of exposure to broadcasted portions of the earlier Republican convention. Figure 1 shows
that while increasing consumption of Gore’s speech improves evaluations of him across
all respondents, the effect is noticeably lower among those who took in larger amounts
of information during the earlier Republican convention. However, a nonsignificant
F-statistic suggests that this conditional effect does not offer enough additional explana-
tory power to justify its inclusion into models of postconvention evaluations of Gore.
Regardless, we view the failure of GOP messaging to significantly derail the positive
impression Gore made on viewers of his speech as evidence that positive impressions
candidates make are difficult to preempt.
       In the case of models of postconvention evaluations of Bush; the F-statistic is
significant, and Figure 2 shows that some of those with the highest investment in
information broadcast during the GOP convention actually gave Bush improved evalua-
tions as their consumption of Gore’s convention speech increased. This phenomenon is in
evidence even after controlling for individual partisan affiliation and pre-Democratic
convention evaluations. The coefficient for the speech exposure variable switching from
negative to positive for those that were exposed to the maximum amount of information
from the GOP convention is a clear example of a backfire effect, and compelling evidence
that positive impressions made by candidates during their conventions are also relatively
resistant to subsequent negative messaging from the opposition party.
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                      |   173

FIGURE 1. The Conditional Effect of Exposure to GOP Convention on the Relationship between
Consumption of Gore’s Convention Speech and Postconvention Evaluations of Gore.

                                     Conclusion

      In this article, we attempted to build on the existing individual-level understand-
ing of how presidential nominating conventions influence individual opinion. As a first
step, we replicated an earlier study of convention effects using panel data from the 2000
election cycle. We found additional evidence that evaluations of the presidential candi-
dates are shaped by a combination of persuasion arising from exposure to convention
speeches and the activation of partisan bias resulting from exposure to the abnormally
politicized atmosphere that pervades the convention period. Next, we iterated on this
model by drawing on the literature regarding political learning and bias in information
processing. We proposed that exposure to information from the first presidential nomi-
nating convention in a given election cycle can exercise a direct effect on individual
opinion that persists into the second convention. In addition, we proposed that this
exposure can also condition how information from the second convention is processed.
We found evidence of these direct and conditional effects during the 2000 conventions.
First, those who reported more exposure to the broadcast portions of the Republican
convention were less impacted by the subsequent Democratic convention, even after
controlling for partisan affiliation and preconvention opinion. Second, ceteris paribus,
increased exposure to the Republican convention broadcasts reduced the persuasive
174   |   PRESIDENTIAL STUDIES QUARTERLY / March 2012

FIGURE 2. The Conditional Effect of Exposure to GOP Convention on the Relationship between
Consumption of Gore’s Convention Speech and Postconvention Evaluations of Bush

impact of consumption of Gore’s speech on the favorability ratings of both candidates.
Among some respondents who consumed the maximum reportable amount of the GOP
convention broadcasts, consumption of Gore’s speech actually caused a backfire effect that
improved postconvention evaluations of Bush.
      These results demonstrate the importance of individual-level investigation to build-
ing a better understanding of campaign effects. The aggregate-level data clearly shows a
large net improvement in favorability for Al Gore after the 2000 conventions; however,
only individual-level investigation that includes measures of exposure to information
from both conventions and measures of individual characteristics allows us to appreciate
that Gore overcame a competing Republican message whose impact crossed party lines
and persisted well beyond the GOP convention. Parties and candidates have a clear
interest in maximizing the impact of their media advantage during conventions; our
research indicates that convening candidates would be well served by making the most of
opportunities to make direct positive impressions on the electorate. Such impressions
appear to be enduring and surprisingly resilient against competing negative messaging.

                                     References
Brambor, Thomas, William Clark, and Matt Golder. 2006 “Understanding Interaction Models:
   Improving Empirical Analyses.” Political Analysis 14 (1): 63-82.
Cera and Weinschenk / PRESIDENTIAL NOMINATING CONVENTIONS                                    |   175

Campbell, James E., Lynne Cherry, and Kenneth Wink. 1992. “The Convention Bump.” American
    Politics Quarterly, 20 (3): 287-307.
Cera, Joseph, and Aaron C. Weinschenk. 2012. “The Individual-level Effects of Presidential Conven-
    tions on Candidate Evaluations.” American Politics Research 40 (1): 3-28.
Ditto, P. H., and D. F. Lopez. 1992. “Motivated Skepticism: Use of Differential Decision Criteria for
    Preferred and Nonpreferred Conclusions.” Journal of Personality and Social Psychology 63 (4): 568-84.
Edwards, K., and E. E. Smith. 1996. “A Disconfirmation Bias in the Evaluation of Arguments.” Journal
    of Personality and Social Psychology 71 (1): 5-24.
Gelman, Andrew, and Gary King. 1993. “Why Are American Presidential Election Polls So Variable
    When Votes Are So Predictable?” British Journal of Political Science 23 (4): 409-51.
Hillygus, D. Sunshine, and Simon Jackman. 2003.“Voter Decision Making in Election 2000: Cam-
    paign Effects, Partisan Activation, and the Clinton Legacy”. American Journal of Political Science 47
    (4): 583-96.
Holbrook, Thomas M. 1996. Do Campaigns Matter? Thousand Oaks, CA: Sage Publications.
Lodge, Milton, Kathleen M. McGraw, and Patrick Stroh. 1989. “An Impression-driven Model of
    Candidate Evaluation.” American Political Science Review 83 (2): 399-420.
Lodge, M., and Taber, C. S. 2000. “Three Steps Toward a Theory of Motivated Political Reasoning.”
    In Elements of Reason: Understanding and Explaining the Limits of Political Rationality, eds. A. Lupia,
    M. D. McCubbins, and Samuel L. Popkin. London: Cambridge University Press.
Nyhan, Brendan, and Jason Reifler. 2010. “When Corrections Fail: The Persistence of Political
    Mmisperceptions.” Political Behavior 32 (2): 303-30.
Panagopoulos, Costas. 2007a. “Follow the Bouncing Ball: Assessing Convention Bumps, 1964-2004.”
    In Rewiring Politics: Presidential Nominating Conventions in the Media Age, ed. Costas Panagopoulos.
    Baton Rouge: Louisiana State University Press, 16-28.
———. 2007b. “Nominating Conventions, Campaign Events, and Political Information.” In Rewiring
    Politics: Presidential Nominating Conventions in the Media Age, ed. Costas Panagopoulos. Baton Rouge:
    Louisiana State University Press, 92-97.
Redlawsk, D. 2002. “Hot Cognition or Cool Consideration? Testing the Effects of Motivated Reason-
    ing on Political Decision Making.” Journal of Politics 64 (4): 1021-44.
Shaw, Daron R. 1999. “A Study of Presidential campaign Event Effects from 1952 to 1992.” Journal
    of Politic, 61 (2): 387-422.
Shafer, Byron E. 1988. Bifurcated Politics: Evolution and Reform in the National Party Convention. Cam-
    bridge, MA: Harvard University Press.
Stimson, James. 2004. Tides of Consent: How Public Opinion Shapes American Politics. New York: Cam-
    bridge University Press.
Taber, C. S., and Lodge, M. 2006. “Motivated Skepticism in the Evaluation of Political Beliefs.”
    American Journal of Political Science 50 (3): 755-69.
Trent, Judith S., and Robert V. Friedenberg. 2000. Political Campaign Communication: Principles and
    Practices. 4th ed. Westport, CT: Praeger.
You can also read