Content uploaded by André Blais
Author content
All content in this area was uploaded by André Blais on May 05, 2017
Content may be subject to copyright.
Content uploaded by Elisabeth Gidengil
Author content
All content in this area was uploaded by Elisabeth Gidengil on Jul 03, 2014
Content may be subject to copyright.
Full Terms & Conditions of access and use can be found at
http://www.tandfonline.com/action/journalInformation?journalCode=upcp20
Download by: [Bibliothèques de l'Université de Montréal] Date: 03 May 2017, At: 07:32
Political Communication
ISSN: 1058-4609 (Print) 1091-7675 (Online) Journal homepage: http://www.tandfonline.com/loi/upcp20
Election Campaigns as Information Campaigns:
Who Learns What and Does it Matter?
Richard Nadeau , Neil Nevitte , Elisabeth Gidengil & André Blais
To cite this article: Richard Nadeau , Neil Nevitte , Elisabeth Gidengil & André Blais (2008)
Election Campaigns as Information Campaigns: Who Learns What and Does it Matter?, Political
Communication, 25:3, 229-248, DOI: 10.1080/10584600802197269
To link to this article: http://dx.doi.org/10.1080/10584600802197269
Published online: 30 Jul 2008.
Submit your article to this journal
Article views: 867
View related articles
Citing articles: 23 View citing articles
Political Communication, 25:229–248, 2008
Copyright © Taylor & Francis Group, LLC
ISSN: 1058-4609 print / 1091-7675 online
DOI: 10.1080/10584600802197269
229
UPCP1058-46091091-7675Political Communication, Vol. 25, No . 3, June 2008: pp. 1– 36Political Communication
Election Campaigns as Information Campaigns:
Who Learns What and Does it Matter?
Election Campaigns as Information CampaignsRichard Nadeau et al.RICHARD NADEAU, NEIL NEVITTE,
ELISABETH GIDENGIL, and ANDRÉ BLAIS
During election campaigns political parties compete to inform voters about their
leaders, the issues, and where they stand on these issues. In that sense, election
campaigns can be viewed as a particular kind of information campaign. Democratic
theory supposes that participatory democracies are better served by an informed
electorate than an uninformed one. But do all voters make equal information gains
during campaigns? Why do some people make more information gains than others?
And does the acquisition of campaign information have any impact on vote intentions?
Combining insights from political science research, communications theory, and social
psychology, we develop specific hypotheses about these campaign information dynam-
ics. These hypotheses are tested with data from the 1997 Canadian Election Study,
which includes a rolling cross-national campaign component, a post-election compo-
nent, and a media content analysis. The results show that some people do make more
information gains than others; campaigns produce a knowledge gap. Moreover, the
intensity of media signals on different issues has an important impact on who receives
what information, and information gains have a significant impact on vote intentions.
Keywords information gains, knowledge gap, electoral campaigns, media coverage,
voting choices
Voters’ lack of political information raises important concerns about contemporary
democratic practices. Those concerns encourage questions not only about how voters
compensate for these information shortfalls (Lupia & McCubbins 1998) but also about
which voters acquire politically relevant information given the opportunity to do so.
Election campaigns involve intense exchanges of political information. Political
parties and leaders have powerful incentives to inform voters about their positions on the
issues of the day, and the media cover and relay these messages. For voters, campaigns are
opportunities to gather politically relevant information that can help them make informed
choices (Bartels, 1996; Popkin, 1994). Election campaigns and information campaigns are
increasingly being conceptualized in similar ways. Holbrook (1996, 2002), for example,
identifies election campaigns as a particular type of information campaign, while Zaller
Richard Nadeau is Professor in the Department of Political Science, University of Montréal.
Neil Nevitte is Professor in the Department of Political Science, University of Toronto. Elisabeth
Gidengil is Hiran Mills Professor in the Department of Political Science, McGill University. André
Blais is Professor in the Department of Political Science, University of Montréal.
Address correspondence to Richard Nadeau, Department of Political Science, University
of Montréal, C.P. 6128, succursale Centre-ville, Montreal, Canada H3C 1J7. E-mail:richard.
nadeau@umontreal.ca
230 Richard Nadeau et al.
(1989) views election campaigns as the encounter of information flows resulting from the
competition between rival information campaigns.
Investigations of the information diffused by parties, its delivery by the media, and its
absorption by voters during campaigns (e.g., Norris et al., 1999) have deepened our
understanding of the role of information in campaigns. But taken singly, these advances
have been less useful for developing precise hypotheses about the dynamics of campaign
information. A richer perspective on the diffusion, penetration, and consequences of
campaign information emerges when these insights are combined (McGuire, 1968;
Tichenor, Donoghue, & Olien, 1970; Kwak, 1999).
A particularly promising line of investigation opens up, we argue, when Converse’s
(1962) and Zaller’s (1989) work on information flows is linked with communication
researchers’ insights concerning the knowledge gap (Tichenor et al., 1970; Gaziano &
Gaziano, 1996; Viswanath & Finnegan, 1996; Kwak, 1999). Integrating these perspec-
tives brings key research questions into sharp focus: Who receives what information, in
what context, and with what effects?
This analysis begins by combining political science and communications research to
develop conceptual and operational definitions of political information. After outlining theo-
ries about the knowledge gap, we propose a Converse–Zaller model to explore the effects of
information on electoral choices. Data from the Canadian Election Study are used to test
eight hypotheses. Four concern the penetration of political information during the campaign
among groups, and four focus on the effects of political information on individual vote
choices. The design of the Canadian Election Study of 1997 is uniquely suited to address
these questions. It is the first to incorporate all the necessary building blocks for testing the
knowledge gap hypothesis during an election campaign. These provide the first explicit
demonstration that information gains are more consequential for moderately sophisticated
voters. The novel methodology, combined with insights from political and communication
sciences, provides a deeper understanding of the still understudied phenomenon of information
acquisition during campaigns (Mendelsohn & Cutler, 2000). It also provides a firmer founda-
tion for preceding research findings while opening up new perspectives for future research.
Communication specialists such as Tichenor et al., (1970) worry that information
campaigns benefit already well-informed individuals and widen the “knowledge gap”
between the “information rich” and the “information poor.” Those concerns turn out to be
well founded. The broader implications of these findings for campaigns, for persuasion,
and for democratic theory are considered in the conclusions.
Electoral Campaigns as Information Campaigns: The Theoretical Approach
General and Campaign-Specific Information
Most typologies of political information (Zaller, 1991; Price & Zaller, 1993) distinguish
between general political information and campaign-specific information. Converse
(1962) explicitly distinguishes between the notion of “mass of stored information” and
that of “current information intake” just as communication researchers (Chaffee, Zhao, &
Leshner, 1994, p. 306; see also Kwak, 1999) distinguish between “pre-campaign
knowledge” and “campaign knowledge.”
Communication researchers and political scientists also agree on two other funda-
mentals. First, respondents’ knowledge of candidates’ and parties’ issue positions are
key indicators of information gains registered during the course of a campaign (Zaller,
1991; Chaffee et al., 1994; Chaffee & Kanihan, 1997; Kwak, 1999, pp. 396–397).1
Election Campaigns as Information Campaigns 231
Second, general political information (see Tichenor et al., 1970; Ettema & Kline, 1977;
Chaffee et al., 1994; Gaziano & Gaziano, 1996) best predicts information gains in
specific contexts such as elections.
General and campaign-specific information thus carry quite different meanings. The
amount of general information about politics, the general stock of information (GSI), indicates
“political awareness,” a concept Zaller applies to voters’ aptitude and motivation for absorbing
information during campaigns (1991). Campaign-specific information (CSI), by contrast,
measures what voters learn about parties’ position on various issues during campaigns.
The distinction between general and campaign-specific information is vital to the
analysis of election campaigns as information campaigns. We argue that the knowledge
gap between the information rich and the information poor grows during the course of a cam-
paign. The general information variable (GSI) allows us to sort individuals into these groups
and then to distinguish those who should make the biggest information gains (CSI) during the
campaign. After specifying the Converse-Zaller model, hypotheses linking electoral choices,
general information and campaign information gains are tested. Reliable measures of both
general and campaign specific information are, of course, essential for identifying those
voters most susceptible to modifying their electoral choices during a campaign.2
The Knowledge Gap
Does campaign-specific information penetrate different segments of the electorate in the
same way? Or is there an information or knowledge gap? Tichenor et al.’s classic account
argues that as “the infusion of mass media information into a social system increases, seg-
ments of the population with higher socio-economic status tend to acquire this information
at a faster rate than the lower status segments, so that the gap in knowledge between the
segments tends to increase rather than decrease” (1970, pp. 159–160). They speculate that
“one principal function of a presidential campaign would be to increase the difference in
level of information at educational extremes” and conclude that “the prospects for closing
knowledge gaps in broad areas of science and public affairs appear dismal” (1970, p. 170).
That pessimism springs both from the nature of political information and the specific
context of election campaigns. The social distribution of electoral information is complex
(Tichenor et al., 1970) and of limited pertinence (Ettema & Kline, 1977), and campaigns
are brief.3 Consequently, Tichenor et al., speculate that the knowledge gap applies
“primarily to public affairs and science news” (1970, p. 160).
Holbrook’s (2002) comprehensive analysis of knowledge gaps during six presidential
campaigns (1976–1996) demonstrates that “knowledge gaps do not always grow over the
course of presidential campaigns and that some events, such as debates, may actually
reduce the level of information inequality in the electorate” (p. 437). But Holbrook’s
approach can be sharpened. His key dependent variable, the number of good and bad points
a voter can elicit about a candidate during a campaign, may measure accumulated impres-
sions rather than information gains. Second, by using education as the key explanatory
variable, it becomes difficult to discriminate between the varying levels of media coverage
devoted to different campaign issues.
Zaller (1989) observes that campaigns can be differentiated both according to their
intensity (high for presidential elections, medium for senatorial campaigns, and generally
weak for elections to the House of Representatives) and the partisan balance of their
communication flows. That same logic, we argue, is also applicable to the analysis of
different issues within a single campaign: During agenda-setting struggles, the media
cover different campaign issues with different levels of intensity.
232 Richard Nadeau et al.
There is no reason to believe that there is a simple linear relationship between the
evolution of the information gap and the intensity of the media issue coverage during a
campaign. Indeed, we hypothesize that when campaign issues are covered intensely and
continuously (strong signal), all categories of voters should be able to make some infor-
mation gains, and so the information gap on that issue should grow relatively little during
the campaign (Hypothesis 1a). Alternatively, when an issue signal is very weak, even
attentive citizens will find it difficult to make significant and continuous gains. Under
those conditions, the expectation is that the gap between the information rich and the
information poor will increase only slightly if at all (Hypothesis 1c). When a signal is
strong enough to be heard by the most attentive but not too weak to be noticed by the less
attentive (medium signal), however, only the information rich will make learning gains.
Thus, the knowledge gap should widen as the campaign unfolds (Hypothesis 1b).
Figures 1a to 1c illustrate the expected increase in campaign-specific information
according to the level of voters’ general political information,4 at varying intensities of
media signals on issues. To the extent that these hypotheses imply the maintenance or
growth of the knowledge gap depending on the issue, and given that most campaigns are
too brief to allow low aware voters to “catch up” even when media signals are strong,
Figure 1. Expected evolution of information gains under various conditions of awareness and medi
a
coverage.
a
) High Coverage: b) Medium Coverage:
c) Low Coverage: d) Combined:
H = high awareness (GSI); MH = medium high awareness (GSI); ML = medium low awareness (GSI);
L = low awareness (GSI).
Time
CSI
L
ML
H
MH
Time
CSI
L
ML
H
MH
Time
CSI
L
ML
H
MH
Time
CSI
L
ML
H
MH
Election Campaigns as Information Campaigns 233
Hypothesis 1d predicts a widening gap between the information rich and the information
poor (Figure 1d).5
Information Intake and Voting Behavior
Given interest in the origins of the knowledge gap, it is remarkable that “little empirical
attempt has been made to investigate consequential aspects of the SES-based knowledge
gap” (Kwak, 1999). In election campaigns, the most relevant focus for assessing the
effects of the unequal penetration of information concerns what impact electoral informa-
tion has on the vote. There is reason to expect that an individual’s general stock of infor-
mation reflects a capacity both to absorb new information and to resist it (Converse, 1962;
McGuire, 1968; Zaller, 1989), and to expect that the acquisition of campaign-specific
information will, for some voters at least, have the potential to change voting intentions
(Holbrook, 1996).6 Thus, campaign-specific information gains should encourage vote
volatility (Hypotheses 2a), while voters’ general stock of information works in the
opposite direction (Hypothesis 2b).7
This perspective suggests that it is those who are moderately well informed who will
be the most susceptible to influence by information diffusion. New information must be
both important and intelligible to constitute a decisive element in evaluating a party or a
candidate. But the extent to which voters deem new information to be important and
intelligible may vary. For low aware voters, new pieces of information obviously form
important additions to their existing stock of political information, but their inability, or
disinclination, to interpret this information may limit its impact.8 The potential impact of
information gains among highly aware voters may also be limited but for different
reasons. For them, information gains are meaningful, but the additional information may
be neither sufficiently new, nor sufficiently important, to dislodge well-established vote
intentions. It is the moderately aware voters, those who are sufficiently experienced to
understand the information that is diffused but not sufficiently informed at the start to
avoid being “surprised,” who may be more susceptible to influence by new information.
The impact of campaign-specific gains on vote instability, we hypothesize, will be
positive and significant for moderately aware voters and nonexistent for those who lie at
the extremes of the information ladder (Hypothesis 2c).
Combining the knowledge gap and the Converse-Zaller model suggests a general
hypothesis about the global impact of information on electoral behavior among particular
sub-groups: The amount of vote change induced by information gains among different
groups depends on the combination of the size of these gains and their potential to change
electoral preferences. Vote volatility attributable to information intake should be greatest
among those with a high (but not very high) level of awareness (Hypothesis 2d), a group
for which information gains are sizeable and susceptibility to influence by information
diffusion significant.
Hypotheses 1a to 1d focus on global patterns of information penetration during the
campaign; they are tested using daily aggregate measures of GSI and CSI. Hypotheses 1a
and 1b determine whether the global stock of general and campaign-specific information
in the electorate evolves or remains stable during the campaign. Hypotheses 1c and 1d
concern the evolution of the knowledge gaps between poorly and well-informed voters
during the campaign, overall and according to varying intensities of media coverage on
issues. Hypotheses 2a to 2d, which examine the links between information intake and
individual voting behavior, are investigated using panel data that record voters’ electoral
preferences during the campaign and after the election. Multivariate tests of the models
234 Richard Nadeau et al.
are used to assess the impact of individual measures of GSI and information gains on
voters’ inclinations to report different electoral preferences before and after the election.
Electoral Campaigns as Information Campaigns: Methodological
Requirements
Analyzing information gains during an election campaign is methodologically demanding.
The research design must be able to (a) distinguish between GSI and CSI, (b) measure
information flows on a daily basis during the course of a campaign, and (c) connect cam-
paign information flows to the dynamics of individual vote choice. The 1997 Canadian
Election Study meets these challenges.9
The Canadian Election Study
The daily rolling cross-sectional design of the Canadian Election Study (CES) makes it
possible to measure fluctuations in campaign information intake by voters. The campaign
wave of the survey contains objective measures of both respondents’ GSI and CSI. The
panel component includes repeated measurements of campaign-specific information as
well as vote intention and reported vote. Consequently, we can evaluate the impact of infor-
mation gains on vote change. Along with standard SES indicators, the CES also includes,
significantly, a content analysis component that measures the strength of television media
signals on the three issues used to gauge campaign-specific knowledge for voters.
Two types of analyses will be performed from these data to test the two subsets of
hypotheses (1a to 1d and 2a to 2d) described above. First, daily averages for GSI and CSI
are used to perform aggregate analyses about the global patterns of information penetra-
tion during the 30 days of the campaign (Hypotheses 1a to 1d; see Note 9, Tables 1 and 3
and Figures 2 and 3). Second, individual-level data are analyzed to gauge the impact of
campaign-specific information gains on a respondent’s propensity to report after the elec-
tion a vote different from the intention expressed during the campaign (Hypotheses 2a to
2d; see Note 9 and Table 4).
General and Campaign-Specific Information
While general and campaign-specific information are conceptually distinct, the operation-
alization of this distinction is not an entirely settled question. For Converse (1962, p. 582),
the “ideal” measure of campaign-specific knowledge entails tapping “all the intake of
political information for our respondents during any political campaign” (p. 587). But
Converse ends up relying on the number of media used by respondents “as a rough index
of [information] intake during the campaign” (p. 588). As Zukin (1977) demonstrates, this
measure is problematic: “first [because] it is confined to the mass media, as interpersonal
information is not considered [and second because] simple exposure may not be tanta-
mount to information intake” (p. 546). Consequently, Macaluso (1977) favors a direct
measure of respondents’ information level, “a procedure that strikes us as eminently desir-
able when one is interested in examining the influence of information on other political
behaviors” (p. 256). Zaller fully exploits this idea, but his own operationalization of
information intake, based on various indicators including “the individual’s volunteered
likes and dislikes about the candidates,” is not entirely satisfactory (1989, p. 203).
Continuing in the spirit of these efforts, we suggest that campaign-specific informa-
tion can be operationalized in a manner that conceptually corresponds with the measures
Election Campaigns as Information Campaigns 235
of people’s general stock of information about politics. What is required is both a battery
of questions that objectively measure GSI and questions that capture respondents’ factual
information specific to an election campaign (CSI).
According to Zaller, an objective measure, first, is “essentially free of the inherent
subjectivity and response set problems that often beset such self-report measures as interest
or media use. Second, it is a measure of ideas that have actually gotten into people’s heads
rather than (as in the case of interest and media use) the mere propensity for ingesting
ideas. And third, neutral information consistently outperforms alternative measures in
predicting relevant criterion variables” (1989, p. 186).10
Our measure of voters’ general stock of political information (GSI) is based on
knowledge of four political actors: (a) the premier of the respondent’s own province, (b)
the first woman prime minister of Canada (Kim Campbell), (c) the name of Canada’s
finance minister (Paul Martin), and (d) the name of the U.S. president (Bill Clinton). Clas-
sifying voters according to their number of correct answers (0, 1 = 1, 2, 3, 4) distinguishes
respondents according to their level of GSI (low, medium low, medium high, high) and
divides the electorate into four groups of roughly equal size (23%, 29%, 27%, and 21%,
respectively, N = 2,957).
Campaign-specific information (CSI) is similarly measured with respondents’ factual
information about what issue positions three political parties took during the 1997 elec-
tion. The first question concerns the constitutional position taken by the Reform party, the
second dealt with the Conservative party’s promise to reduce taxes by 10%, and the third
asked about the New Democratic party’s promise to cut Canada’s unemployment rate in
half by 2001.
Responses to these questions are examined separately, but they are also grouped into
a single measure of campaign-specific information. Though not exhaustive, this battery of
questions is particularly useful here because, as the results of the content analysis show
(see below), each party’s issue position was covered with a different level of media inten-
sity. Consequently, it is possible to test the general knowledge gap hypothesis as well as
specific hypotheses concerning the link between the intensity of media coverage and the
expected evolution of the media coverage during the campaign.
These batteries of information questions capture voters’ level of awareness about pol-
itics in general and campaigns in particular (see Zaller, 1991, p. 134, Table 1).11 Because
political awareness is a relatively stable attribute, voters’ general stock of information
should move little during the course of a campaign. Neither political parties nor the media
have any incentive to diffuse general information that is unlikely to influence partisan
choices, and voters have no incentive to acquire general information that is of no utility to
decision making (Lupia & McCubbins, 2000). The case for campaign-specific informa-
tion is clearly quite different. Voters do have an interest in learning where parties stand on
issues, just as parties have an interest in informing voters about what they perceive to be
“winning” issues. Consequently, we should expect to find a significant accumulation of
campaign-specific information during the course of the election campaigns.
These auxiliary hypotheses, examined in Table 1 and Figure 2 with data from the
rolling cross-sectional survey, are broadly confirmed. The stability of the general stock of
information contrasts sharply with increases in campaign-specific information. These
results, first, empirically ground the conceptual distinction between general political and
campaign-specific information. Second, the overall increase in electoral information
during campaigns clearly demonstrates that when the information is sufficiently available
and relevant, some voters do bother to integrate it, possibly to clarify their voting choices
(Popkin, 1994; Holbrook, 1996).
236 Richard Nadeau et al.
Table 1
Evolution of general and campaign-specific information
Separate analyses Stacked analysis:
informationGSI CSI
Constant .47 (.02)** .25 (.01)** .47 (.01)**
Time .03 (.03) .10 (.02)** .03 (.02)
CSI −.23 (.02)**
CSI × Time .06 (.03)*
Adjusted R2.02 .45 .88
N30 30 60
Note. Entries are OLS regression coefficients with standard errors in
parentheses. General stock of information (GSI) is a 4-point scale (standard-
ized from 0 to 1) adding scores on four questions of factual knowledge (each
coded 1 if respondents gave the right answer and 0 if they did not or didn’t
know). Campaign-specific information (CSI) is the added score (standard-
ized from 0 to 1) on three questions related to campaign promises for which
the respondent mentioned the right party or mentioned it with another party
(coded 1), or did not mention the right party at all or didn’t remember (coded
0, including no opinions). Both variables are aggregate daily averages based
on the responses provided on a given day by a subsample of about 110
respondents; see Note 9). Time corresponds to days of the campaign in
ascending order, standardized to run from 0 (first day of the campaign) to 1
(last day of the campaign).
*p < .05; **p < .01 (two-tailed test).
Source: Canadian Election Study, 1997.
Figure 2. Evolution of CSI and GSI.
Source: Canadian Election Study, 1997.
0,15
0,20
0,25
0,30
0,35
50597
50797
50997
51197
51397
51597
51797
51997
52197
52397
52597
52797
52997
Date of interview
CSI
0,00
0,20
0,40
0,60
0,80
1,00
GSI
CSI
GSI
Election Campaigns as Information Campaigns 237
Content Analysis: Strong and Weak Signals
Television is a principal source of campaign information in Canada as elsewhere
(Ansolabehere, Behr, & Iyengar, 1993). Content analysis of television news thus
provides critical information for evaluating both the substance and amplitude of the
media signals broadcast to voters during a campaign. The content analysis of the 1997
Canadian election focused on the prime time nightly newscasts of Canada’s four
principal television networks.12 Two networks are public (Radio-Canada in Quebec
and CBC outside Quebec), and two are private (TVA in Quebec and CTV outside
Quebec).
The four networks’ late-night election news items were classified first, according to
whether the dominant angle dealt with the issues or with the electoral race.13 The analysis
then measured the intensity of media coverage for each of the three issues that are of
central concern. The aggregate results for the French (Radio-Canada and TVA) and
English-language networks (CBC and CTV) are presented in Table 2.14 The first part of
the table indicates the relative importance given to the issues and to the electoral race
(leaders’ tour, polls, strategies etc.) during the 1997 campaign. Clearly, as in other set-
tings (see Norris et al., 1999), the electoral race theme dominated campaign coverage
(76% of news items for both the French and English networks).15 The second part of the
table compares the media coverage of the three issues under study and shows that these
particular issues dominated the electoral coverage devoted to issues, in Quebec and the
rest of Canada. Equally important are the significant variations in the intensity of the
media signals for each issue: The network signals were “strong” for the Reform party’s
constitutional position, “medium” for the Conservatives’ tax position, and “weak” for the
NDP’s jobs policy.
Electoral Campaigns as Information Campaigns: The Findings
The results confirm three core findings. First, an overall knowledge gap does emerge
between the information rich and the information poor during the course of an election
campaign. Second, moderately aware voters are the most susceptible to influence by
information diffusion. And third, the greatest amount of information change attributable to
information intake occurs among those who occupy those rungs just below the top of the
awareness ladder.
Table 2
Content analysis of TV news
CBC/CTV SRC/TVA
Total news 417 (100) 319 (100)
Campaign 317 (76) 243 (76)
Issues 100 (24) 76 (24)
Specific issues: total 60 (100) 37 (100)
Unity (Reform) 37 (62) 24 (65)
Taxes (PC) 17 (28) 9 (24)
Unemployment (NDP) 6 (10) 4 (11)
Note. Values in parentheses are percentages.
Source: Canadian Election Study, 1997.
238 Richard Nadeau et al.
Campaign Knowledge Gap
Communications research expects that the overall knowledge gap between the informa-
tion rich and the information poor will widen during campaigns (Hypothesis 1d). And
Zaller’s work produces the expectation that knowledge gaps on specific issues will remain
the same when the media coverage of an issue is either strong or weak (Hypotheses 1a and
1c) but will widen when messages are of average intensity (Hypothesis 1b).
Figure 3d depicts the evolution of campaign-specific information across the campaign
for all three issues combined (using 5-day moving averages) with voters differentiated
according to their level of general political information (low, medium low, medium high,
high). The overall information gains are clearly most pronounced among voters who are
already well informed at the beginning of the campaign. Figures 3a to 3c illustrate the evo-
lution of the knowledge gap for each single issue. These curves correspond closely to the
stylized expectations presented earlier (Figures 1a to 1c). The gap between the groups
widens when it comes to knowledge of the Conservatives’ (medium signal) issue, whereas
the initial gap in knowledge of the New Democratic (weak signal) and Reform parties’
positions (strong signal) is sustained throughout the campaign.
Knowledge gap dynamics are examined more directly in Table 3. Here the dependent
variables represent the daily knowledge gaps between the most aware (medium high and
high) and least aware (medium low and low) groups of voters (see Note 4). Time is the inde-
pendent variable. The results confirm the widening of the knowledge gap (from .20 to .33 on
a 0–1 scale) (Hypothesis 1d), and this evolution is mainly driven by the moderately covered
Figure 3. Evolution of CSI by levels of GSI (5-day moving averages).
Source: Canadian Election Study, 1997.
0,00
0,20
0,40
0,60
0,80
1,00
7
9505
7
9805
7
9115
7
9415
7
9715
79025
7
9325
7
9625
7
9925
Date of interview
H
M
H
M
L
L
0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
79505
7
9805
79115
79415
79
715
79
025
7
9325
79625
799
25
Date of interview
CSI-PC
CSI-NDP
CSI
CSI-Reform
H
M
H
ML
L
a
) CSI-Reform: b) CSI-PC:
c) CSI-NDP: d) CSI:
0,00
0,05
0,10
0,15
0,20
0,25
0,30
0,35
05 579
509
97
7931
5
15 779
529
17
79525
25 979
Date of interview
H
M
H
M
L
L
0,00
0,10
0,20
0,30
0,40
0,50
0,60
0,70
7
95
0
5
79
8
05
7
91
1
5
79
4
15
7
97
1
5
79
0
25
7
93
2
5
7
96
2
5
7
99
2
5
Date of interview
H
M
H
M
L
L
Election Campaigns as Information Campaigns 239
taxation issue (Hypotheses 1a to 1c). The large increase in the knowledge gap in this latter
case (from .17 to .40 on a 0–1 scale) explains more than 80% of the increase in the total
information gap during the campaign.16 The distinctive pattern for taxation is confirmed by
the results in the second part of Table 3; the gap between the information rich and poor
increases significantly more for this issue than for national unity and unemployment.17
The knowledge gap hypothesis unequivocally does apply to the election setting.18
Zaller’s speculation (1989) about interaction effects between the intensity of media cover-
age and the knowledge gap is also clearly confirmed. Finally, the positive regression
coefficients (though not significant) for the issues of national unity and unemployment
signify that the growing knowledge gap reflects the cumulative impact of large increases
for moderately covered issues and small increases in the other cases. Overall, these results
provide further reason to be pessimistic about the potential for campaigns to increase the
quantity and the diversity of the information stock of the least aware voters.19
The Impact of Political Information
But are these information dynamics consequential for the stability of vote choice?20 Four
hypotheses explore that question. One possibility is that electoral volatility is positively
Table 3
Evolution of the knowledge gap: A macro-level analysis
Knowledge gap
CSI Unity (Reform) Taxation (PC) Unemployment
(NDP)
Constant .20 (.02)** .39 (.03)** .17 (.03)** .05 (.03)
Time .13 (.03)** .07 (.05) .23 (.05)** .06 (.05)
Adjusted R2.30 .03 .40 .02
N
30 30 30 30
Knowledge gap:a stacked analysis
PC vs. Reform PC vs. NDP Reform vs. NDP All
Constant .39 (.03)** .05 (.03) .05 (.03) .05 (.03)
Time .07 (.05) .06 (.04) .06 (.05) .06 (.05)
PC −.22 (.04)** .12 (.04)** — .12 (.04)**
PC × Time .16 (.07)*.17 (.06)** — .17 (.07)*
Reform — — .34 (.04)** .34 (.04)**
Reform × Time — — .02 (.07) .02 (.07)
Adjusted R2.50 .68 .82 .77
N
60 60 60 90
Notes. Entries are OLS regression coefficients with standard errors in parentheses.
The dependent variables are the aggregate daily knowledge gaps (see Note 9) between the most
aware (high and medium high) and the least aware voters (medium low and low). The definition for
time is provided in Table 1.
*p < .05; **p < .01 (two-tailed test).
Source: Canadian Election Study, 1997.
240 Richard Nadeau et al.
related to campaign-specific information (Hypothesis 2a) and negatively related to the
general stock of information (Hypothesis 2b). Another possibility is that information gains
will have a greater impact on voters with moderate levels of general political information
(Hypothesis 2c) and that the amount of vote change attributable to information gains is
highest for the group just below the top of the awareness ladder (Hypothesis 2d).
The results of multivariate tests of those hypotheses are presented in Table 4. Here,
the dependent variable is vote volatility, which takes the value of 1 if respondents’
reported vote in the post-election wave of the survey is different from their declared vote
intention when interviewed during the campaign. The model in column 1 includes all
voters. It links respondents’ vote volatility to nine variables—the elapsed time between
the campaign and the post-electoral interviews, the strength of partisan attachment, the
preelection vote choice (supporting a losing party or not), voters’ levels of general and
campaign-specific information, and four other measures of information reception:
education, campaign interest, television news watching, and newspapers reading.21
The results confirm the impact of the first five variables on the volatility of vote
choice; all have the expected sign and are statistically significant.22 The time between
interviews, the strength of partisan attachment, being on the losing side, and the level of
general political information all decrease the probability of change in choice during the
campaign. Note that the acquisition of campaign-specific information is the only measure
of ability and motivation positively and significantly linked to electoral instability. The
hypothesis that general political information has a negative effect on electoral change and
an indirect positive effect (through information gains) on this same variable is thus
confirmed.
The hypothesis that the impact of information gains will be more important for voters
with average levels of GSI is also confirmed. The weak impact of information gains on
electoral change for respondents as a whole is attributable to the different effect that this
variable has on those who are moderately well informed compared to others. Information
gains do not induce a questioning of choices either among those who are too poorly
informed at the beginning to give meaning to this new information or among those who
already possess too much to be influenced by it (Table 4, columns 2, 4, and 6). For those
occupying the middle ground, however, the impact of information gains is significant
(at .057 or better) and strong (see columns 3, 4, and 7)23. These voters, having acquired
information on taxation, economic, and constitutional positions of the parties under study,
have a probability of changing their vote that is roughly 23 percentage points higher than
someone from the same group but who remained insulated from campaign information flows.
The impact of information gains on vote change can be estimated under different con-
ditions of awareness by establishing, for each group, just how much vote volatility is
induced by information intake. These estimates are computed from the logistic regression
coefficients in Table 4 and the overall information gains for each subgroup.24 These calcu-
lations show that the impact of gains in campaign-specific information on “explained”
vote change is very limited among groups at the extremities of the information ladder
(respectively 0 and −1.0 percentage points; both results are derived from nonsignificant
coefficients). It is more important for the groups in the middle of the scale (respectively 1
and 2.5 percentage points). And, as expected, the impact of information intake is the
strongest among those who display a medium high level of awareness.
The results suggest that there are significant gaps or thresholds to consider. The first,
the knowledge gap, appears to distinguish those who learn (high and medium high levels
of awareness) from those who learn little or nothing (medium low and low levels of
awareness) during campaigns. The second distinguishes those who are susceptible to
241
Table 4
Regression analysis of the impact of information gains on vote change during the 1997 campaign
Levels of GSI Levels of GSI
All Low Medium low Medium high High L and H ML and MH
Constant −2.19 (.39) −4.78 (1.11)** −1.42 (.82) −2.03 (.73)** −3.69 (1.04)** −4.55 (.81)** −1.75 (.55)**
Time .03 (.01)** .07 (.02)** .02 (.01) .02 (.01)*.03 (.01)** .04 (.01)** .02 (.01)**
Party ID strength −1.13 (.17)** −2.58 (.52)** −1.06 (.35)** −.82 (.28)** −1.20 (.37)** −1.63 (.28)** −.92 (.22)**
Loser .39 (.17)*1.85 (.48)** −.19 (.34) .30 (.29) .37 (.39) .78 (.29)** .09 (.22)
Schooling .16 (.44) 1.85 (1.26) −.76 (.93) −.38 (.76) 1.12 (.97) 1.43 (.76) −.53 (.58)
Interest .58 (.43) .77 (1.07) 1.18 (.83) .61 (.80) .50 (1.15) .37 (.75) .88 (.57)
TV news .01 (.37) −1.24 (1.03) −.70 (.72) .21 (.66) .06 (.86) −.16 (.63) −.19 (.48)
Newspapers −.38 (.34) 1.58 (.99) −.41 (.69) −.67 (.55) −.14 (.83) .38 (.61) −.54 (.43)
GSI −.68 (.25)** ——————
Information gains .86 (.35)*1.32 (1.12) 1.35 (.71) 1.26 (.55)*−1.11 (.96) −.31 (.68) 1.31 (.43)**
Pseudo-R2.12 .39 .10 .10 .11 .22 .09
N1,304 204 317 405 378 582 722
Note. Entries are logistic regression coefficients with standard errors in parentheses. A dummy variable (not shown) is included in each of the last two regressions
(GSI-L and GSI-ML, respectively), the logistic regression coefficients are 1.24 for GSI-L (SE = .35) and −.14 for GSI-ML (SE = .22). Vote change, the dependent
variable, is coded 1 if respondents changed their vote between the campaign interview and the post-election interview and 0 if their actual vote was the same as their
vote intention. Information gains is the difference (standardized from 0 to 1, with negative values recoded 0) between the CSI index and the same index as measured
in the post-election survey using the same three questions. Schooling is an 11-point scale running from 0 (no schooling) to 1 (professional degree or PhD). Party ID
strength is a dummy variable coded 1 if respondent feels a very strong or a fairly strong identification with a political party and 0 otherwise. Loser takes the value of
1 for respondents supporting a losing party in the campaign survey and 0 otherwise. Interest, TV news, and newspapers are 10-point scales running from 0 (no inter-
est or no attention) to 1 (a great deal of interest or attention) using questions measuring respondents’ interest in the election campaign, attention to news about the
election campaign on TV, or attention to news about the election campaign in the newspapers.
*p < .05; **p < .01 (two-tailed test).
Source: Canadian Election Study, 1997.
242 Richard Nadeau et al.
influence by information gains (medium high and medium low levels of awareness) from
those who are not (high and low levels of awareness). And the third differentiates those for
whom information gains had a sizeable impact (medium high levels of awareness) from
those for whom the impact of information intake was more limited, either because of a
lack of susceptibility to new information (low and high levels of awareness) or because of
limited information gains (medium low levels of awareness).
Evidently, the group that stands to benefit the most from a more equal penetration of
campaign information is the one that occupies the rung just above the bottom of the
awareness ladder. That they can make significant information gains when the media signal
is strong (the unity issue) clearly indicates that efficient information campaigns do matter.
Discussion and Conclusions
Viewing election campaigns as information campaigns is methodologically demanding.
To assess both global patterns of information penetration and the impact of information
gains on individual vote volatility requires a multimethod approach using individual-level
and aggregate data. The aggregate findings offer a useful “big picture” of information
dissemination during campaigns but they are based on a limited number of cases (see
Note 9), and they are not suitable for complex modeling. Individual-level data with larger
samples are more flexible, but do not permit the repeated measurements necessary to
account for electoral information dynamics. One promising avenue for future research is
to exploit more fully the complementary strengths of both of these kinds of data.
Viewing election campaigns as information campaigns not only encourages a broader
research agenda, but it also invites a more expansive consideration of the theoretical and
empirical contributions of researchers who examine these questions from different
vantage points. Combining Converse and Zaller’s work on communication flows with
theories of the knowledge gap certainly provides both a richer and clearer understanding
of campaign information dynamics.
Our major findings confirm Chaffee and Kanihan’s conjecture that “the concept of
knowledge gap . . . is applicable to political knowledge as well” (1997, p. 426). They
also document how, and when, campaign knowledge gaps occur. And there is suggestive
evidence to explain “the gap between knowledge and behavior,” which is, according to
Hornik (1989, p. 113), “the central theoretical problem in the field of purposive commu-
nication.”
Different segments of the electorate evidently do respond differently to a certain type
of campaign-specific information. Those at the top and bottom of the information ladder
are similarly unresponsive to new information about issues, but the reasons for their non-
responsiveness may well be quite different. Beyond a certain threshold, exposure to
parties’ positions is not sufficient to modify well-documented political choices. It is not
that highly aware voters are immune to all kinds of information flows, perhaps, but rather
that parties’ positions per se do not form the type of rich, complex, and “surprising” pieces
of information that encourage the reconsideration of their partisan choices.
Not all campaign information flows, of course, or media signals are about issues.
Poorly informed voters may be more susceptible to cognitively cheap and simple informa-
tion about leaders. Moderately well-informed voters may be more responsive to information
about issues. And well-informed voters are more inclined to respond to the more complex
information about arguments because that factual information is necessary to construct
arguments (Lupia & McCubbins, 2000, pp. 53–54). If election campaigns carry heteroge-
neous messages about issues, leaders, and arguments, then future research needs to
Election Campaigns as Information Campaigns 243
establish whether different types of information have differential impacts on different
segments of the electorate.
From this perspective, our demonstration that the notion of the vulnerability of the
poorest among the information poor to information flows is unfounded is perhaps a first
step. That conclusion was based primarily on macro data showing that individuals with a
low level of awareness generally made few information gains while generally displaying
similar levels of volatility to those observed in adjacent groups in terms of sophistication.
Thus, some conjecture that the information poor would be extremely vulnerable to new
information if only that information were to reach them (Converse, 1962; Zaller, 1992).
Our results, however, suggest that this hypothesis may be based on spurious correlation. If
making information gains does not lead low aware voters to change their mind, the impli-
cation is that changing the attitudes of the citizens at the bottom of the information scale is
not merely a problem of reaching these citizens. To draw on Lupia and McCubbins’s
(2000, p. 52) useful distinction between knowledge and information, these voters suffer
from lack of knowledge, which prevents them of being able to take advantage of available
information. In this sense, this group suffers from a “knowledge gap” as well as an “infor-
mation gap” during campaigns. That characterization does not apply to those located one
step above the bottom of the awareness scale. Their sensibility to new information
signifies that they suffer less from a lack of knowledge than from an “information” gap.
Thus, it is the individuals in this group who might benefit most from an equalization of the
information gains during campaigns. For more complex pieces of information (such as
arguments about policies), the threshold between those who suffer or not from the
“knowledge” gap is likely on a higher rung of the awareness scale ladder.
These distinctions have implications both for theories of persuasion (Mutz, Sniderman,
& Brody, 1996; O’Keefe, 2002) and of information campaigns (Salmon, 1989). The
consensus is that persuasion occurs when the right message is sent to the right person. Our
results show, however, that efficient strategies of communications should also consider
the capabilities of the recipients of this information. Disseminating complex information
to individuals who are equipped neither to absorb nor to interpret it is suboptimal. One
long-run strategy might be to augment their capacity to take advantage of complex infor-
mation. A short-run alternative is to present them with simpler and more tailored pieces of
information. At the opposite end of the sophistication spectrum, though, providing simple
information is also suboptimal. But between these two groups lies a large and receptive
audience that evidently does benefit from learning simple pieces of factual knowledge.
Moreover, the fact that many of those located just above the bottom of the awareness scale
can make significant information gains when media signals are loud (see Figure 3a)
provides some comfort to those who believe that (good) information campaigns matter.
These observations resonate with the debates about what voters should know to vote
competently (Lupia & McCubbins, 1998, 2000). As Bartels (1996) and Kuklinski and
Quirk (2000) show, judgmental shortcuts do not necessarily pave the way to enlightened
choices. If heuristics rules are an inadequate substitute for such basic information as
knowledge of parties’ positions, then our evidence concerning the distribution and the
incidence of information gains is particularly relevant. And, if the explicit objective of
information campaigns is to reach the least favorable segments of the population (Salmon,
1989), then our results raise trenchant questions about the efficiency of election
campaigns as information campaigns.
Increasing or persistent knowledge gaps are undesirable outcomes from a variety of
vantage points. The worry that a “relative deprivation of knowledge may lead to a relative
deprivation of power” is a concern shared by communications researchers and political
244 Richard Nadeau et al.
scientists alike (Donoghue, Tichenor, & Olien, 1975, p. 4). A common understanding of
the benefits of widespread political information for the workings of democracy is also
widely shared. Gaziano’s (1984, p. 556) concern that “inequalities in knowledge . . . run
counter to the fundamental assumption that an informed citizenry is essential to democ-
racy” moves from Delli Carpini and Keeter’s (1996, p. 8) assumption that “information is
the currency of citizenship.”
These considerations underscore the dual nature of campaigns and provide broader
meaning to the question “Do campaigns matter?” Campaigns do matter from a purely par-
tisan perspective, and conceptualizing election campaigns as information campaigns is
important for understanding how they matter. Campaigns also matter from the vantage
point of democratic citizenship, and interpreting electoral campaigns as information cam-
paigns is vital for understanding the dynamics that produce knowledge gaps. The criteria
for judging campaigns clearly reflect these dual perspectives. Staging colorful campaign
events, or depriving voters of pieces of information, might be wise strategies for media
outlets intent on winning the ratings contest or for parties intent on winning the agenda-
setting battle. Those very same strategies, however, turn out to be obstacles to voters’
ability to make enlightened choices (Page, 1978, p. 187). Political ads exemplify the prob-
lem. They are tailored to attract low information voters’ attention and their informational
content is not negligible (West, 2005). The ultimate goal of political ads is much more
about “priming” and “framing” issues than about informing voters (Kuklinski et al., 2000;
Norris et al., 1999). Extending the length of political campaigns is also an option. As
David Moore argued more than 20 years ago, “longer campaigns may indeed be doing
their voters a service by providing a period of time that results in a more equitable oppor-
tunity for all voters to learn what they need before voting” (1987, p. 198). Unfortunately,
the length of electoral campaigns became shorter rather than longer in many democracies
over the past decades (Nevitte et al., 2000). Campaigns matter, in our view, because they
are unique opportunities to “lift the bottom” (Delli Carpini & Keeter, 1996, pp. 280, 287)
and to bring voters back in. The challenge is to determine why most campaigns fall short
of these objectives and why campaigns are rarely great educational experiences.
Notes
1. This measure resembles that used by Berelson and his collaborators (1954, p. 308) and
indicates “voter enlightenment” for Chaffee et al. (1994, p. 306).
2. Lupia and McCubbins (2000, p. 52) argue that knowledge refers to the ability to make
predictions, whereas information refers to data. Thus, the knowledge gap might be better
characterized as an information gap, resulting from structural knowledge gaps among groups.
3. Tichenor et al. (1970, p. 170) suggest that “media coverage tends to wane before the
knowledge gap closes,” which prompted Moore (1987, p. 198) to propose longer campaigns to give
low-educated voters more time to learn about issues.
4. Given that parties’ position on issues represents a type of information that is neither exceed-
ingly simple (such as recognizing leaders) nor particularly complex (arguments justifying a policy),
the main threshold in terms of information gains should be between voters with high and medium
high levels of awareness and those with medium low and low levels of awareness.
5. The generality of Hypotheses 1a and 1d rests on the assumption that the circumstances
conducive to decreasing knowledge gaps (e.g., “ceiling effects”; see Ettema & Kline, 1977) are
infrequent during electoral campaigns.
6. For some voters, new information may be decisive. Conservative voters who are opposed to
tax cuts but unaware that the party is proposing a 10% tax cut may change their vote upon receiving
that “new” information.
Election Campaigns as Information Campaigns 245
7. Concerning the inertia effect produced by the “mass of stored information about politics”
possessed by individuals, Converse wrote (1962, p. 583): “From a psychological point of view, any
cognitive modeling of these propositions could rapidly capitalize upon the many hypotheses
concerning the structure and interconnectedness as the mass of information increases.”
8. This is consistent with Khazee’s (1981, p. 517) observation that for the information poor,
“information enters [their] cognitive space having undergone little or no interpretation.”
9. The CES data were collected by the Institute for Social Research at York University in
Toronto. The campaign-period telephone survey began on April 27, the day the election was called,
and ended on June 1, the last day of the campaign; 3,949 interviews were conducted, approximately
110 per day of the campaign, using a rolling cross-sectional strategy. The response rate was 59%.
The post-election survey recontacted respondents from the campaign survey in the 8 weeks after the
election; 80% of the campaign survey respondents were re-interviewed (n = 3,170). The questions
on the New Democratic party’s and the Reform party’s issue positions were introduced into the
campaign wave questionnaire on May 3; thus, we analyze the last 4 weeks of the campaign (from
May 3 to June 1). The campaign response rate and the level of panel attrition in the study met usual
standards. A detailed description of the methodological guidelines followed to ensure the quality of
the data can be found in Northrup (1998).
10. If factual measures of information are superior to indicators of cognitive engagement, then
our indicator of information gains should be more strongly linked to vote change than variables like
education, campaign interest, or media exposure. The superiority of our variable of general informa-
tion about politics (GSI) as a predictor of campaign-specific information (CSI) as well as the
dominance of our measure of information gains as a predictor of vote change indicates the validity
of these information measures.
11. A regression analysis (OLS) using level of campaign-specific information as the dependent
variable and the usual determinants of knowledge as independent variables shows that the GSI
clearly outperforms education, interest in the campaign, television news watching, and newspaper
reading in predicting CSI (detailed results are available upon request).
12. The intercoder reliability level is .90. All content analysis material is available upon
request.
13. Reports of the “electoral race” variety are those that give priority to campaign activities,
party leaders or candidates, polls, election advertisements, races in certain regions, or other features
of the contest between the political parties. News of the “issue” variety includes the themes, debates,
and positions defended or promoted by the various parties or by voters.
14. The results show a large convergence between the broadcast patterns of the networks.
15. News about the electoral race rarely contained information about issues. Moreover, the
extremely rare traces of information about issues in these stories are distributed in much the same
way as for the news that mainly focuses on issues.
16. An analysis of specific groups shows that, as expected, the knowledge gaps increased sig-
nificantly more between than among low (low and medium low) and high (medium high and high)
groups. Following an anonymous reviewer’s suggestion, we have checked if our key findings con-
cerning the global patterns of information penetration are the same for strong and weak partisans.
The coefficients among both groups are very similar for the models concerning the evolution of
aggregate measures of GSI, CSI, and the knowledge gaps during the campaign. Detailed results are
available from the authors upon request.
17. We discount the possibility that low aware voters make more information gains than highly
aware voters in the case of intensively covered issues. The coefficient tapping the relationship
between time and knowledge is higher for the highly aware group (.17 as compared to .10 for the
low aware group). These data suggest that decreasing knowledge gaps are rare and probably
confined to “easy” information (name recognition) and unusually long campaigns.
18. Individual-level data reanalysis shows that the impact of GSI on CSI is significantly higher
among respondents interviewed later in the campaign. These results (available upon request) also
reveal that the impact of GSI on the knowledge gaps became significantly higher only for the
moderately covered issue of taxation.
246 Richard Nadeau et al.
19. Information gains about the widely covered issue of national unity account for 73% of the
learning gains among the least aware voters and for 31% among the more knowledgeable voters
(55% and 36% for the intermediate groups, respectively). Thus, the information gains among the
least informed are limited in quantity and diversity.
20. Sixteen percent of respondents changed their vote between the pre- and post-election waves
of the survey. The distributions for the vote variables before and after the election (N = 1,410) corre-
spond closely to the official election results and are consistent with accounts of the dynamics of the
1997 campaign (Nevitte et al., 2000).
21. We use logistic estimations since the dependent variable is dichotomous (see Menard,
1995). The operationalizations of the time variable and the pre-election choice variable are aimed at
capturing a pattern, common to post-election surveys, where the margin of victory for the winning
party is overstated because of the propensity of losers to report having voted for the winning party.
We thank an anonymous reviewer for her useful suggestions on this point.
22. The propensity to gain information is unrelated to initial levels of CSI. Regressions for
those with and without initial campaign-specific information show that the impact of information
gains on vote change in the two groups is statistically the same as the coefficient for the entire
sample.
23. To test for this, we ran a regression with all four groups including an interactive term
designed to assess the differential impact of the information gains variable on the vote stability of
the moderately sophisticated voters (middle groups by information gains). The results show that the
coefficient for the interactive variable is correctly signed, substantially meaningful, and statistically
significant. Moreover, the coefficients for the information intake variable are not statistically
distinguishable within the middle and the extreme groups.
24. We multiply the logistic regression coefficients in Table 4 by the average information gains
during the campaign for each group defined in terms of their level of awareness (respectively 0, .04,
.11, and .12 on a 0–1 scale for the low, medium low, medium high and high groups).
References
Ansolabehere, S., Behr, R., & Iyengar, S. (1993). The media game: American politics in the
television age. New York: Macmillan.
Bartels, L. M. (1996). Uninformed votes: Information effects in presidential elections. American
Journal of Political Science, 40, 194–230.
Berelson, B. R., Lazarsfeld, P. F., & McPhee, W. N. (1954). Voting: A study of opinion formation in
a presidential campaign. Chicago: University of Chicago Press.
Chaffee, S. H., & Kanihan, S. F. (1997). Learning about politics from the mass media. Political
Communication, 14, 421–430.
Chaffee, S. H., Zhao, X., & Leshner, G. (1994). Political knowledge and the campaign media of
1992. Communication Research, 21, 305–324.
Converse, P. E. (1962). Information flow and the stability of partisan attitudes. Public Opinion
Quarterly, 26, 578–599.
Delli Carpini, M. X., & Keeter, S. (1996). What Americans know about politics and why it matters.
New Haven, CT: Yale University Press.
Donoghue, G. A., Tichenor, P. J., & Olien, C. N. (1975). Mass media and the knowledge gap:
A hypothesis reconsidered. Communication Research, 2, 3–23.
Ettema, J. S., & Kline, F. G. (1977). Deficits, differences, and ceilings: Contingent conditions for
understanding the knowledge gap. Communication Research, 2, 179–202.
Gaziano, C. (1984). Neighborhood newspapers, citizen groups and public affairs knowledge gaps.
Journalism Quarterly, 61, 556–566.
Gaziano, C., & Gaziano, E. (1996). Theories and methods in knowledge gap research since 1970. In
M. B. Salwen & D. W. Stacks (Eds.), An integrated approach to communication theory and
research (pp. 127–143). Mahwah, NJ: Erlbaum.
Holbrook, T. (1996). Do campaigns matter? Thousand Oaks, CA: Sage.
Election Campaigns as Information Campaigns 247
Holbrook, T. (2002). Presidential campaigns and the knowledge gap. Political Communication, 19,
437–454.
Hornik, R. C. (1989). The knowledge-behavior gap in public information campaigns: A develop-
ment communication view. In C. T. Salmon (Ed.), Information campaigns (pp. 113–138).
Newbury Park, CA: Sage.
Khazee, T. A. (1981). Television exposure and attitude change. Public Opinion Quarterly, 45, 507–518.
Kuklinski, J. H., & Quirk, P. J., (2000). Reconsidering the rational public: Cognition, heuristics, and
mass opinion. In A. Lupia, M. D. McCubbins, & S. L. Popkin, (Eds.), Elements of reason
(pp. 153–182). Cambridge: Cambridge University Press.
Kuklinski, J. H., Quirk, P. J., Jerit, J., Schwieder, D., & Rich, R. (2000). Misinformation and the
currency of citizenship. Journal of Politics, 62, 790–816.
Kwak, N. (1999). Revisiting the knowledge gap hypothesis: Education, motivation, and media use.
Communication Research, 26, 385–413.
Lupia, A., & McCubbins, M. D. (1998). The democratic dilemma: Can citizens learn what they need
to know? Cambridge: Cambridge University Press.
Lupia, A., & McCubbins, M. D. (2000). The institutional foundations of political competence: How
citizens learn what they need to know. In A. Lupia, M. D. McCubbins, & S. L. Popkin, (Eds.),
Elements of reason (pp. 47–66). Cambridge, England: Cambridge University Press.
Macaluso, T. F. (1977). Political information, party identification and voting defection. Public
Opinion Quarterly, 41, 255–260.
McGuire, W. J. (1968). Personality and susceptibility to social influence. In E. F. Borgatta & W. W.
Lambert, (Eds.), Handbook of personality theory and research, (pp. 1130–1187). Chicago:
Rand-McNally.
Menard, S. (1995). Applied logistic regression analysis. Thousands Oaks, CA: Sage.
Mendelsohn, M., & Cutler, F. (2000). The effect of referendums on democratic citizens: Information,
politicization, efficacy and tolerance. British Journal of Political Science, 30, 685–698.
Mondak, J. J., & Anderson, M. R. (2004). The knowledge gap: A reexamination of gender-based
differences in political knowledge. Journal of Politics, 66, 492–512.
Moore, D. W. (1987). Political campaigns and the knowledge-gap hypothesis. Public Opinion
Quarterly, 51, 186–200.
Mutz, D. C., Sniderman, P. M., & Brody, R. A. (Eds.). (1996). Political persuasion and attitude
change. Ann Arbor: University of Michigan Press.
Nevitte, N., Blais, A., Gidengil, A., & Nadeau, R. (2000). Unsteady state: The 1997 Canadian
federal election. New York: Oxford University Press.
Norris, P., Curtice, J., Sanders, D., Scammel, M., & Semetko, H. A. (1999). On message:
Communicating the campaign. London: Sage.
Northrup, D. (1998). The 1997 Canadian Election Survey: Technical documentation. Toronto:
Institute for Social Research, York University.
O’Keefe, D. J. (2002). Persuasion: Theory and research (2nd ed.). Thousand Oaks, CA: Sage.
Page, B. I. (1978). Choices and echoes in presidential elections: Rational man and electoral
democracy. Chicago: University of Chicago Press.
Popkin, S. L. (1994). The reasoning voter: Communication and persuasion in presidential
campaigns, (2nd ed.). Chicago: University of Chicago Press.
Price, V., & Zaller, J. (1993). Who gets the news? Alternative measures of news perceptions and
their implications for research. Public Opinion Quarterly, 57, 133–164.
Salmon, C. T. (Ed.). (1989). Information campaigns. Newbury Park, CA: Sage.
Tichenor, P. J., Donoghue, G. A., & Olien, C. N. (1970). Mass media flow and differential growth in
knowledge. Public Opinion Quarterly, 34, 159–170.
Viswanath, K., & Finnegan, J. R. Jr. (1996). The knowledge gap hypothesis: Twenty-five years
later. In B. R. Burleson (Ed.), Communication yearbook 19 (pp. 187–227). Thousand Oaks,
CA: Sage.
West, D. M. (2005). Air wars: Television advertising in election campaigns, 1952–2004 (4th ed.).
Washington, DC: Congressional Quarterly Press.
248 Richard Nadeau et al.
Zaller, J. (1989). Bringing Converse back in: Modeling information flow in political campaigns. In
J. Stimson (Ed.), Political analysis (pp. 181–234). Chicago: University of Chicago Press.
Zaller, J. (1991). Political awareness, elite opinion leadership, and the mass survey response. Social
Cognition, 8, 125–153.
Zaller, J. (1992). The nature and origins of mass opinion. Cambridge, England: Cambridge
University Press.
Zukin, C. (1977). A reconsideration of the effects of information on partisan stability. Public
Opinion Quarterly, 41, 244–254.1
