Content uploaded by Costas Panagopoulos
Author content
All content in this area was uploaded by Costas Panagopoulos on Dec 31, 2018
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=fbep20
Download by: [108.29.138.203] Date: 30 August 2016, At: 11:09
Journal of Elections, Public Opinion and Parties
ISSN: 1745-7289 (Print) 1745-7297 (Online) Journal homepage: http://www.tandfonline.com/loi/fbep20
Convention effects: examining the impact of
national presidential nominating conventions
on information, preferences, and behavioral
intentions
Aaron Weinschenk & Costas Panagopoulos
To cite this article: Aaron Weinschenk & Costas Panagopoulos (2016): Convention effects:
examining the impact of national presidential nominating conventions on information,
preferences, and behavioral intentions, Journal of Elections, Public Opinion and Parties, DOI:
10.1080/17457289.2016.1223678
To link to this article: http://dx.doi.org/10.1080/17457289.2016.1223678
Published online: 28 Aug 2016.
Submit your article to this journal
View related articles
View Crossmark data
Convention effects: examining the impact of national
presidential nominating conventions on information,
preferences, and behavioral intentions
Aaron Weinschenk
a
and Costas Panagopoulos
b
a
Political Science, University of Wisconsin, Green Bay, Green Bay, WI, USA;
b
Political Science,
Fordham University, Bronx, NY, USA
ABSTRACT
We ask whether and how US presidential nominating conventions matter in
contemporary US elections. Using individual-level panel data, we find
evidence that the conventions exert important effects on the electorate by
influencing post-convention intentions to participate in electoral politics,
knowledge about the candidates, and candidate favorability ratings, even
after controlling for pre-convention intentions, knowledge, and candidate
ratings. We conclude that conventions remain important campaign events
that play a role in facilitating democratic processes in America.
Introduction
The study of political campaigns has occupied an important place in political
science for decades now, but political scientists are still learning about the
ways in which campaigns influence voters. Much of the extant research has
focused on the effects of US presidential campaigns, which is not surprising,
given the salience and importance of presidential elections (Hillygus 2005;
Hillygus and Jackman 2003; Holbrook 1994,1996; Holbrook and McClurg
2009; Johnston, Hagen, and Jamieson 2004; Panagopoulos 2009,2012,
2013; Shaw 1999b; Wlezien and Erikson 2002). Many studies have focused
on the effects of campaign spending and other dimensions of campaigning,
including advertisements, candidate appearances, and campaign field offices
(Freedman, Franz, and Goldstein 2004; Gerber et al. 2011; Holbrook and
McClurg 2005; Masket 2009; Masket, Sides, and Vavreck 2016; Ridout et al.
2004; Shaw 1999b; Weinschenk 2015), although scholars have also been inter-
ested in determining whether and to what extent specific campaign events
impact political preferences (Erikson and Wlezien 2012; Hillygus and
Jackman 2003; Holbrook 1996,1999; Panagopoulos 2009,2012; Shaw
© 2016 Elections, Public Opinion & Parties
CONTACT Aaron Weinschenk weinscha@uwgb.edu Political Science, University of Wisconsin,
Green Bay, Green Bay, WI, USA
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES, 2016
http://dx.doi.org/10.1080/17457289.2016.1223678
1999a). During presidential general elections, the two types of events that
attract the greatest attention by virtue of their visibility and potential to
shape the dynamics of candidate preferences are the nominating conventions
and candidate debates.
In this study, we are interested in understanding how US presidential nomi-
nating conventions affect individual voters. We focus on conventions in the
US context, but note that there have been some excellent analyses of party
conventions in other contexts (see, e.g. Courtney 1995; Cross 1996; Perlin
1988; Purnnett 1992). Existing research on US conventions has tended to
focus on the impact of conventions on aggregate-level voter preferences
(Campbell, Cherry, and Wink 1992; Erikson and Wlezien 2012; Holbrook
1996; Johnston, Hagen, and Jamieson 2004; Panagopoulos 2007,2012;
Stimson 2004). During each election cycle, scholars and journalists often
measure the extent to which candidates get a “convention bump”following
their party’s nominating convention. The general consensus is that conven-
tions do provide a boost in candidate support; indeed, as Panagopoulos
and Endres (2015) note, on average, Democratic candidates get an 11-point
bump after their conventions and Republican candidates get a 10-point
bump after their conventions.
Although aggregate-level research on presidential conventions has pro-
vided a great deal of insight into campaign effects, these studies only tell
part of the story about conventions. To date, just a handful of studies have
used individual-level data to understand whether and how conventions
impact voters (see, e.g. Cera and Weinschenk 2012,2013; Hillygus and
Jackman 2003). Like aggregate-level studies, existing individual-level
studies have focused on candidate preferences, although individual-level
studies on conventions have been able to determine which types of individ-
uals are more (or less) likely to make up their minds about the candidates
during conventions (Panagopoulos and Endres 2015) or change their prefer-
ences after the conventions are over. For instance, Hillygus and Jackman
(2003) illustrate that, following the 2000 conventions, politically interested,
undecided independents became more likely to support Al Gore. Although
candidate preferences are important, we believe it is worthwhile to consider
the effects of conventions on a broader range of political attributes, including
knowledge, preferences, and behavioral intentions. While existing evidence
suggests that conventions influence vote choice (at both the aggregate
and individual levels), the effects of conventions on participation, political
knowledge, and candidate favorability ratings are much less clear. Do poten-
tial voters become mobilized to participate in the election as a consequence
of conventions? Do they acquire information about the candidates during the
conventions? Are they persuaded by the information about the candidates
they encounter during conventions? Given the collection of high-quality,
nationally representative panel data around the conventions, it is now
2A. WEINSCHENK AND C. PANAGOPOULOS
possible to provide a more nuanced analysis of how conventions affect indi-
vidual voters.
Before outlining how this article will proceed, it is important to note that
presidential conventions have been important events in presidential selection
for much of American history, in part because nominees are formally selected
by the conventions (Panagopoulos 2007). Interestingly, some of the perceived
importance of conventions has eroded in the modern era as conventions have
become coronations rather than deliberations (Panagopoulos 2008; Panago-
poulos and Endres 2015). This has led some scholars and political analysts
(e.g. Maisel, Rouse, and Wilson 2011) to argue that conventions may not
matter as much as they did in the past. Our analysis provides insights into
why conventions still matter.
We proceed in a straightforward manner. First, we present aggregate and
individual-level data on convention viewership. Second, we examine the
individual attributes that shape exposure to conventions. Lastly, we
analyze the impact of exposure to the convention speeches on a number
of important political attributes, many of which have not been considered
in previous individual-level analyses. Rather than treating conventions as
“intervening events”(Hillygus and Jackman 2003, 588), we measure exposure
to the conventions and examine its effects directly. Our analyses provide
unequivocal evidence that presidential conventions matter to potential
voters in significant ways. Although household viewership of the conven-
tions has been declining over time and few individuals watch every
moment of the convention speeches, the information generated during
these campaign events impacts a variety of important political attributes,
including intentions to get involved in electoral politics, knowledge about
candidate policy positions, and assessments of the candidates. Although
our analysis focuses on one county and one type of campaign event, we
encourage scholars to apply the framework developed here to other con-
texts and to other events that act as focal points for attention to a campaign
and at which candidates deliver speeches. Exposure to such events likely
matters for the indicators we explore in this paper, though we note that
this is an empirical question that is ripe for future study. Based on the
results presented in this article, we conclude that conventions remain impor-
tant campaign events that play a role in facilitating democratic processes in
America.
Does anyone pay attention to the convention anymore?
One criticism that is often leveled against modern presidential nominating
conventions is that no one pays attention anymore. In his book Tides of
Consent, Stimson (2004, 109) provides a description of how conventions
used to capture public attention:
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 3
At the dawn of the television age in the early 1950s, which is pretty much the
dawn of serious political polling as well, presidential nominating conventions
were raucous affairs, uncontrolled and unpredictable. Americans could tune in
their snowy black and white TV’s and see history happening in front of them.
And they did tune in. An electorate normally tuned out and uninterested
found the conventions pretty interesting and exciting, the Olympic Games of
politics.
He goes on to note, “As the conventions become ever more boring and
predictable, TV coverage has fallen off and viewership has declined. The
opportunity to see history being made is gone. The conventions merely
ratify what has long been known, long ago decided”(110). Figure 1
shows how household viewership of the presidential conventions has
changed over time. We use Nielsen ratings, which measure the percent of
households with TVs watching the convention during primetime hours.
While it is clear that viewership has dwindled, “it really is a nonsense argu-
ment to say no one watches the conventions,”as one political analyst has
pointed out.
1
It is also possible to get a sense of how individuals experience the nominat-
ing conventions by looking at survey data on exposure to the convention
Figure 1. Nielsen ratings of conventions (percent of households with TVs watching the
convention during primetime hours). Note: Graph made by authors using data from
http://www.centerforpolitics.org/crystalball/articles/reviewing-the-convention-ratings/.
1
http://usatoday30.usatoday.com/news/politics/story/2012-08-26/political-conventions-boost-ratings/
57331382/1.
4A. WEINSCHENK AND C. PANAGOPOULOS
speeches. During the 2004 presidential election, the National Annenberg Elec-
tion Study (NAES) conducted panel surveys around the nominating conven-
tions. In 2004, the NAES conducted four sets of interviews for the
convention panel studies: one before the Democratic National Convention
(DNC), one after the DNC (where the respondents interviewed before the con-
vention were re-interviewed), one before the Republican National Convention
(RNC), and one after the RNC (where the respondents interviewed before the
convention were re- interviewed).
2
In both the DNC and RNC post-convention
surveys, respondents were asked about the extent to which they watched
different convention speeches. In the DNC panel, respondents were asked
how much of Bill Clinton’s speech they saw or heard (did not see or hear,
few minutes, half hour, one hour, or entire speech), how much of John
Edwards’speech they saw or heard (did not see or hear, few minutes, 20
minutes, entire speech), and how much of John Kerry’s speech they saw or
heard (did not see or hear, few minutes, half hour, one hour, or entire
speech). The RNC study also asked three questions about the convention
speeches. The questions asked about exposure to Zell Miller’s speech (did
not see or hear, few minutes, 10 minutes, or entire speech), Dick Cheney’s
speech (did not see or hear, few minutes, 20 minutes, or entire speech),
and George Bush’s speech (did not see or hear, few minutes, half hour, one
hour, or entire speech).
3
Tables 1 and 2provide a look at the variation in
exposure to the DNC and RNC convention speeches. There is quite a bit of
variation in exposure to the convention speeches, but it is interesting to
note that in both conventions, a fairly large number of people saw at least
some of the convention speeches. Again, the idea that no one pays attention
to the conventions is inaccurate.
Table 1. Individual-level data on exposure to 2004 DNC speeches.
Exposure Did not see or hear A few minutes Half hour One hour Entire speech
Speech
Clinton 47.81% 19.34% 9.37% 5.81% 17.67%
Edwards 46.98% 18.39% 9.02% –25.62%
Kerry 37.13% 18.62% 10.32% 6.17% 27.76%
Notes: Data from 2004 Annenberg DNC Panel Study. For Edwards’speech, the column for “half hour”cor-
responds to “20 minutes”since the response categories were: did not see or hear, a few minutes, 20
minutes, and entire speech.
2
Unfortunately, the NAES did not conduct a survey in which the same respondents were interviewed
across both the RNC and DNC. None of the panel studies conducted by the NAES have done this,
although it would be incredibly useful to track the same people across both conventions in a given year.
3
Zell Miller was the Governor of Georgia and a US Senator. Although he was a Democrat, Miller backed
Bush over Kerry in the 2004 election. Despite being less prominent than Cheney or Bush, we believe
it is important to have multiple measures of speech exposure when possible. As we illustrate in a
later section, the reliabilities for the speech measures are quite high, which is comforting.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 5
Who tunes in and who tunes out?
Given the variation in exposure to the conventions above, it is interesting to
briefly consider what differentiates those who watch all of the convention
speeches from those who do not tune in at all. In Table 3, we present the
results of two ordinary least squares (OLS) regression models in which
measures of exposure to the convention speeches are used as dependent
variables. To construct measures of convention exposure, we averaged the
three speech measures discussed above together and re-scaled the variables
to run from 0 to 1, where 0 corresponds to not seeing any of the speeches and
1 corresponds to seeing all of the speeches. This produced an overall measure
of speech exposure for the RNC and an overall measure for the DNC. For the
RNC, Cronbach’s alpha for the speech measure is .83 and for the DNC, Cron-
bach’s alpha is .82, indicating high levels of reliability. We use age, race, parti-
sanship (standard 7-point scale; higher values correspond to a stronger
attachment to the Democratic party), income, education, political knowledge
(based on four factual questions), sex, the extent to which a respondent
follows politics, ideology (1–5, with 1 corresponding to “very conservative”
and 5 corresponding to “very liberal”), and the strength of respondent parti-
sanship as predictors. These variables capture important demographic and
political differences across individuals. Both regression models also include
state-fixed effects.
4
Overall, the results in Table 3 indicate that a number of variables predict
levels of exposure to the convention speeches. The most important predictors
of speech viewing are variables capturing political orientations; those who
Table 2. Individual-level data on exposure to 2004 RNC speeches.
Exposure Did not see or hear A few minutes Ten minutes One hour Entire speech
Speech
Miller 49.46% 14.91% 11.16% –24.46%
Cheney 47.98% 21.66% 8.17% –22.19%
Bush 37.83% 18.92% 9.25% 4.04% 29.97%
Notes: Data from 2004 Annenberg RNC Panel Study. For Cheney’s speech, categories are: did not see or
hear, few minutes, 20 minutes, or entire speech and for George Bush’s speech, categories are: did not
see or hear, few minutes, half hour, one hour, or entire speech.
4
In all models presented in this paper, we opt to use fixed-effects models (controlling for state) to account
for the idea that people in different states may differ in their levels of information, participation propen-
sities, campaign exposure, and candidate evaluations. This is the same approach taken by Huber and
Arceneaux (2007). We wanted to make sure that the campaign effects we are detecting are due to
the speeches rather than some differences in campaigning and/or information across states. This is a
conservative estimation strategy and the results from models estimated without state-fixed effects
are nearly identical to those reported in the paper. None of the statistically significant coefficients
that we report become insignificant in the models that do not include fixed effects and the magnitude
of the coefficients is very similar across estimation approaches. Results from the models can be obtained
from the authors on request. Due to the similarity of results, we choose to present the results from the
models that control for state.
6A. WEINSCHENK AND C. PANAGOPOULOS
report being interested in politics and who have strong attachments to a pol-
itical party watch more of the convention speeches than their counterparts.
Interestingly, the results in Table 3 provide some evidence of selective
exposure to political information. In both models, the measure of respondent
partisanship is statistically significant (p< .05, one-tailed). During the DNC,
Democrats watched more of the speeches than Republicans. During the
RNC, Republican identifiers watched more of the speeches than Democrats.
This is not to say that no Republicans watched the DNC or that no Democrats
watched the RNC. Instead, it appears that partisans watch more of speeches
that are given by co-partisans. Figure 2 shows the predicted effects of parti-
sanship on speech exposure (controlling for the other variables in the
models) for each convention. The solid line represents the predicted effect
of partisanship and the dashed lines represent the 95% confidence intervals.
As the panels illustrate, partisanship has a fairly pronounced impact on
exposure to the convention speeches. Although there appear to be a
number of individual-level factors that shape speech viewership, it is impor-
tant to take the analysis to the next level by considering whether exposure
to the speeches influences political knowledge, preferences, or behavioral
intentions. Indeed, just because some individual attributes predict speech
exposure, there is still an opportunity for the convention speeches to exert
Table 3. OLS regression models of exposure to the 2004 convention speeches.
DNC RNC
Variables b/se b/se
Income 0.00
0.01
0.01*
0.01
Education 0.00
0.01
−0.01
0.01
Age 0.00*
0.00
0.00*
0.00
White 0.02
0.05
−0.01
0.05
Sex −0.03
0.03
0.02
0.03
Follow politics 0.12*
0.02
0.13*
0.02
Partisanship strength (folded) 0.03*
0.02
0.03*
0.02
Partisanship (7-point measure) 0.05*
0.01
−0.03*
0.01
Political knowledge −0.01
0.02
0.02
0.02
Ideology 0.02
0.02
−0.04*
0.02
Constant −0.45*
0.13
−0.24*
0.12
No of Obs. 448 505
Adjusted R
2
0.21 0.23
State-fixed effects? Yes Yes
Notes: Cell entries are fixed-effects regression coefficients, controlling for state, predicting speech
exposure *p< .05 (one-tailed). Data from NAES panel studies.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 7
an independent influence on voters. Potential voters may be mobilized, per-
suaded, or learn new information as they watch the speeches. If the speeches
do not have an impact on potential voters, one might take that as evidence
that the conventions are not particularly useful institutions. If, on the other
hand, the conventions do exert effects on potential voters, it would lend
support to the argument that conventions remain an important and relevant
part of the presidential selection process.
The impact of conventions on the electorate
Most studies on campaign effects are based on the idea that exposure to cam-
paign information should impact voters. As Holbrook and McClurg nicely sum-
marize, “As campaigns become more active …voters should become better
informed, more enthused, and more motivated …” (2005, 693). Although
measures of campaign information can sometimes be fairly blunt (e.g. state
of residence), in this study, we are interested in using and assessing a more
direct measure of individual exposure to campaign information –specifically,
the convention speeches. The speeches given during the nominating conven-
tions provide an ideal opportunity to inform voters, generate excitement, and
shape how potential voters view the candidates. The candidates and their
allies get to speak directly to potential voters, without having their message
filtered through media outlets. Theoretically, we expect that exposure to
Figure 2. Effect of partisanship on exposure to convention speeches.
8A. WEINSCHENK AND C. PANAGOPOULOS
the convention speeches will have a number of effects on individuals’knowl-
edge, preferences, and behavioral intentions. In brief, exposure to campaign
information should have mobilization effects, informational effects, and per-
suasive effects (Holbrook 2010). Below, we elaborate on the specific hypoth-
eses we develop in the context of presidential conventions.
First, based on the literature on the mobilizing role of campaigns (Hillygus
2005; Holbrook and McClurg 2005; Holbrook and Weinschenk 2014), we
expect that during the conventions, as exposure to campaign information
(provided through the convention speeches) increases, potential voters will
become more likely to participate. By providing political information to
voters, campaigns should lower the costs of participation, increasing the like-
lihood that citizens get involved (Downs 1957). In this way, campaigns may
serve an important civic function.
In addition to a mobilizing effect, we expect that the convention speeches
will help to inform voters, another potential civic function of campaigns.
The convention speeches provide an opportunity for candidates and their
allies to spell out their policy positions, differentiate themselves from
opponents, and provide information on their qualifications for office. As
Stimson (2004) notes,
A public nearly always tuned out tunes in for a few days [during the conven-
tions] and those are opportunities to learn about people and programs, times
to change views. The more there is to learn, when candidates are unknown,
or suffer from inaccurate images as with vice presidents, the more important
the occasion. Nothing else in campaigning compares. (137–138)
Interestingly, Holbrook (2010, 13) has noted, “…little academic research
has been done on the role of conventions in voter education (a topic ripe
for future investigation).”We expect that as exposure to the conventions
speeches increases, accurate knowledge of candidate policy positions will
improve from the pre-convention period. Previous studies on the educative
effects of campaigns have provided some evidence that voters learn as
they are exposed to campaign events. Holbrook (1999), for example, finds
that voters acquire information about the candidates during debates. Given
that presidential conventions occur before the debates –when political infor-
mation is more scarce –it seems reasonable to think that the conventions will
improve voter knowledge (Holbrook 1996).
Finally, we hypothesize that exposure to the conventions will have a per-
suasive effect. Previous research on campaign advertisements has illustrated
that voters can be persuaded by exposure to campaign ads (Franz and Ridout
2007; Huber and Arceneaux 2007). Given that the conventions focus on por-
traying the convening candidate as positively as possible (while highlighting
the negative features of the opposing candidate), we expect that, as exposure
to the convention speeches during each convention increases, evaluations of
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 9
the convening candidate should improve (and evaluations of the opposing
candidate should worsen). Aggregate-level research on conventions has indi-
cated that the positive information and media coverage generated during
each party’s convention impacts preference dynamics. For instance, Johnston,
Hagen, and Jamieson (2004) note that during the 2000 election, support for
Bush surged after his convention speech and support for Gore increased
after his convention speech. As Wlezien and Erikson (2002, 988) point out,
“polls can move around quite a lot during campaigns, especially during the
late summer, in the period surrounding the conventions.”Conventions rep-
resent focal periods during an election cycle when voters can be influenced
by campaign information.
Participation intentions, knowledge, and candidate evaluations
Next, we focus on three central classes of variables of interest: participation
intentions, knowledge, and candidate evaluations. To measure participation
intentions, we use a question that asked respondents,
How likely are you to participate in this year’s presidential campaign, by working
to help one of the candidate’s campaigns, by donating money to a campaign, by
attending a campaign event of some kind, or by trying to convince others to
vote for him as well? Would you say you are very likely to do at least one of
these things, somewhat likely, or not very likely?
Responses are coded so that higher values correspond to a greater likeli-
hood of participating. This question was asked in both convention studies and
in both the pre- and post-convention surveys in each study. Due to the ordinal
nature of this measure, we estimate an ordered logit model (with state-fixed
effects).
To measure political knowledge, we created measures based on respon-
dent answers to nine factual questions about candidate policy positions
(e.g. “To the best of your knowledge, who favors allowing workers to invest
some of their Social Security contributions in the stock market –George
W. Bush, John Kerry, both, or neither?”). For each question, correct answers
were coded as 1 and incorrect answers and “don’t know”responses were
coded as 0. The nine knowledge questions were asked in both convention
studies and in both the pre- and post-convention surveys in each study. We
summed the nine factual knowledge measures together in each survey, yield-
ing pre- and post-convention measures of candidate knowledge. Higher
values correspond to more knowledge about policy positions. We estimate
the models using OLS regression (with state-fixed effects) due to the continu-
ous nature of these measures.
To measure how respondents felt about Bush and Kerry, we used candidate
favorability ratings. The two questions asked,
10 A. WEINSCHENK AND C. PANAGOPOULOS
On a scale of zero to 10, how would you rate [George W. Bush/John Kerry]? Zero
means very unfavorable, and 10 means very favorable. Five means you do not
feel favorable or unfavorable. Of course you can use any number between
zero and 10.
These questions were asked in both convention studies and in both the
pre- and post-convention surveys in each study. We estimate the models
using OLS regression (with state-fixed effects) due to the continuous nature
of the measures.
Results and analysis
We use the measures of participation intentions, knowledge, and candidate
evaluations described above as dependent variables in the empirical
models we estimate below. Because of the tight timing of the
surveys around the conventions (interviews were administered during the
10 days before the DNC and again during the 10 days after the convention
and for the RNC, respondents were interviewed during the 10 days before
the convention and were re-interviewed during the 11 days following the
convention), our models are quite simple.
5
For each convention and depen-
dent variable, we estimate a model that includes a lagged measure of the
dependent variable (which allows us to assess changes over time) and a
measure of direct exposure to the convention speeches. As a robustness
check, we also estimate more elaborate models by adding a number of
control variables. To be clear, all analyses that follow make use of the
2004 DNC and RNC panel studies that were conducted as a part of the
NAES.
Table 4 contains the results of the models of participation intentions. The
table contains two models for the DNC and two for the RNC. Overall, the
results indicate that exposure to the convention speeches has a statistically
significant effect (p< .05, one-tailed) on intentions to participate in the cam-
paign during the DNC. In both models, as exposure to the DNC convention
speeches increases, intentions to participate also increase, even after control-
ling for pre-convention participation intentions. In the RNC, the speech vari-
able is significant in the basic model, but the significance level drops once
controls are added. Thus, it appears that the DNC convention had a more pro-
nounced effect on participation intentions than the RNC. There are a number
of possible explanations for the different effects across the DNC and RNC. Pre-
vious research (Holbrook 1996,1999) has found that earlier campaign events
typically have a more pronounced impact than later ones; since the DNC
occurred first in 2004 (by convention, the out-party goes first), it could be
5
The 2004 RNC was held from 30 August–2 September. The 2004 DNC was held from 26 to 29 July.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 11
the case that the differences stem from the sequence of the 2004 conven-
tions. It is also worth noting that the different effects may stem from the
fact that the Republican candidate, George W. Bush, was the incumbent
and was better known than the Democratic candidate, John Kerry. Thus,
Kerry was the more novel (and less familiar) candidate and may have attracted
more interest than Bush. Previous research (Holbrook 1999; Jacobson 2013)
has demonstrated that information generated by candidates who are unfami-
liar to the electorate typically has a more pronounced effect on voters than
information generated by candidates who are familiar to the electorate. A
third possibility may have to do with the pool of potential voters during
each convention. If Republicans are more likely to turnout than Democrats,
it could be the case that there is a boost in participation intentions anytime
a Democratic convention takes place before a Republican convention. In
short, the pool of Democratic voters who have not yet committed to partici-
pating in the election may be larger than the Republican pool, and the DNC
may have been effective at “nudging”the pool of uncommitted Democrats
Table 4. Ordered logit models of post-convention participation intention, 2004.
DNC, Model 1 DNC, Model 2 RNC, Model 1 RNC, Model 2
Variables b/se b/se b/se b/se
Pre-convention participation intention 2.25*
0.17
2.10*
0.18
1.92*
0.15
1.76*
0.15
Speech exposure 1.84*
0.38
1.20*
0.43
0.57*
0.31
0.31
0.38
Income –0.10
0.08
–0.02
0.06
Education –0.03
0.07
–−0.10
0.06
Age –0.00
0.01
–−0.02
0.01
White –−0.22
0.44
–0.13
0.37
Sex –−0.03
0.28
–−0.07
0.24
Follow Politics –0.42*
0.14
–0.71*
0.18
Partisanship strength –0.42*
0.14
–0.66*
0.13
Political knowledge –0.06
0.15
–0.09
0.14
Any post-convention coverage –1.06*
0.35
–0.27
0.29
Partisanship –0.04
0.08
–0.01
0.07
Ideology –0.14
0.17
–0.19
0.15
No of Obs. 448 448 504 504
Pseudo R
2
0.38 0.41 0.31 0.36
State-fixed effects? Yes Yes Yes Yes
Notes: Cell entries are fixed-effects ordered logit coefficients, controlling for state, predicting post-conven-
tion participation intention. Cut points not shown to save space; full model results available from authors
on request *p< .05 (one-tailed). Data from NAES panel studies.
12 A. WEINSCHENK AND C. PANAGOPOULOS
to participate the in election, yielding a significant speech effect in the DNC
model.
6
Table 5 provides the results of OLS regression models that estimate the
effect of exposure to the convention speeches on knowledge about candidate
policy positions. Again, we present a basic model for each convention and
then a more elaborate model to check the robustness of results to the
inclusion of controls. Across the models, the speech variable has a statistically
significant effect in the expected direction. As exposure to the convention
speeches increases, knowledge about candidate policy positions improves,
even after controlling for pre-convention knowledge levels. This result holds
across the DNC and RNC models, although again the speech variable has a
slightly larger effect in the DNC model than in the RNC model.
Table 5. OLS regression models of post-convention policy knowledge, 2004.
DNC, Model 1 DNC, Model 2 RNC, Model 1 RNC, Model 2
Variables b/se b/se b/se b/se
Pre-convention knowledge 0.59*
0.04
0.46*
0.04
0.61*
0.03
0.48*
0.04
Speech exposure 0.77*
0.21
0.64*
0.23
0.57*
0.17
0.43*
0.19
Income –−0.02
0.04
–0.06
0.03
Education –0.07*
0.03
–0.07*
0.03
Age –−0.00
0.00
–0.00
0.00
White –0.36+
0.23
–0.02
0.19
Sex –0.19+
0.14
–0.13
0.13
Follow politics –0.24*
0.10
–0.34*
0.10
Partisanship strength –0.11+
0.07
–0.06
0.06
Political knowledge –0.23*
0.08
–0.17*
0.08
Any post-convention coverage –0.26+
0.18
–−0.04
0.16
Partisanship –0.03
0.04
–0.05+
0.04
Ideology –−0.05
0.08
–0.03
0.07
Constant 2.36*
0.19
0.46
0.59
2.26*
0.20
−0.21
0.49
No. of Obs. 443 443 430 430
Adjusted R
2
0.46 0.49 0.48 0.52
State-fixed effects? Yes Yes Yes Yes
Notes: Cell entries are fixed-effects regression coefficients, controlling for state, predicting post-convention
policy knowledge *p< .05 (one-tailed) +p< .10 (one-tailed). Data from NAES panel studies.
6
We thank two anonymous reviewers for suggesting ideas about why there may have been differential
effects for the RNC and DNC.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 13
In Tables 6 and 7, we turn to an analysis of how the speeches impact evalu-
ations of Bush and Kerry. Table 6 shows the results of a model in which post-
convention evaluations of Kerry are used as the dependent variable. The first
two columns present the results for the DNC and the last two columns show
the results for the RNC. As expected, exposure to the speeches exerts a stat-
istically significant effect (p< .05, one-tailed) in the expected direction on
evaluations of Kerry. During the DNC, as exposure to the DNC speeches
increases, respondents’views of Kerry improve; during the RNC, as exposure
to the RNC speeches increases, views of Kerry worsen. Given the content of
speeches, it makes sense that the Democratic speeches would help Kerry
and that exposure to Republican speeches would negatively impact evalu-
ations of Kerry. There does not appear to be an interaction between respon-
dent partisanship and speech exposure. Previous research has suggested that
the persuasive effects of campaign information may be enhanced when the
candidate delivering the information and the listener are co-partisans and
diluted when the partisanship of the candidate and the listener are mis-
matched (Cera and Weinschenk 2012). The interaction between speech
exposure and party identification was not significant in the DNC Model 2,
(p= .42) or the RNC Model 2, (p= .09).
The results in Table 7 show the effects of speech exposure on post-conven-
tion evaluations of Bush (controlling for pre-convention evaluations). Again,
we present results for the DNC and RNC. As expected, exposure to the
Table 6. OLS regression model of post-convention evaluations of John Kerry, 2004.
DNC, Model 1 DNC, Model 2 RNC, Model 1 RNC, Model 2
Variables b/se b/se b/se b/se
Pre-convention Kerry rating 0.83*
0.02
0.66*
0.03
0.85*
0.02
0.73*
0.02
Speech exposure 0.72*
0.19
0.62*
0.19
−0.54*
0.15
−0.33*
0.15
Income –−0.03
0.03
–−0.03
0.02
Any post-convention coverage –−0.06
0.15
–−0.15
0.13
White –−0.53*
0.20
–0.05
0.16
Sex –−0.06
0.12
–0.08
0.10
Partisanship –0.26*
0.04
–0.20*
0.04
Ideology –0.21*
0.07
–0.12*
0.06
Constant 1.00*
0.12
0.97*
0.32
0.96*
0.12
0.51*
0.28
No. of Obs. 697 697 764 764
Adjusted R
2
0.78 0.80 0.83 0.84
State-fixed effects? Yes Yes Yes Yes
Notes: Cell entries are fixed-effects regression coefficients, controlling for state, predicting post-convention
Kerry ratings *p< .05 (one-tailed) +p< .10 (one-tailed). Data from NAES panel studies.
14 A. WEINSCHENK AND C. PANAGOPOULOS
Democratic speeches during the DNC has a negative impact on evaluations of
Bush. In both models, the speech variable is statistically significant at p< .05
(one-tailed). Exposure to the Republican speeches during the RNC, on the
other hand, has a positive effect on assessments of Bush. In both models,
the coefficients on the speech measure are statistically significant at p< .05
(one-tailed). Once again, there does not appear to be an interaction effect
between respondent partisanship and exposure to the speeches in the
models shown in Table 7 (interaction between speech exposure and party
identification not significant in either the DNC Model 2, (p= .99) or the RNC
Model 2, (p= .62).
It is worth noting that our results in Tables 6 and 7connect to Wlezien and
Erikson (2002), Erikson and Wlezien (2012) work on presidential elections.
Although Wlezien and Erikson (2002) use aggregate data to examine the
dynamics of vote preferences over the course of presidential elections
(while we use individual-level data), the authors find evidence that polls
can move around quite a bit during the conventions. Our analysis of candidate
favorability ratings provides additional evidence that the conventions are
important in shaping how voters evaluate the candidates. The speeches
given during the conventions seem to exert persuasive effects on potential
voters –leading them to update their assessments of the candidates as
new information comes in. Although the respondents in the panel studies
we use were not re-interviewed again after the post-convention survey was
Table 7. OLS regression model of post-convention evaluations of George W. Bush, 2004.
DNC, Model 1 DNC, Model 2 RNC, Model 1 RNC, Model 2
Variables b/se b/se b/se b/se
Pre-convention Bush rating 0.85*
0.02
0.72*
0.02
0.89*
0.01
0.76*
0.02
Speech exposure −0.47*
0.18
−0.36*
0.19
0.49*
0.15
0.45*
0.16
Income –−0.03
0.03
–−0.06*
0.02
Any post-convention coverage –0.08
0.15
–−0.14
0.13
White –0.45*
0.19
–0.05
0.16
Sex –−0.14
0.12
–−0.02
0.10
Partisanship –−0.23*
0.04
–−0.23*
0.04
Ideology –−0.20*
0.07
–−0.17*
0.06
Constant 1.14*
0.14
3.02*
0.42
0.64*
0.09
3.13
0.37
No. of Obs. 702 702 767 767
Adjusted R
2
0.83 0.85 0.86 0.88
State-fixed effects? Yes Yes Yes Yes
Notes: Cell entries are fixed-effects regression coefficients, controlling for state, predicting post-convention
Bush ratings *p< .05 (one-tailed) + p< .10 (one-tailed). Data from NAES panel studies.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 15
administered, it is interesting to consider how long the campaign-induced
effects we observe here may persist. Interestingly, Wlezien and Erikson
(2002, 2012) find evidence that, in the aggregate, the effects of the conven-
tions persist to Election Day. As they note, “Although the convention
season is the time for multiple bounces in the polls, one party ends up with
an advantage when the dust clears. And this gain is a net convention bump
rather than a bounce”(2012, 77).
As a final bit of analysis, we include in Figures 3 and 4visualizations of the
predicted effects of exposure to the convention speeches on the dependent
variables from the models displayed above (only statistically significant effects
are shown). Figure 3 shows the predicted effects during the DNC and Figure 4
displays the predicted effects during the RNC (controlling for the other vari-
ables in the models). The solid lines represent the predictions and the
dashed lines represent the 95% confidence intervals. Overall, the predictions
shown in these figures represent important campaign effects. Although some
of the effects are more pronounced than others, it is important to remember
that these changes are occurring over a fairly short time span and that our
models include lagged dependent variables. We take our results as an indi-
cation that the conventions exert meaningful effects on the electorate.
Although our dataset does not enable us to determine how long these
changes persist, we believe that studying decay effects would be an excellent
Figure 3. Predicted effects of speech exposure on post-convention measures during
DNC.
16 A. WEINSCHENK AND C. PANAGOPOULOS
question for future study. Obtaining panel data around campaign events rep-
resents an important way to learn about the effects of those events, but it
would be useful to re-interview people multiple times after campaign
events to determine if campaign effects are durable or if they decay over time.
Do conventions matter? Yes.
In this study, we investigated the extent to which presidential nominating
conventions impact potential voters. Most previous studies on US presidential
conventions have focused on aggregate convention bumps. Generally speak-
ing, those analyses have found that conventions have an impact on the
dynamics of candidate support, often exerting effects that persist through
Election Day (Erikson and Wlezien 2012; Stimson 2004). Our research comp-
lements and builds upon existing work on convention effects. We used indi-
vidual-level data to get a sense of how the conventions impact the electorate.
Few studies on conventions have leveraged individual-level survey datasets to
understand how conventions impact the electorate. We gauged exposure to
the conventions by developing a measure of speech viewing and focused
on the effects of exposure to the conventions on three important political
indicators –participation intentions, knowledge of candidate positions, and
candidate assessments. Our analyses, which made use of high-quality, panel
Figure 4. Predicted effects of speech exposure on post-convention measures during
RNC.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 17
data collected immediately before and after the 2004 conventions, provided
solid evidence that the conventions do impact potential voters. As exposure
to the convention speeches increases, the electorate becomes more likely to
participate in the campaign and more knowledgeable about the candidates
and their policy positions. People also appear to update their assessments
of the candidates based on information generated during the conventions.
Although the conventions influence how voters feel about the candidates,
they also help voters acquire information and mobilize them to engage in
electoral politics. Thus, the effects of nominating conventions are broader
than previously thought. Conventions are important opportunities for poten-
tial voters to engage with candidates and represent focal events that have the
capacity to perform central civic functions.
Our study has a number of limitations. For instance, as we illustrated above,
there are attributes (e.g. partisanship, ideology) that lead some people to self-
select into watching the conventions. In short, exposure to the conventions is
not randomly assigned. Although we did not have access to experimental
data on the conventions, our use of panel data enabled us to include
lagged dependent variables in our empirical models. This approach will not
assuage all concerns about selection issues, but it does enable us to control
for baseline levels of knowledge, participation intentions, and candidate
favorability and to examine the extent to which speech exposure leads to
changes over time, something that cross-sectional datasets do not allow.
Future studies may be able to build on our statistical models when trying
to study campaign effects. Another limitation is that we have only studied
one particular election. In order to be able to make comparisons across elec-
tions, researchers will need to replicate our study using data collected in other
years and contexts. In addition, we acknowledge that we have focused solely
on the USA, but we believe that our central argument extends beyond a single
national context and applies to focal events which attract voting audiences
who can be exposed to campaign information by candidates and party
leaders. We expect our findings about the effects of exposure to intense cam-
paign communications, persuasive rhetoric and informational content may
well generalize to these environments, and we encourage scholars to
examine the effects of candidate speeches on mobilization, knowledge, and
persuasion in other contexts. We believe that our approach and modeling
strategy is applicable to other kinds of campaign events or speeches
broadly defined.
In the end, we believe that additional work needs to be done on presiden-
tial conventions. Scholars can engage questions about the generalizability of
our findings by conducting similar analyses in other presidential election
cycles. Some preexisting datasets can potentially facilitate analyses along
these lines, but we would also encourage the collection of new data. As we
noted, it would be interesting to conduct a panel study in which respondents
18 A. WEINSCHENK AND C. PANAGOPOULOS
are interviewed multiple times across both major party conventions. Such a
dataset would allow for an even more nuanced look at how campaign
events affect individual voters. It would also be interesting to examine the
extent to which the effects of candidate speeches vary based on individual
attributes. Scholars may be able to develop and test hypotheses about
traits that influence how receptive individuals are to convention speeches.
There has been some work on differences in receptivity to campaign infor-
mation (Fridkin and Kenney 2004; Hillygus 2005), but little is known about
the extent to which different traits (e.g. demographics, political attitudes,
and personality attributes) shape receptivity to convention speeches.
Disclosure statement
No potential conflict of interest was reported by the authors.
Notes on Contributors
Aaron Weinschenk is Assistant Professor of Political Science at the University of Wis-
consin-Green Bay. His research focuses on campaigns, elections, and political partici-
pation. He has written more than a dozen journal articles and one book on these topics.
Costas Panagopoulos is Professor of Political Science at Fordham University. He has
authored more than 60 journal articles and 7 books on campaigns, elections, and
voting behavior. He directs the Center for Electoral Politics and Democracy and also
the Graduate Program in Elections and Campaign Management at Fordham.
References
Campbell, J. E., L. Cherry, and K. Wink. 1992.“The Convention Bump.”American Politics
Quarterly 20: 287–307.
Cera, J., and A. Weinschenk. 2012.“The Individual-level Effects of Presidential
Conventions on Candidate Evaluations.”American Politics Research 40: 3–28.
Cera, J., and A. Weinschenk. 2013.“Polls and Elections: The Conditional Effects of
Competing Messages During Presidential Nominating Conventions.”Presidential
Studies Quarterly 42: 161–175.
Courtney, J. 1995.Do Conventions Matter?: Choosing National Party Leaders in Canada.
Montreal: McGill-Queen’s Press.
Cross, W. 1996.“Direct Election of Provincial Party Leaders in Canada, 1985-1995: The
end of the Leadership Convention?”Canadian Journal of Political Science 29: 295–
315.
Downs, A. 1957.An Economic Theory of Democracy. New York: Harper.
Erikson, R., and C. Wlezien. 2012.The Timeline of Presidential Elections: How Campaigns
do (and do not) Matter. Chicago: University of Chicago Press.
Franz, M., and T. Ridout. 2007.“Does Political Advertising Persuade?”Political Behavior
29: 465–491.
Freedman, P., M. Franz, and K. Goldstein. 2004.“Campaign Advertising and Democratic
Citizenship.”American Journal of Political Science 48: 723–741.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 19
Fridkin, K., and P. Kenney. 2004.“Do Negative Messages Work? The Impact of Negativity
on Citizens’Evaluations of Candidates.”American Politics Research 32: 570–605.
Gerber, A., J. Gimpel, D. Green, and D. R. Shaw. 2011.“How Large and Long-Lasting are
the Persuasive Effects of Televised Campaign ads? Results from a Randomized Field
Experiment.”American Political Science Review 105: 135–150.
Hillygus, S. 2005.“Campaign Effects and the Dynamics of Turnout Intention in Election
2000.”The Journal of Politics 67: 50–68.
Hillygus, S., and S. Jackman. 2003.“Voter Decision Making in Election 2000: Campaign
Effects, Partisan Activation, and the Clinton Legacy.”American Journal of Political
Science 47: 583–596.
Holbrook, T. 1994.“Campaigns, National Conditions, and U.S. Presidential Elections.”
American Journal of Political Science 38: 973–998.
Holbrook, T. 1996.Do Campaigns Matter? Thousand Oaks, CA: Sage.
Holbrook, T. 1999.“Political Learning from Presidential Debates.”Political Behavior 21:
67–89.
Holbrook, T. 2010.“Do Campaigns Really Matter?”In The Electoral Challenge: Theory
Meets Practice, edited by Stephen Craig and David Hill, 2nd ed, 1–20. Thousand
Oaks, CA: CQ Press.
Holbrook, T., and S. McClurg. 2005.“The Mobilization of Core Supporters: Campaigns,
Turnout, and Electoral Composition in United States Presidential Elections.”
American Journal of Political Science 49: 689–703.
Holbrook, T., and S. McClurg. 2009.“Living in a Battleground: Presidential Campaigns
and Fundamental Predictors of Vote Choice.”Political Research Quarterly 62:
495–506.
Holbrook, T., and A. Weinschenk. 2014.“Campaigns, Mobilization, and Turnout in
Mayoral Elections.”Political Research Quarterly 67: 42–55.
Huber, G., and K. Arceneaux. 2007.“Identifying the Persuasive Effects of Presidential
Advertising.”American Journal of Political Science 51: 957–977.
Jacobson, G. 2013.The Politics of Congressional Elections. 8th ed. New York: Pearson.
Johnston, R., M. Hagen, and K. Jamieson. 2004.The 2000 Presidential Election and the
Foundations of Party Politics. Cambridge: Cambridge University Press.
Maisel, S., J. Rouse, and R. Wilson. 2011.“Unconventional Wisdom: The Future of
Presidential Nominating Conventions.”The New England Journal of Political Science
V: 229–264.
Masket, S. 2009.“Did Obama’s Ground Game Matter? The Influence of Local Field
Offices during the 2008 Presidential Election.”Public Opinion Quarterly 73: 1023–
1039.
Masket, S., J. Sides, and L. Vavreck. 2016.“The Ground Game in the 2012 Presidential
Election.”Political Communication 33: 169–187.
Panagopoulos, C. 2007.“Follow the Bouncing Ball: Assessing Convention Bumps, 1964-
2004.”In Rewiring Politics: Presidential Nominating Conventions in the Media age,
edited by Costas Panagopoulos, 16–28. Baton Rouge: LSU Press.
Panagopoulos, C. 2008.“Presidential Nominating Conventions: Past, Present and
Future.”The Forum 5: 1–12.
Panagopoulos, C. 2009.“Campaign Dynamics in Battleground and Nonbattleground
States.”Public Opinion Quarterly 73: 119–129.
Panagopoulos, C. 2012.“Campaign Context and Preference Dynamics in U.S.
Presidential Elections.”Journal of Elections, Public Opinion & Parties 22: 123–137.
Panagopoulos, C. 2013.“Campaign Effects and Preference Dynamics in the 2012
Election.”The Forum 10: 36–39.
20 A. WEINSCHENK AND C. PANAGOPOULOS
Panagopoulos, C., and K. Endres. 2015.“The Enduring Relevance of National
Nominating Conventions.”The Forum 13: 559–576.
Perlin, G. 1988.Party Democracy in Canada: The Politics of National Party Conventions.
Scarborough: Prentice-Hall Canada.
Purnnett, R. 1992.Selecting the Party Leader: Britain in Comparative Perspective. London:
Harvester Wheatsheaf.
Ridout, T. N., D. V. Shah, K. M. Goldstein, and M. M. Franz. 2004.“Evaluating Measures of
Campaign Advertising Exposure on Political Learning.”Political Behavior 26:
201–225.
Shaw, D. R. 1999a.“A Study of Presidential Campaign Event Effects from 1952 to 1992.”
The Journal of Politics 61: 387–422.
Shaw, D. R. 1999b.“The Effect of TV ads and Candidate Appearances on Statewide
Presidential Votes, 1988-96.”American Political Science Review 93: 345–361.
Stimson, J. 2004.Tides of Consent: How Public Opinion Shapes American Politics.
New York: Cambridge University Press.
Weinschenk, A. 2015.“Campaign Field Offices and Voter Mobilization in 2012.”
Presidential Studies Quarterly 45: 573–580.
Wlezien, C., and R. Erikson. 2002.“The Timeline of Presidential Election Campaigns.”The
Journal of Politics 64: 969–993.
JOURNAL OF ELECTIONS, PUBLIC OPINION AND PARTIES 21
